Treasure Chest of Six Sigma Growth Meths., Tools, Best Practs. - L. Hambleton (Pearson, 2008)

Treasure Chest of Six Sigma Growth Methods, Tools, and Best Practices A Desk Reference Book for Innovation and Growth Lynne Hambleton

PRENTICE HALL UPPER SADDLE RIVER, NJ • BOSTON • INDIANAPOLIS • SAN FRANCISCO NEW YORK • TORONTO • MONTREAL • LONDON • MUNICH • PARIS • MADRID CAPETOWN • SYDNEY • TOKYO • SINGAPORE • MEXICO CITY

The authors and publisher have taken care in the preparation of this book, but make no expressed or implied warranty of any kind and assume no responsibility for errors or omissions. No liability is assumed for incidental or consequential damages in connection with or arising out of the use of the information or programs contained herein. The publisher offers excellent discounts on this book when ordered in quantity for bulk purchases or special sales, which may include electronic versions and/or custom covers and content particular to your business, training goals, marketing focus, and branding interests. For more information, please contact: U. S. Corporate and Government Sales (800) 382-3419 [email protected] For sales outside the U. S., please contact: International Sales [email protected] This Book Is Safari Enabled The Safari® Enabled icon on the cover of your favorite technology book means the book is available through Safari Bookshelf. When you buy this book, you get free access to the online edition for 45 days. Safari Bookshelf is an electronic reference library that lets you easily search thousands of technical books, find code samples, download chapters, and access technical information whenever and wherever you need it. To gain 45-day Safari Enabled access to this book: • Go to http://www.prenhallprofessional.com/safarienabled • Complete the brief registration form • Enter the coupon code JUKU-BLUI-EID7-3NJU-CV7J If you have difficulty registering on Safari Bookshelf or accessing the online edition, please e-mail [email protected].

Publisher John Wait Editor in Chief Karen Gettman Acquisitions Editor Bernard Goodwin Editorial Assistant Michelle Housley Development Editor George E. Nedeff Marketing Manager Curt Johnson Publicist Andrea Bledsoe Cover Designer Louisa Adair Managing Editor Gina Kanouse Copy Editor Language Logistics, LLC Graphic Artist Laura Robbins Senior Indexer Cheryl Lenser Compositor Eric S. Miller Manufacturing Buyer Dan Uhrig

Visit us on the Web: www.phptr.com Library of Congress Cataloging-in-Publication Data: Hambleton, Lynne. Treasure chest of six sigma growth methods, tools & best practices : a desk reference book for innovation and growth / Lynne Hambleton. p. cm. Includes bibliographical references and index. ISBN 978-0-13-230021-6 (pbk. : alk. paper) 1. Six sigma (Quality control standard) 2. Strategic planning. 3. Business planning. 4. Management. I. Title. HD62.15.H354 2008 658.4’013—dc22 2007016916 Copyright © 2008 Pearson Education, Inc. All rights reserved. Printed in the United States of America. This publication is protected by copyright, and permission must be obtained from the publisher prior to any prohibited reproduction, storage in a retrieval system, or transmission in any form or by any means, electronic, mechanical, photocopying, recording, or likewise. For information regarding permissions, write to: Pearson Education, Inc. Rights and Contracts Department One Lake Street Upper Saddle River, NJ 07458 Trademarks All terms mentioned in this book that are known to be trademarks or service marks have been appropriately capitalized. Prentice Hall publishing cannot attest to the accuracy of this information. Portions of the input and output contained in this publication/book are printed with permission of Minitab Inc. Minitab™ and the Minitab logo® are registered trademarks of Minitab Inc. Minitab QUALITY COMPANION™ and the QUALITY COMPANION logo™are registered trademarks of Minitab Inc. Included in this book are Crystal Ball™ screen captures courtesy of Decisioneering, Inc. Use of a term in this book should not be regarded as affecting the validity of any trademark or service mark. 10-Digit ISBN 0-132-30021-4 13-Digit ISBN 978-0-132-30021-6 Text printed in the United States on recycled paper at R.R. Donneley & Sons in Crawfordsville, Indiana. First printing, July 2007

I dedicate this book first and foremost to my loving husband, Bill, and our two wonderful sons, Corbin and Garrett. I also dedicate this book to Skip and Kathy Creveling, whose friendship and support are invaluable gifts; and to Janet Nelson, a fellow consultant and CSSBB, who is courageously and gracefully battling cancer.

This page intentionally left blank

Preface .......................................................................................................xv Introduction Different Methods for Different Purposes..................1

Part I

Six Sigma Methodology Overview: Choosing the Right Approach to Address the Requirements

Section 1 Section 2 Section 3 Section 4

Define-Measure-Analyze-Improve-Control (DMAIC)....13 Lean and Lean Six Sigma ....................................................29 Design for Six Sigma (DFSS) ...............................................45 Six Sigma for Marketing (SSFM) ........................................67

Part II

Six Sigma Tools and Techniques: Choosing the Right Tool to Answer the Right Question at the Right Time

The Six Sigma Encyclopedia of Business Tools and Techniques .....115 Summary Tool Matrix ............................................................................115 A Activity Network Diagram (AND) - 7M Tool . . . . . . . . . . . . . 127 Affinity Diagram - 7M Tool..........................................................136 Analysis of Variance (ANOVA)...................................................142 Arrow Diagram..............................................................................159 B Benchmarking ................................................................................160 Box Plots—Graphical Tool ...........................................................165 Brainstorming Technique .............................................................168 C Capability Analysis .......................................................................173 Cause-and-Effect Diagram - 7QC Tool .......................................173 Cause-and-Effect Prioritization Matrix ......................................188 Cause and Prevention Diagram ..................................................198 Checklists - 7QC Tool....................................................................204 Conjoint Analysis ..........................................................................207 Control Charts - 7QC Tool............................................................217 Cost/Benefit Analysis...................................................................238

v

vi

Treasure Chest of Six Sigma Growth Methods, Tools & Best Practices

D

F

G

H

I K M

N P

Q

Critical Path Method (CPM) ........................................................242 Critical-to-Quality (CTQ) .............................................................242 Data Collection Matrix..................................................................248 Design of Experiment (DOE) .......................................................250 Dotplot ............................................................................................280 Failure Modes and Effects Analysis (FMEA).............................287 5-Whys ............................................................................................305 Fault Tree Analysis (FTA) .............................................................309 Fishbone Diagram - 7QC Tool .....................................................316 Flowchart - 7QC Tool ....................................................................316 Gantt Chart.....................................................................................317 GOSPA (Goals, Objectives, Strategies, Plans and Actions)......320 Graphical Methods........................................................................323 Histogram - 7QC Tool ...................................................................330 House of Quality (HOQ) ..............................................................335 Hypothesis Testing ........................................................................335 Interrelationship Diagram - 7M Tool ..........................................369 KJ Analysis .....................................................................................375 Market Perceived Quality Profile (MPQP) ................................390 Matrix Diagrams - 7M Tool ..........................................................399 Measurement System Analysis (MSA) .......................................412 Monte Carlo Simulation ...............................................................431 Multi-vari Chart.............................................................................439 Normal Probability Plot ...............................................................444 Pareto Chart - 7QC Tool................................................................445 PERT (Program Evaluation and Review Technique) Chart ....453 Poka-Yoke .......................................................................................462 Porter’s 5 Forces ............................................................................464 Prioritization Matrices - 7M Tool ................................................470 Process Capability Analysis .........................................................486 Process Decision Program Charts (PDPC) - 7M Tool ...............515 Process Map (or Flowchart) - 7QC Tool .....................................522 Pugh Concept Evaluation and Selection....................................534 Quality Function Deployment (QFD).........................................543

vii

R

S

T V W Y

RACI Matrix (Responsible, Accountable, Consulted, Informed)........................................................................................554 Real-Win-Worth (RWW) Analysis...............................................560 Regression Analysis ......................................................................571 Risk Mitigation Plan......................................................................601 Rolled Throughput Yield..............................................................610 Run Chart - 7QC Tool....................................................................611 7M - Seven Management Tool .....................................................615 7QC - Seven Quality Control Tool...............................................616 Sampling .........................................................................................618 Scatter Diagram - 7QC Tool .........................................................640 Scorecards .......................................................................................653 SIPOC (Supplier-Input-Process-Output-Customer).................663 SMART Problem & Goal Statements for a Project Charter .....665 Solution Selection Matrix .............................................................672 Stakeholder Analysis ....................................................................681 Statistical Tools...............................................................................684 Stratification - 7QC Tool ...............................................................697 SWOT (Strengths-Weaknesses-Opportunities-Threats)...........699 Tree Diagram - 7M Tool ................................................................712 TRIZ.................................................................................................715 Value Stream Analysis ..................................................................727 Voice of Customer Gathering Techniques..................................737 Work Breakdown Structure (WBS) .............................................753 Y = f (X) ...........................................................................................758

Part III

Best Practices Articles

The Anatomy of Quality Loss in a Product ...............................763 The Anatomy of Variations in Product Performance ...............777 Benchmarking—Avoid Arrogance and Lethargy .....................789 Building Strength via Communities of Practice and Project Management...................................................................................799 Complex Organizational Change Through Discovery-based Learning .........................................................................................827 Lean Six Sigma for Fast Track Commercialization High Risk-High Reward, Rapid Commercialization: PROCEED WITH CAUTION! .....................................................835

viii

Treasure Chest of Six Sigma Growth Methods, Tools & Best Practices

Listening to the Customer First-Hand; Engineers Too ........... 851 The Practice of Designing Relationships....................................873 A Process for Product Development...........................................887 Selecting Project Portfolios using Monte Carlo Simulation and Optimization ..................................................................................921

Part IV

Appendixes

Appendix A Statistical Distribution Tables...................................939 Appendix B Glossary.......................................................................951 Appendix C References ...................................................................979 Index ........................................................................................................981

Acknowledgments Thank you to my friends and professional colleagues who contributed to this book. They took precious time out of their hectic schedules to share the wisdom they have gained through their business experiences. Some contributed inadvertently by brainstorming concepts with me, namely Dan Rose, Joe Szostek, and Chris Tsai. Thank you to the peer reviewers for reading the early drafts to test for understanding. This invaluable manuscript input came from colleagues such as Eric Maass and Scott Wise. Other colleagues authored Best Practices articles, featured in Part III (listed alphabetically): • Thank you to Donna Burnette and David Hutchens, for agreeing to share some of their professional insights on the critical components of learning that have earned them firm national recognition. Their article on discover-based learning programs adds an invaluable perspective on how best to digest and utilize new knowledge and skills. This approach transforms otherwise dry, dense content into a fun and memorable experience. • Thank you to Mike Cook, whose witty, provocative article on the importance of collaboration and designing relationships adds color and refreshing change of pace to this book and reminds us of the important human element involved in the work. • Thank you to Clyde (“Skip”) Creveling for not only sharing his creativity and thought-leadership in writing an article on how to “fast track” a product development process, but also his unending professional support and guidance. • Thank you to both Larry Goldman and Karl Luce for sharing best practices experience on project selection and the portfolio management process using Monte Carlo simulation and optimization techniques. Their insights help give a competitive advantage to any reader. • Thank you to Bill Jewett for his sage and practical approach to writing not one, but four articles. His depth and breadth of experience as a practitioner, manager, and consultant were shared in two articles about robust ix

x

Treasure Chest of Six Sigma Growth Methods, Tools & Best Practices design—specifically on quality loss and on performance variation. In addition, the best practices of collecting and leveraging the Voice of the Customer article, as well as a benchmark product commercialization process article round out an engineering snapshot of product development for a non-technical business person.

• Thank you to Sy Zivan, one of the benchmarking pioneers in the 1980s from Xerox Corporation, for sharing his knowledge on the best practices of the benchmarking process. His article reflects his latest thinking on how benchmarking has evolved over the years. Thank you to each of these professionals and any others I inadvertently have missed mentioning for adding your unique and invaluable perspectives, all of which enhanced this book. Thank you to my family and friends who put up with me during the writing process. Those who wove a supportive, energizing network around me include L. Berkowitz, K. Creveling, D. Croessmann, L. Judson, L. Markt, M. McCandless, and especially my husband, W. Magee. Most importantly, thank you also to the professionals at Prentice Hall for their support and hard work to make this book a reality—Heather Fox, publicist; Bernard Goodwin, editor; Michelle Housley, editorial assistant; and George Nedeff, development editor (listed alphabetically).

About the Author Lynne Hambleton is a business consultant with special focus on strategy development and execution and change management to improve operational processes and expand commercial opportunities. She has held several management positions in Xerox Corporation where she worked for almost 25 years. She also has worked in education, healthcare, and energy public sectors and start-ups. Hambleton’s experience spans general management, marketing, field operations, strategic planning, alliance development, and sales/channel management. She also has served as an adjunct professor of strategic planning at Rochester Institute of Technology’s School of Business. Ms. Hambleton received a Master’s degree in Business Administration, with an emphasis in industrial marketing; a Master’s degree in Adult & Higher Education/Organizational Development; and a Bachelor of Science degree in psychophysiology, all from University of North Carolina— Chapel Hill. Ms. Hambleton is also an active PMI-certified Project Management Professional (PMP) (1998); a Master Black Belt; and Certified Six Sigma Black Belt (CSSBB) from Villanova University (2006). Hambleton’s additional publications include Six Sigma for Marketing Processes, An Overview for Marketing Executives, Leaders, and Managers (coauthors C.M. Creveling and B. McCarthy), Prentice Hall, 2006; the chapter titled, “Supporting a Metamorphosis through Communities of Practice,” in Leading Knowledge Management and Learning, by Dede Bonner, 2000; and the article, “How Does a Company the Size of Xerox Design a Curriculum in Project Management for the Entire Organization?” printed in In Search of Excellence in Project Management, Volume 2, by Harold Kerzner, 1999. Ms. Hambleton lives in Rochester, New York, and can be reached best via email at [email protected] or visiting www.mageemanagement.com.

xi

This page intentionally left blank

Preface The Treasure Chest is part of Prentice-Hall’s Six Sigma for Innovation and Growth Series. This book serves as a consolidated “how to” reference book of Lean Six Sigma, covers growth and innovation tools, provides an overview of methods and the tools to which they align, and offers an overview of additional best practices used to manage a successful Six Sigma growth initiative. The Treasure Chest of Six Sigma Growth Methods, Tools, and Best Practices guides you in selecting the right tool to answer the right question at the right time. The right question drives the requirement or need to be addressed—regardless whether the requirement comes from your customer or a business need. Understanding the overall objective, or requirement, helps determine which Six Sigma methodology to use. The right time dictates what else you know given where you are in your approach. Understanding whether you are in the planning, designing, implementing, or maintaining phases of an overall process determines which process step, thereby the context of the question being asked. Finally, the right tool should be the last question asked, as it is based on first knowing what is required and at what point you are in the process. The Treasure Chest is a desk-reference book for people interested in growth, operations excellence, and business-process improvement. This book speaks to the general business practitioner, business analyst, manager, and leader, regardless of the business context. It is for profit or nonprofit enterprises; large or small firms; whether in headquarters-function, plant, or field location, regardless of functional discipline. The book covers a range of applications from strategic planning aspects of business (offering portfolio renewal) to presenting development and launch preparation, from post-launch operations management to offering discontinuance. Whether contributing a new design (product and/or service), proposing a new process, evaluating a portfolio of offerings, or managing a current portfolio of offerings, this book compiles the resources that help drive growth proactively and presents them in a quick-reference format for easy navigation. This book takes the hassle out of researching the methodology and tools so you can immediately begin to find solutions for your discipline. The Treasure Chest speaks primarily to business people who need practical “hands-on” guidance and answers to the following questions:

xiii

xiv

Treasure Chest of Six Sigma Growth Methods, Tools & Best Practices

1. How do you select the appropriate tool based on the business need (or question being asked) and the required deliverables? 2. How do you use the tool, what inputs or data are required, and what comprises a step-by-step procedure for each tool or technique? 3. How do you analyze the tool’s output and decide on the next course of action? This book was specifically written for general business disciplines, such as marketing, strategic planning, pricing, finance, customer administration, sales, services, support, maintenance, and parts and supplies distribution. This book also is targeted to the technical engineering and research community searching for candidate tools that support communication, project management, risk mitigation planning, and change management requirements.

Common Language Communication presents a challenge when a collaborative team speaks a different language, different filter, different perspective, different interpretation. Successful innovation and growth rely on the integration and collaboration of multiple disciplines, often represented in a cross-functional team. Such teams may be comprised of internal, functionally distinct professionals or any combination of external partners, clients, and sometimes even competition. Regardless if the goal requires creating something from scratch or fine-tuning the management of current offerings, the combination of multiple disciplines, capabilities, and perspectives greatly enhance the end results of this work. However, this collaborative work requires a common language to understand and integrate the diversity. The methods and tools presented in this book assist with interpretation of different perspectives and provide a common platform, foundation, and language from which multiple views can work in harmony. Interestingly, each discipline has its own unique language, different thought-processes, and/or different tools that characterizes its work. Sometimes, perhaps with minor adaptations, a tool considered commonplace by one discipline can be viewed as an “a-ha” eureka discovery to a second group when properly applied. Tools commonly used by marketing groups can excite a technical team if introduced at the appropriate time. For example, a communication summary tool or prioritization tool may be overlooked because it is a “soft tool,” but it actually can fit perfectly when communicating “big picture” thinking. Similarly, if the time is right, marketing teams enjoy the rigor of techniques from the technical counterparts to provide a fresh perspective. Treasure Chest embraces tools used by both the technical and non-technical communities and describes when and how to use them. It contains the business tools and methods for innovation and growth to facilitate best practices sharing and a “common language” across multi-disciplined teams. It also integrates some key technical tools appropriate for general business use (or understanding).

Introduction Different Methods for Different Purposes The Evolution of Six Sigma Six Sigma (or “Lean Six Sigma,” as some refer to it) evolved into a rich set of different standardized methods, tools, and best practices. Six Sigma started as a problem-solving approach to reduce variation in a product and manufacturing environment. That application has expanded to process improvement and other areas of the business, including product or process redesign, research, and technology design, offering portfolio renewal, product development, and post-launch operations management. Six Sigma application stretches beyond the manufacturing enterprises into the services industry and non-profit organizations. Regardless of the application, businesses search for simplicity without jeopardizing the need for robust data. Six Sigma offers a set of methods and tools from which to choose. Six Sigma methods build from a common core foundation yet allow flexibility to adapt to changing environment needs. Part of the flexibility stems from a plethora of candidate tools available, depending on the situation. The tool library ranges from rigorous statistical and quantitative tools to “soft” qualitative ones. The purpose of this book is to help in selecting the most appropriate method and the most appropriate tool within the suite of available candidate tools.

Common Approach to Leverage Everyone’s Contribution, Regardless of Business Model A method establishes the foundation for how work gets accomplished. It defines the who, what, when, where, why, and how of a process; wherein a process describes a series of logically sequenced tasks to complete work. It answers the questions such as, “what gets done;” “who does the work;” 1

2

Introduction

“when the work starts and stops;” “where the work is done;” “why the work is being done;” and “how the work is to be completed.” A welldeployed method orchestrates and integrates the people working in a process in an efficient and effective (streamlined) set of activities. It organizes the work as defined by a set of customer requirements. The work produces the agreed-to deliverables according to their “acceptance” criteria. A well-constructed method defines a set of tasks that circumvent redundancies and gaps. Activities done beyond the prescribed work to produce the required deliverables arguably could be called unnecessary or a “waste.” A method defines which tool best supports a task and will produce the desired results, providing a common language of terms and tools and a common way of working for those involved in the process.

Overview of Six Sigma Method The methods used in Six Sigma (including Lean Six Sigma) contain several common principles, such as data-driven decision-making and project management fundamentals. Part I, “Six Sigma Methodology Overview— Choosing the Right Approach to Address the Requirements,” uses these principles to organize its content.

Tool-Task-Deliverables Linkage Six Sigma methods represent a structured thought process that starts with thoroughly understanding the requirements (or key business questions) before proceeding. The requirements, in turn, define the deliverables to be produced and the tasks needed to produce those deliverables and, last, the supporting tools to be used to complete the tasks and produce the deliverables. This structure is often called the Tools-Tasks-Deliverables combination to indicate the interdependencies. The Tools-Tasks-Deliverables linkage is executed in “reverse” or from right-to-left, starting with Deliverables. Hence, a tool is selected only after the requirements, deliverables, and tasks are well understood to ensure that the appropriate tool is used for a given task and to avoid the “rut” of treating everything as if it were a nail when the only tool you use is a hammer. The various Six Sigma methods suggest a variety of applicable tools to choose from, but rarely does a given project require the utilization of every tool. No tool fits every situation. Determining which tool fits best depends on the situation. Thus, tool selection is done only after the requirements, resulting deliverables, and tasks are completely understood. [Part II of this book provides not only an inventory of potential tools, but also information on how to apply and interpret results to help you in tool selection.] Remember: Use the right tool at the right time to help ask and answer the right questions.

Over view of Six Sigma Method

Result-metrics The result-metrics focus is a distinguishing principle of Six Sigma methods. These fact-based metrics determine whether (internal or external) customer requirements are achieved. Performance typically is evaluated via a statistical metric of the process or offering (e.g. product, services, or information). High-level process and performance metrics define what critical-toquality is and encompass the critical parameters necessary to meet requirements. Eventually, these metrics should be translated into a language that is “meaningful” to a process worker involved in providing either the inputs or process deliverables (outputs). Depending on the requirements, the result-metrics may be “hard” or “soft” measurements— quantitative or qualitative; continuous or attribute data. A good litmus test for translated critical-to-quality metrics is whether a “new hire” understands clearly what is expected of him/her to meet requirements with no “fuzzy” or nebulous evaluation of what characterizes “good” or “poor” performance.

Process-centric Another principle employs a process-centric view. Understanding how inputs to a process are integrated and how value is added to a product, information, or services offering is as important as what is being added. Understanding the combination of what and how inputs and other key variables come together to produce the final outputs (or deliverables) enables a more accurate forecast of whether customer requirements (or targets) will be satisfied. Prior “results” alone are poor predictors of future outcomes, and without knowledge of the process, any forecast is blinded; any successful forecast would be by chance. Because business prefers accurate forecasts of performance, a process-centric view becomes an integral ingredient.

Adaptive and Iterative Methods used in Six Sigma are adaptive and iterative. Adaptive implies the fact that it can be tailored to a variety of situations and business contexts. Moreover, any given Six Sigma method can be integrated with another process or methodology as an underpinning to identify, gather, analyze, and report on critical parameters in a proactive or reactive manner. For example, if your firm has an existing standard product development process or customer account selling approach, Six Sigma can supplement it and make it more robust. The adaptive nature of these methods also speaks to the wide array of industries and situations in which they can be applied. The breadth of industries includes military, government, automotive, aerospace, high-tech, manufacturing, office products, financial

3

4

Introduction

services, e-commerce, logistics and supply chain, healthcare, and pharmaceutical industries. Within companies, multiple disciplines have embraced Lean Six Sigma: manufacturing, engineering, finance, administration, customer operations, maintenance, services deployment, and marketing and sales. The approach can even be applied to personal and social situations. Of the Top 100 companies in the 2005 Fortune 500 list, 70 of them have been in the top 100 for five or more years. Interestingly, of those 70 companies, 63% of them publicly acknowledge implementing Six Sigma to some degree. Through further analysis, we have found that these same 44 Six Sigma users also reported 49% higher profits (compounded annually) on average than their peers. The iterative nature of the Six Sigma methods stems from the fact that more information on a variable or potential root cause gets revealed as the project progresses. Hence, one path of inquiry based on one assumption may prove to be a dead-end or altered, as more data on the current state becomes known. Although Six Sigma methods use a project-structure with phase-gates, a fundamental principle across the various approaches encourages informed updates to prior step deliverables, as appropriate, and promotes proper communication. Use the best information available at the time but continue to ask questions and keep an open mind. Six Sigma projects involve a discovery process wherein an individual serves as a “sleuth,” investigating, exploring, hypothesizing, and testing assumptions.

Data-driven Decision-making Given the uncertain nature of projects, when seeking facts that answer key business questions, revisions to earlier project work reflect the evolving discovery of fact-based results. For example, a business operations review may focus on a set of key metrics to manage a process. If a chronically missed target evokes a Six Sigma project, an interim project deliverable could identify that some of the metrics associated with the key variables driving the desired business outcomes are missing, hence the “dashboard” requirements become refined to reflect the vital few parameters, comprised of both leading and lagging metrics. Next, the project could focus on establishing baseline data for the “new” critical metrics to re-evaluate performance and better understand any cause-and-effect relationship(s).

Project-based Methods As previously referenced, Six Sigma methods tend to use a project structure. A project structure has a distinct beginning and end to the work performed. The requirements phase determines the boundaries of this definitive timeframe. A project team, with defined roles, forms only for the duration of the project’s timeframe. A project structure adds the rigor of requiring completed deliverables approvals (often gained in a phase-gate review

Over view of Six Sigma Method

meeting with the project and key stakeholders) before exiting a given phase-gate or step and starting another one. Project structure borrows heavily from the project management discipline and its nine knowledge areas to manage the lifecycle of the project: scope, time, budget/cost, risk, quality, communications, human resources, procurement, and integration. The project context of Six Sigma methods incorporates a rather shortterm perspective (averaging a three, six, or twelve-month project scope). A project may involve an improvement or enhancement to something [often focused on reducing defects, minimizing variance from a target, or improving velocity (speed)], clean-sheet innovation, or design and creation (such as product or services development or portfolio assessment). The technical community (for example, engineering or manufacturing) has embraced a category of Six Sigma methods called Design for Six Sigma (DFSS). A newly emerging field is Six Sigma for Marketing (SSFM). SSFM may be a misnomer, because the various methods apply to the remaining (“non-engineering”) business disciplines, such as marketing, sales, strategic planning, services, and customer operations. Examples of project-based methods include • DMAIC (Define-Measure-Analyze-Improve-Control, and its variants DMAIIC (with “II” representing Improve-Innovate) and Lean Six Sigma) • Lean (and its variants PDCA (Plan-Do-Check-Act) / PDSA (PlanDo-Study-Act) and Lean Six Sigma) • DFSS category with DMADV (Define-Measure-Analyze-DesignVerify), CDOV (Concept-Design-Optimize-Verify) (and their variants DMEDI (Define-Measure-Explore-Develop-Implement), PIDOV (Plan-Identify-Design-Optimize-Validate), ICOV (IdentifyCharacterize-Optimize-Verify), and IIDOV (Invent-InnovateDevelop-Optimize-Verify)) • SSFM category with UAPL (Understand-Analyze-Plan-Launch) and (sometimes) IDEA (Identify-Define-Evaluate-Activate)

Operational-based Methods Managing an ongoing operation, however, is emerging as a new application area. Hence, the application of the Six Sigma method and tools to an operational process may last for years, rather than months as with a short-term project. The objective of operational-based Six Sigma is to manage or sustain an improvement of a launched product and/or services offering, for example, or to adapt and respond to environmental changes. This operational focus of Six Sigma is being applied to business areas such as customer operations (for sales, services, support, administration, financing, and related business disciplines) and strategic

5

6

Introduction

planning (for offerings portfolio management). Some might argue that portfolio management can be handled as a project defined by an annual planning cycle, but others view it as an ongoing area. Nonetheless, Six Sigma discipline has added a competitive advantage to those firms that have begun to apply its method and tools to this process area. The operations-based methods currently fall within the SSFM (Six Sigma for Marketing) category with LMAD (Launch-Manage-AdaptDiscontinue, for customer operations) and sometimes IDEA (IdentifyDefine-Evaluate-Activate, for strategic planning), which can be considered operational given that the management and revitalization process of a firm’s offerings portfolio is cyclical and can span multiple years.

How Do the Various Six Sigma Methods Fit Together? In summary, this Introduction overviews the major Six Sigma approaches being used currently. Each Six Sigma method has a valid purpose in today’s business world, the selection of which approach best fits a need depends on the key business question being asked at the time. They all fit together and inform one another. The integrated view is as follows: An enterprise’s strategic platform defines its business and offerings, so typically a process flow starts with the business strategy process of portfolio definition and renewal (IDEA — Identify-Define-Evaluate-Activate). From there, funding gets earmarked for research, tactical, and operational activities. Research and Technical Development (R&TD) efforts are funded to develop forward-looking capabilities that eventually feed product development and commercialization. The approach used to guide Research’s activities is called I2DOV (Invent-Innovate-DevelopOptimize-Verify). The specific offering’s design, development, and commercialization efforts split into two branches: 1) the technical team that uses CDOV (Concept-Design-Optimize-Verify) to guide its activities and 2) the marketing and business areas that use the UAPL (Understand-AnalyzePlan-Launch) approach. Finally, the operational and supporting infrastructure and business areas of a post-launch customer value chain environment use the LMAD (Launch-Manage-Adapt-Discontinue) approach to guide and direct their activities. If any of these areas of an enterprise encounter a trouble spot in an existing process or offering, the Lean Six Sigma DMAIC (Define-MeasureAnalyze-Improve-Control) method and any of its variants (that is, DMAIIC (Define-Measure-Analyze-Improve-Innovate-Control), DMADV (Define-Measure-Analyze-Design-Verify), DMEDI (Define-MeasureExplore-Develop-Implement)) focus on the problem and its root causeand-effect to determine the best correction.

Right Tool at the Right Time

Figure 1 depicts how each method integrates with one another. Unique Focus on Proactive Growth

Strategic Offering Portfolio Renewal Process

Tactical Offering Commercialization Processes

L-M-A-D Marketing, Selling and Customer Value Chain

U-A-P-L Marketing and Business

I-D-E-A

Operational PostLaunch Process

Production Engineering and Manufacturing

C-D-O-V Techinical

Strategic Research and Technology Development Process

I2-D-O-V Lean Six Sigma Problem-solving Process

D-M-A-I-C Unique Focus on Reactive Cost Control and Variation Reduction

Figure 1: Integrated Portfolio of Six Sigma Methods

The various Six Sigma approaches have their appropriate applications. Moreover, their candidate tools and methods sometimes overlap or feature tool variants of another’s, as they all build from the core Six Sigma fundamentals.

Right Tool at the Right Time Part II of this book explores how to select the most appropriate tool to answer the right question at the right time. It organizes the tools in alphabetical order and, using an encyclopedia-style article format, describes what question each tool tries to answer and how to use the tool. Take a closer look at the structure of the various Six Sigma (or Lean Six Sigma) methods to understand their similarities and differences and when best to apply each of them. Start with Part I to understand the various methods and their structure, requirements, deliverables, and list of candidate tools. Afterward examine Part II, the heart of this book, to decide which of the candidate tools might be appropriate for your project. Part II is structured as a desk reference that inventories the different tools and techniques. Each tool “article” describes the main purpose of the tool, how best to use it, how to interpret the results of the tool, and any variations on how to apply the tool or technique. Most importantly, Part I lists the candidate tools aligned with particular requirements and task-deliverable

7

8

Introduction

combinations, but Part II identifies in more detail which key question each tool tries to answer. Remember—an effective Six Sigma practitioner scrutinizes the candidate tool set and selects the right tool, at the right time, to answer the right question.

Special Note Throughout this book, the word “product” references to a generic company “offering” and represents both a tangible product as well as a services offering. This book discusses technology-based products frequently because of marketing’s interdependency with the technical community. In parallel, R&D, design and production/services support engineering should be using growth and problem prevention-oriented forms of Six Sigma in their phases and gate processes. The Six Sigma approach serves as a common language between the marketing and technical disciplines. The term “solutions” usually involves both technology and services; thus, “product” and “services” encompass the scope of a given solution. Regardless of the offering, the Six Sigma approach we are outlining is the same and can be applied to either a tangible product or a services offering.

In addition, the term “Six Sigma” refers to the generic field or discipline and encompasses the many different approaches. People may distinguish Six Sigma (SS), Lean Six Sigma (LSS), Design for Six Sigma (DFSS) and Six Sigma for Marketing (SSFM) from one another. However, this book uses “Six Sigma” categorically unless otherwise noted.

What this Book Covers The Treasure Chest is organized into three parts: 1) Six Sigma methodology, 2) tools and techniques, and 3) best practices applicable to Six Sigma deployment. The book uses an encyclopedia-like format made up of over 60 topics.

Part I Six Sigma Methodology Overview—Choosing the Right Approach to Address the Requirements This segment of the book presents an overview the various Six Sigma approaches and describes the purpose of each. With the knowledge of the different methods, this book connects the requirements of a method to the appropriate candidate tools and techniques. This Part serves as the foundation for selecting the right tool for a given purpose.

What this Book Covers

It discusses the various technical, business and marketing Six Sigma methods, including DMAIC, Lean Six Sigma, Design for Six Sigma (DFSS), and Six Sigma for Marketing (SSFM). Within each approach, a general description, common applications, and key requirements provide an overview of its structure and purpose. In addition, the key requirements determine each method’s unique tools-tasks-deliverables combination. The Method Section provides the foundation for comparing and contrasting the different approaches. Once a method is selected, it establishes a team’s common language, regardless of whether or not people are part of a hetero- or homogeneous team. The method defines the requirements for its work activities. It sets expectations, describes required deliverables and their due dates, and identifies who does what when. A common method is the single most critical unifying theme for a team’s work. If it is well understood and followed by each team member, then collaboration, coordination, and communication can occur fluidly.

Part II Six Sigma Encyclopedia of Business Tools and Techniques— Choosing the Right Tool to Answer the Right Question at the Right Time This Part features an in-depth look at a robust library of tools, organized in alphabetical order for easy reference. By design, it represents the bulk of this book, with an exhaustive review of the Growth and Lean Six Sigma tools and techniques. Each tool (tangible item) or technique (for example, brainstorming) features a “how to” description explaining how to utilize the tool and interpret typical outcomes. Each description identifies the deliverable the tool or technique supports. Part II features the following topics for each of the 60+ tools included in this book: • The question the tool helps to answer • Any alternative names or variants associated with the tool • When best to use the tool or technique • A brief description and useful real-world examples • How to implement the tool for your application • How to analyze and apply its output • Helpful hints and tips that encourage you to think outside of the box • Supporting candidate tools that link to the featured tool, depending on the question needing to be answered, by providing either input to or using the output from this featured tool Part II begins with a useful summary table of tools organized by the type of question it helps to answer. The Treasure Chest organizes the candidate tools and techniques alphabetically for easy reference. The encyclopedia includes an array of soft tools and techniques, graphical methods, and statistical tools. The statistical tool descriptions leverage some of the appropriate software

9

10

Introduction

tools such as MINITAB, Minitab’s new Quality Champion, Decisioneering Crystal Ball, Visio, and other Microsoft applications, such as Excel.

Part III Best Practices Articles This portion of the book contains a series of articles written by well-reputed professionals that complements and extends beyond the world of Six Sigma to provide that competitive advantage in growth and operational excellence. It features a collection of technical and humanistic topics ranging from the latest thinking on benchmarking strategy to determining and offering the best portfolio. Articles describe how best to accelerate the development of an offering, how to ensure design robustness, and how to govern projects. Part III includes a discussion on best practices to introduce, deploy, and sustain a major culture change, such as deploying Six Sigma thinking, by featuring a set of articles to support a change initiative including communities of practice and project management; simulation approach to training of new skills, knowledge, and attitudes; and designing collaborative work relationships. A final note, the “Six Sigma for Marketing” (or SSFM) terminology in the marketplace may mislead prospective practitioners wishing to use Six Sigma to drive innovation and growth. While the primary application of SSFM involves processes typically associated with marketing, the respective work often encompasses additional functional disciplines within a company. Depending on the size of the firm and its business model, the professionals involved in 1) portfolio renewal, 2) offering development and commercialization preparation, and finally 3) post-launch operations management throughout an offering’s lifecycle reach beyond just marketing. General business professionals involved in these three processes also represent disciplines such as strategic planning, pricing, finance, customer administration, customer service and support, professional services, and logistics and supply chain. In fact, the third process involving post-launch operations spans the entire customer value chain. Hence, when this book references “SSFM,” it follows the marketplace terminology of the emerging Six Sigma focus. In the context of SSFM, the identification of “marketing” (and sometimes sales) distinguishes the new Six Sigma application as different from the classic variation reduction, problem-solving, and cost cutting approach, and as different from the technical DFSS (Design for Six Sigma). However, the Treasure Chest intends the reference to marketing as a “loose” association and prefers the broader reference of “general business” to better articulate its broader scope and applicability. This book works nicely as a sequel to the Six Sigma for Marketing Processes, co-authored by CM Creveling, L. Hambleton, and B. McCarthy.

Part IV Appendixes The Appendixes contain a set of references such as charts and statistical tables for hypothesis testing, a glossary, and a list of references.

Pa r t I Six Sigma Methodology Overview: Choosing the Right Approach to Address the Requirements Section 1

Define-Measure-Analyze-Improve-Control (DMAIC)

Section 2

Lean and Lean Six Sigma

Section 3

Design for Six Sigma (DFSS)

Section 4

Six Sigma for Marketing (SSFM)

11

This page intentionally left blank

1 Define-Measure-AnalyzeImprove-Control (DMAIC) Six Sigma’s most common and well-known methodology is its problem-solving DMAIC approach. This section overviews the methodology and its high-level requirements, given that the requirements define the appropriate deliverables, which dictate the tasks and the tool selection to aid in the task. This section also outlines the DMAIC standard toolset, through the understanding of the tooltask-deliverables linkage, to facilitate appropriate selection of a tool when referencing the “how to” tool articles in Part 2 of this book.

What Is the Main Objective of this Approach? The DMAIC (Define-Measure-Analyze-Improve-Control) is the classic Six Sigma problem-solving process. Traditionally, the approach is to be applied to a problem with an existing, steady-state process or product and/or service offering. Variation is the enemy—variation from customer specifications in either a product or process is the primary problem. Variation can take on many forms. DMAIC resolves issues of defects or failures, deviation from a target, excess cost or time, and deterioration. Six Sigma reduces variation within and across the value-adding steps in a process. DMAIC identifies key requirements, deliverables, tasks, and standard tools for a project team to utilize when tackling a problem.

Brief Description of DMAIC Applications This classic or traditional Six Sigma methodology was designed to solve a problematic process or product and/or service offering to regain control. It addresses improvements in productivity (how many), financial (how much money), quality (how well) and time (how fast)—PFQT. Originally costs dominated the financial aspects, but lately project focus has shifted 13

14

Define-Measure-Analyze-Improve-Control (DMAIC)

to revenues and growth. The 5-step DMAIC [pronounced “duh-MAY-ick”] method often is called the process improvement methodology. The classic strategy reduces process variance (in total, across the activities and within-step) to bring it back on target—the customer specification or requirement. Knowing that Six Sigma resolves more issues than just cycle time, Figure 1-1 highlights its impact on cycle time by contrasting a problematic process versus its post-Six Sigma improved state. Step 1

Step 2

Step 3

Step 4

Problematic Process

Process After Six Sigma

Step 1

Step 2

Step 3

Step 4

Cycle Time

Figure 1-1: Six Sigma’s Impact on Cycle Time

The DMAIC approach is designed to allow for flexibility and iterativework, if necessary. As more is learned through the 5-step process, assumptions or hypotheses as to the root cause of the problem may be disproved, requiring the project team to revisit them and modify or to explore alternative possibilities. For example, root cause to a sales force effectiveness issue may have been hypothesized as a sales training problem in a specific geographic region. Rather than jumping to conclusions without facts by implementing a new sales training program, the Six Sigma project team wisely decides to gather facts about the problem first. After some investigation and analysis, the team discovers that the root cause points to an issue with sales management direction, not lack of sales representatives’ knowledge and skills. If the project team acted upon the original assumption, time and money would have been wasted on developing a mismatched solution that would have produced poor results; the team’s hard work would have gone to waste. Instead, the team did a mid-course correction based on facts, adjusted its hypothesis, and developed a solution directly aimed at the true root cause—hence favorable results ensued. DMAIC builds on three fundamental principles: • Results-focused; driven by data, facts, and metrics. • Work is project-based (short-term in nature, with length depending

on scope and complexity) and project-structured, versus an ongoing process.

Brief De scription of DMAIC Applications

15

• Inherent combination of tools-tasks-deliverables linkage that

The DMAIC methodology uses a process-step structure. Steps generally are sequential; however, some activities from various steps may occur concurrently or may be iterative. Deliverables for a given step must be completed prior to formal gate review approval. Step Reviews do occur sequentially. The DMAIC five steps are Step 1. DEFINE the problem and scope the work effort of the project team. The description of the problem should include the pain felt by the customer and/or business as well as how long the issue has existed. Hence, identify the customer(s), the project goals, and timeframe for completion.

The appropriate types of problems have unlimited scope and scale, from employee problems to issues with the production process or advertising. Regardless of the type of problem, it should be systemic—part of an existing, steady-state process wherein the problem is not a one-time event, but has caused pain for a couple of cycles. Step 2. MEASURE the current process or performance. Identify what data is available and from what source. Develop a plan to gather it. Gather the data and summarize it, telling a story to describe the problem. This usually involves utilization of graphical tools. Step 3. ANALYZE the current performance to isolate the problem. Through analysis (both statistical and qualitatively), begin to formulate and test hypotheses about the root cause of the problem. Step 4. IMPROVE the problem by selecting a solution. Based on the identified root cause(s) in the prior step, directly address the cause with an improvement. Brainstorm potential solutions, prioritize them based on customer requirements, make a selection, and test to see if the solution resolves the problem. Step 5. CONTROL the improved process or product performance to ensure the target(s) are met. Once the solution has resolved the problem, the improvements must be standardized and sustained over time. The standard-operating-procedures may require revision, and a control plan should be put in place to monitor ongoing performance. The project team transitions the standardized improvements and sustaining control plan to the process players and closes out the project. A DMAIC project typically runs for a relatively short duration (three to nine months), versus product development projects (using UAPL or DFSS) and operational line management (using LMAD), which can run

DMAIC

varies by step in the method.

16

Define-Measure-Analyze-Improve-Control (DMAIC)

years. Given the relatively shorter duration to other types of Six Sigma methodologies, we distinguish the DMAIC as having five steps, rather than phases. The DMAIC method is primarily based on the application of statistical process control, quality tools, and process capability analysis; it is not a product development methodology. It can be used to help redesign a process—any process, given that the redesign fixes the initial process problem. To be implemented, the method requires four components: • A measurement system (a gauge) of the process or product/service

offering in trouble. • Standard toolset that supports tasks to produce deliverables

(including statistical, graphical, and qualitative tools and techniques). • An ability to define an adjustment factor(s) to correct the process or

product/service offering back on target. • A control scheme to maintain the improvement or correction over

time by implementing a control plan with a monitoring system to audit the response performance against statistical control limits and defined action plans if needed.

What Key Overall Requirements Define this Approach? Requirements come from the customer and the business, depending on the problem scenario. The (internal and external) customer requirements get translated into what is critical-to-quality (CTQ). These CTQs define the criteria to evaluate what good looks like—how well the project scope and deliverables meet requirements. Hence, the project team must meet the requirements of a phase before declaring completion and closing it out. The DMAIC method was designed and structured to answer the following overall business questions: • What does the customer define as the problem? (Secondarily, is the problem sustained over time, is it chronic, or is it a one-time occurrence?) • What characterizes the current problem (e.g., process and performance metrics), and how has it changed over time? (Secondarily, is the process in control, and does it have a good measurement system? Is the process capable of producing the customer requirements?)

What Requirement Determine s the Key Activitie s in this Approach?

17

(Secondarily, is the process capable of producing the customer requirements?) • What controls should be implemented to sustain this improvement, including a warning system, action plan, and communication plan needed in case requirements fail to be met? (Secondarily, can the improvements be sustained over time?)

What Requirement Determines the Key Activities in this Approach? The preceding key business questions determine the DMAIC architecture. Figure 1-2 depicts a high-level process flow of the DMAIC method through its five steps. Define

Measure

Analyze

Yes Problem over-time? No

Improve No

In Control? Yes

No

Remove Known Special Causes

Process Capable? Yes

Close Project

Yes Close Project

Good measurement system?

Control Yes No

No

Process Capable?

Sustained Improvement?

Yes Close Project

No Remove Measurement System variation

Figure 1-2: High-Level DMAIC Process Flow

Table 1-1 shows the linkage between the high-level business requirements and the five-step DMAIC method.

DMAIC

• What are the root causes, and what improvement actions correct them to meet customer requirements again?

18

Define-Measure-Analyze-Improve-Control (DMAIC) Table 1-1: DMAIC Requirements-Step Linkage Requirements

Resulting High Level Task (Step)

What does the customer define as the problem?

1. DEFINE • Describe in the words of the external or internal customer—Voice of Customer (VOC). • Define the boundary conditions set forth by the business, including regulatory environment—Voice of Business (VOB). • Understand the current process. What has happened over time, examine process control charts to identify incidents of common and special cause variation— Voice of the Process (VOP).

What characterizes the current problem (that is, process and performance metrics), and how has it changed over-time?

2. MEASURE • Measure the problem; describe it with facts, data, and performance metrics. Determine if the process in control and if the measurement system is accurate. • Considered iterative until metrics are gathered over time.

What are the root causes?

3. ANALYZE Determine if the process capable of producing the customer requirements. If not, consider it iterative until root causes are identified and verified with facts and data.

What improvement actions correct the root causes to meet customer requirements again?

4. IMPROVE Determing if the process is capable of producing the customer requirements. If not, consider it iterative until improvements are identified and verified with facts, data, and performance metrics.

What controls should be implemented to sustain this improvement, including a warning system, action plan, and communication plan needed in case requirements fail to be met?

5. CONTROL • Demonstrate how the improvements and/or changes can be sustained. • Manage Risks

Figure 1-3 provides a DMAIC icon that reinforces both the overall flow of a method and the purpose of each step and respective interrelationships. It summarizes the five-step DMAIC process and its notable iterative nature. Throughout the remainder of this text, Figure 1-3 will symbolize the DMAIC approach and indicate a particular step within it if appropriate.

Define

Measure

Analyze

Improve

Control

Figure 1-3: DMAIC Icon

What Tools Are Aligned to Each Step of the Process? Given the preceding High Level Task Step(s), the following series of tables summarize the subsequent tool-task-deliverables combination associated with each individual step within the five-step approach. The detail behind how to use each tool can be found in Part II, “Six Sigma Encyclopedia of Business Tools and Techniques: Choosing the Right Tool to Answer the Right Question at the Right Time.” Table 1-2: Define Tools-Tasks-Deliverables

Step 1: DEFINE Define

Measure

Analyze

Improve

Control

What does the customer define as the problem?

Deliverables

Tasks

Candidate Tools and Techniques

Project Charter Approved Identify Problem • SMART (contract with the statement/Opportunity • Project Charter Form Sponsor regarding and Goal statement. containing: Problem problem, project scope, Statement (As-Is), project goal(s), key Desired State (To Be), deliverables, timeframe, and the Business Reasons and budget) • Big “Y” over time

continues

19

DMAIC

What Tools Are Aligned to Each Step of the Proce ss?

20

Define-Measure-Analyze-Improve-Control (DMAIC) Table 1-2: Continued Deliverables

Tasks

Candidate Tools and Techniques

High-level Process Map Constructed

Develop High-level Process Map

• Process Map • RACI Matrix

Critical Parameters Hypothesized

Gather VOC and Business Requirements

• VOC/VOB Gathering techniques • Current process control charts (VOP) • Stakeholder Analysis • CTQ

Project Charter Published and Communicated

Develop Communication Plan

Communication Plan template

High-Level Project Plan Defined and Approved

Finalize Project Charter

• Project Charter Form • High-level Process Map • SIPOC • Project RACI Matrix

Table 1-3: Measure Tools-Tasks-Deliverables

Step 2: MEASURE Define

Measure

Analyze

Improve

Control

What characterizes the current problem, and how has it changed over time?

Deliverables

Tasks

Candidate Tools and Techniques

Data Collected

• Identify Sources • of Data • Collect Baseline • Data from existing process • • Determine current • Process Performance; • is it in control? • • Remove any known special causes; verify if process is in control

Y = f(X); Big “Y” and little “Ys” Data Gathering Plan template Control Charts Statistical Sampling Graphical Methods QFD (Quality Function Deployment)

Deliverables

Tasks

Candidate Tools and Techniques

Process Map Defined In-depth With Current Measures

Develop Detailed Process Map

• Detailed Process Map • RACI Matrix, revised

Current Measurement System Capability Evaluated

• Validate measurements and collection system • Is the process capable of meeting requirements?

• Measurement System Analysis (MSA) • Process Capability Analysis

Project Charter and Plan updated, as necessary

• Revise Problem and • Project Charter; its plan Goal statements as and milestones needed • Project RACI Matrix • Update Project Plan, as needed

Table 1-4: Analyze Tools-Tasks-Deliverables

Step 3: ANALYZE Define

Measure

Analyze

Improve

Control

What are the root causes of the current problem?

Deliverables

Tasks

Candidate Tools and Techniques

Data Analyzed

• Validate gaps in • requirements vs. current metrics • • Establish Y=f(X) • Quantify Opportunity • to close gaps •

Y = f(X); Big “Y”; little “Ys” and the “Xs” Critical Gap/Step Analysis Pareto Charts Statistical Analysis: Normal Distribution, Variation • Correlation and Regression

continues

21

DMAIC

What Tools Are Aligned to Each Step of the Proce ss?

22

Define-Measure-Analyze-Improve-Control (DMAIC) Table 1-4: Continued Deliverables

Tasks

Candidate Tools and Techniques

Process Analyzed

• Develop Detailed • Detailed Process Map Process Map (inputs, outputs, met• Establish Y=f(X) rics, process step • Quantify Opportunity owners) to close gaps • RACI Matrix, revised • Process Mapping of Critical Parameters • Y = f(X) • Pareto Charts • Process Capability Analysis

Root Cause Analyzed

• Conduct Root Cause Analysis • Prioritize Root Causes • Quantify Opportunity to close gaps

Project Charter and Plan updated, as necessary

• Revise Problem and • Project Charter; Goal statements as its plan and milestones needed • Project RACI Matrix • Update Project Plan, as needed

• Brainstorming Technique • Cause and Effect Diagrams • Five Whys • Affinity Diagram (KJ) • Hypothesis Testing of key causes and/or critical parameters (vital few Xs) • Inferential statistics (Correlation and Regression) • DOE • FMEA

What Tools Are Aligned to Each Step of the Proce ss?

23

Table 1-5: Improve Tools-Tasks-Deliverables

Define

Measure

Analyze

Improve

Control

the root causes to meet customer requirements again? Deliverables

Tasks

Candidate Tools and Techniques

Potential Solution Generated

Develop Potential Improvements or solutions for root causes

• Brainstorming Technique • Positive Deviance • TRIZ

Potential Solution Evaluated

• Develop Evaluation Criteria • Measure results • Evaluate improvements meet targets • Evaluate for Risk

• • • •

Solution Selected

Select and Implement • Pugh Concept improved process and Evaluation metrics • Solution Selection matrix • Force Field diagram • QFD • Measurement System Analysis (MSA) • Process Capability Analysis

Improved Path Forward Implemented

Develop Detailed Future Process Map of improvement

Project Charter and Plan updated, as necessary

• Revise Problem and • Project Charter; its plan Goal statements as and milestones needed • Project RACI Matrix • Update Project Plan, as needed

Basic DOE Pilots/Tests FMEA Cost/Benefit Analysis

• Detailed Process Map • RACI Matrix, future • Procedure manual (standard operating procedure) • Implementation and Transition Plan

DMAIC

Step 4: IMPROVE What improvement actions correct

24

Define-Measure-Analyze-Improve-Control (DMAIC) Table 1-6: Control Tools-Tasks-Deliverables

Step 5: CONTROL Define

Measure

Analyze

Improve

Control

What controls should be implemented to sustain this improvement?

Deliverables

Tasks

Candidate Tools and Techniques

Control Plan Defined

• Document New Measurement Process • Define control plan

• • • • •

Control Plan Design Control Charts (SPC) FMEA/Risk Analysis Communication Plan Stakeholders Analysis

Improvements/ Validate metrics and Innovation Implemented collection systems

• Measurement System Analysis (MSA) • Process Capability Analysis • Cost/Benefit Analysis

Training Conducted

Train

• Training/Transition plan

Process Documented

Document recommend- • Process Map ation or improvement • RACI summary and highlight • Procedure manuals changes from As-Is to Improved

Tracking System Deployed

Establish Tracking Procedure

• Scorecard or Dashboard • Data Mining (MINITAB graphical data analysis)

Lessons Learned Documented and Project Closed

• Revise Problem and Goal statements to reflect actual • Update Project Plan to reflect actual • Record lessons learned and file along with final project documentation

• Project Charter; its plan and milestones • Project RACI Matrix • New SIPOC

What Are Some of the Key Concepts that Characterize this Approach?

25

There are some key characteristics that distinguish DMAIC from other Six Sigma methods. The following overview wraps up the DMAIC highlights and introduces some of its variants.

How Is the Problem Defined? The problem statement in a Project Charter typically speaks to defects or variance from a target over time with an existing, steady state, process, or product. (The charter is part of a standard Six Sigma toolset used to document the project scope. See Also “SMART,” in Part II, p. 665) Typically, the customer should determine the target; however, at times the business, industry standard, or regulatory agency may set it. Time-based problem statements indicate the problem may be chronic (has persisted for a period of time), which helps create a case for change (versus a one-time occurrence) to incite interest in and resources to tackle the issue. Common metrics include DPMO (Defects per Million Opportunities (or units)), PPM (Parts per Million), Mean Time between Failures (MTBF), Cost, Percent Variance, or Errors.

What Is Commonly Measured? Typically, three key items are measured: • Output (or Outcome)—The end result of the process (or product) requiring improvement • Process—The workflow (of activities and items) that produces the output • Inputs—The raw materials and information used by the process to produce the output The relationship of these three key items often is described as an equation: Y = f(x), which reads, “Y is a function of X.” The “Y” refers to the output(s); the “X” refers to the key measures from the process variables (inputs and/or the process itself). See Also “Y=f(x),” in Part II, p. 758. The DMAIC project goal is to identify the critical (or vital few) Xs—the root cause of the problem—and select their optimal level(s) to best drive the desired improvement in the output performance (sometimes called the “Big Y”). This language sounds foreign to many people not comfortable with mathematically-structured sentences; however, it is readily used in most Six Sigma texts. A simpler articulation is the goal of a DMAIC project is to improve PFQT—Productivity (how many), Financial (how much money), Quality (how well), and Time (how fast).

DMAIC

What Are Some of the Key Concepts that Characterize this Approach?

26

Define-Measure-Analyze-Improve-Control (DMAIC)

Are There any DMAIC Variations? There are two prevalent variations to the traditional DMAIC method. Both build on the DMAIC fundamentals but add new dimensions to extend its applications. The first is DMAIIC, wherein innovation is added for situations where a simple improvement modification is inadequate and a new design may be required. (Note that in the technical, engineering arena that an innovation adaptation typically aligns with the DMADV method. DMADV distinguishes itself from DMAIIC by not only its often unique environment scenario, but also it usually calls for a requirement of building a new process (or product design) from scratch at the start of the project; whereas, DMAIIC often is unaware of the redesign requirement until much later into the project lifecycle. See Also “DFSS,” in Section 3, p. 45) The second is Lean Six Sigma, which adds concepts of velocity, value-add, and flow to the DMAIC concepts. 1. DMAIIC—Adding an “I” for Innovation—Many organizations have found that improving a current process or product may not be enough to deliver the desired results, and at times innovation is needed. Since the project teams have just completed the Define-Measure-Analyze stages of the process and are in the midst of Improve, rather than starting over from scratch, project teams have found that the work done to this point is a good foundation for innovation work. Hence, some companies have built on the DMAIC framework already in-place and added a second “I” for innovation to keep the project team progressing. Therefore, the variation is Define-MeasureAnalyze-Improve/Innovate-Control. Figure 1-4 shows the DMAIIC flow diagram and depicts how the Innovate tasks integrate into the classic DMAIC model. 2. Lean Six Sigma—Adding Lean Concepts—By incorporating lean concepts into DMAIC, the project adds a dimension of velocity (i.e. improved cycle time), value-add, and flow to what Six Sigma already offers.

Both concepts share similar views on customer-focus, process-centric work, and appropriate tools. Lean simply adds a deeper set of tools to eliminate waste between process steps handoffs. Often DMAIC provides a project the big picture view (what the customer values balanced by business values) and process stabilization and capability—while Lean introduces speed and flow concepts at a more detailed level. The Define-Measure-Analyze-Improve-Control structure still holds true for Lean Six Sigma projects. See Also “Lean and Lean Six Sigma,” Section 2, p. 29, for more details on Lean.

Summar y Measure Yes Problem over-time? No

Improve/ Innovate

Analyze No

In Control? Yes

No

Remove Known Special Causes

Process Capable? Yes

Yes

Control Yes

Process Capable?

Yes No Close Project

Close Project Good measurement system?

No Sustained Improvement?

Yes Close Project

Can goals be achieved with existing process?

No Remove Measurement System variation

No Resign process and/or product/ services offering

Figure 1-4: High-level DMAIC Process Flow

Summary Many view DMAIC as the foundation of Six Sigma. DMAIC is best used as an iterative problem-solving method to combat variation in an existing, steady-state process. Students of the quality and process improvement often start by learning the DMAIC approach because most other methodologies derive from its fundamental structure and concepts. Components of the standard DMAIC toolset also can be found in the tool suite of other Six Sigma approaches.

DMAIC

Define

27

This page intentionally left blank

Section 2 Lean and Lean Six Sigma The Lean and Lean Six Sigma approach expands Six Sigma concepts into a philosophy focused on velocity, value-add, and process flow. This section overviews Lean concepts and its high-level requirements, given the requirements define the appropriate deliverables, which dictate the tasks and the tool selection to aid the task. This section outlines the Lean standard toolset through the understanding of the tool-task-deliverables linkage to facilitate appropriate selection of a tool when referencing the “how to” tool articles in Part II of this book.

What Is the Main Objective of this Approach? The Lean method streamlines a process to its essential value-adding activities from the perspective of the customer. It drives efficient process performance by focusing on doing things right the first time. Lean practitioners scrutinize the flow of value-adding tasks by selecting those critical tasks to execute and those to eliminate. Paying customers define what is classified as value-add by indicating that they are willing to pay for it. A value-added activity is the process of physically changing an object per the customers’ requirements, and the task is done right the first time. In Lean terms, work is a value-added activity; waste (or non-value-added activities) is the opposite. The purpose is to improve speed of a continuously flowing process that operates on-demand (from customer-pull perspective) and with minimum to no waste. The pull-perspective refers to just-in-time production in that nothing gets produced until the customer orders it—providing what the customer wants, when they want it, and how many they want without any lead-time. The goal is to balance resources to customer demand—defined as “flow.” The Lean method reduces or eliminates wasteful activities, such as wait-time, unnecessary hand-offs and transportation, excess inventory, and overproduction. 29

30

Lean and Lean Six Sigma

Brief Description of Typical Applications Lean methods traditionally are applied to factory operations, production, and supply chain processes to trim out their non-value activities. However, the approach works well with administrative and other non-production areas. Lean Production or Just-in-Time (JIT) was popularized in Japan, post-World War II, to improve factory efficiency. The Japanese marketplace sought a great variety of automotive offerings, and Toyota developed and applied Lean concepts to improve production flexibility and efficiency to serve market demands better. The Toyota Production System challenged the idea of large lot production from Ford and introduced flexibility through diversified small lot production. Both Taiichi Ohno and Shigeo Shingo pioneered most of these concepts at Toyota in the 1970s. Toyota focused on systematically and continuously reducing waste (or “muda” in Japanese). Muda included product defects, overproduction, excess inventories, unnecessary processing, wasted movement of people, excess transporting of goods, and waiting. MIT coined the term “lean manufacturing” in the 1990s while studying Toyota’s manufacturing successes. Figure 2-1 shows how Lean improves the speed of a problematic process by eliminating waste, minimizing hand-offs, and reducing unnecessary activities. In this case, Step 3 is problematic (too close to the lower specification limit). Lean is able to remove Step 3 without jeopardizing quality and improve cycle time—the (former) Step 4 becomes the (new) Step 3 in the improved process. Step 1

Step 2

Step 3

Step 4

Problematic Process

Step 1

Step 3 (new)

Step 2

Process After Lean

Cycle Time

Figure 2-1: Lean’s Impact on Cycle Time

Some compare Lean and Six Sigma differences as a matter of a degree of complexity, size, and scale of the initiative. Lean is the simpler of the two. The optimal approach integrates Lean with Six Sigma to add not only more tools to the arsenal, but also a different perspective. Six Sigma offers a big picture context, balancing the customer and business requirements and criteria to evaluate overall impact of leaning, such that Lean initiatives no longer stay at the local level and potentially out of synch with the rest of the organization. For example, a common manufacturing mistake is to allow lean initiatives to outrank any other process voice, such that the process loses its ability to design and produce robust products.

What Requirement Determine s the Key Activitie s in this Approach?

31

See Also “The Anatomy of Variations in Product Performance” and “Lean Six Sigma for Fast Track Commercialization High Risk-High Reward, Rapid Commercialization: PROCEED WITH CAUTION!” articles, in Part III, “Best Practices Articles,” p. 777 and 835, respectively, for further elaboration on this point.

What Key Overall Requirements Define this Approach? The Lean approach was designed and structured to answer the following overall business questions: • Is the process delivering the most optimal value to the customer? • Is the process operating as efficiently as possible, absent of any unnecessary waste?

• Does the process strive for continuous improvement such that the process players (internal and partners) are empowered to adjust to changes or opportunities for improvement?

What Requirement Determines the Key Activities in this Approach? The preceding key business questions determine the Lean architecture. Figure 2-2 depicts a high-level Lean workflow: Analyze Non-Value Add

Map Current Flow

ID Customer Value

No

No Optimal Value?

Will customer pay for this?

Yes

No

Close Project

Yes

Label as ValueAdd

Label as NonValue-Add

Critical to business? Yes Label as NonValue-Add Business

Standardize and Stabilize Process

Eliminate Waste Yes No Pulled System?

No Implement Just-in-Time

Sustained Improvement?

Yes Close Project

Figure 2-2: High-Level Lean Process Flow

Lean often is not associated with a rigorous method, but rather its concepts are often integrated in with other methodologies, such as DMAIC or DFSS. However, a pure Lean approach generally follows the following high-level approach. Table 2-1 shows the link between the high-level business requirements with the Lean approach.

LSS

• Is the process output produced only at the pull of the customer?

32

Lean and Lean Six Sigma Table 2-1: Lean Requirements-Step Linkage Requirements

Resulting Key Activities

Is the process delivering the most optimal value to the customer?

1. Identify customer value. • Collect and summarize exactly what they would be willing to pay for. 2. Map the current process workflow. • Diagram the current process. • Group activities into either value-add or non-value-add categories. • Identify value currently delivered.

Is the process operating as efficiently as possible, absent of any unnecessary waste? Is the process output produced only at the pull of the customer?

3. Analyze the non-value-add activities and determine problem(s). • Look for bottlenecks (i.e., Theory of Constraints). • Group into non-value categories and problem(s) identified. • Apply appropriate waste elimination strategies and tools. 4. Eliminate the appropriate waste. • Apply appropriate Lean Six Sigma strategies for improvement. • Eliminate just-in-case activities and resources. • Map and document the future state process. 5. Standardize the process and develop a transition plan. • Work with all the process players, including suppliers, to collaborate and gain agreement on how to achieve improvements. • Document changes.

Does the process strive for continuous improvement such that the process players are empowered to adjust to changes or opportunities for improvement?

6. Validate the process and inspect for stabilization. • Create a system that automatically adjusts to relevant internal and external environmental changes.

What Tools Are Aligned to Each Step of the Process? Given the preceding high-level activities, the following series of tables summarize the subsequent tool-task-deliverables combination associated with each individual step within the Lean approach. The detail behind how to use each tool can be found in Part II, “Six Sigma Tools and Techniques: Choosing the Right Tool to Answer the

What Tools Are Aligned to Each Step of the Proce ss?

33

Right Question at the Right Time,” of this book. Table 2-2 summarizes the tool-task-deliverable linkage for Lean’s first requirement. Table 2-2: Lean Tools-Tasks-Deliverables for Requirement 1 Requirement: Is the process delivering the most optimal value to the customer? Step 1: Identify customer value. Tasks

Candidate Tools and Techniques

Customer Value documented

• Define exactly what • VOC Gathering the customer would techniques be willing to buy • Takt time (customer (both current and demand rate) potential customers); required volume • Determine the desired average completion rate

Table 2-3 summarizes the tool-task-deliverable linkage for Lean’s Requirement 2. Table 2-3: Lean Tools-Tasks-Deliverables for Requirement 2 Step 2: Map the current process workflow. Deliverables

Tasks

Candidate Tools and Techniques

Value Stream process flow documented

Diagram the current • Detailed process map customer value stream • Value Stream mapping from two perspectives: • Spaghetti map goods/services and the production process (and its players)

Value and Nonvalue-add activities documented

• Measure and collect • Value Stream matrix time associated with • Detailed process map or each activity in the swim-lane process map process • Identify and categorize activities

Identify value currently delivered

Calculate process • Total Lead Time capacity and Takt time • Process Cycle Efficiency • Workstation Turnover Times

LSS

Deliverables

34

Lean and Lean Six Sigma

Table 2-4 outlines the tool-task-deliverable linkage for Lean’s third requirement. Table 2-4: Lean Tools-Tasks-Deliverables for Requirement 3 Requirements: • Does the process operate as efficiently as possible, absent of any unnecessary waste? • Is the process output produced only at the pull of the customer? Step 3: Analyze the non-value-add activities and determine problem(s). Deliverables

Tasks

Candidate Tools and Techniques

Non-value-adds • Categorize non• Categories of waste categorized and problem(s) value-add activities • Theory of Constraints identified into business and pure waste • Determine the type of problems that exist Lean strategy developed

• Develop and apply • appropriate flow, • balance, capacity, • waste elimination, • and process pull • strategies • • Validate to ensure that • the faster processes, • less waste, continuous flow, on-demand • improvements are • achieved. [that is, • reduced cycle time, reduced product • defects (reworks and • maintenance), reduced • changeover time, • reduced equipment • failure, reduced • accidents, and so on.] •

Value stream matrix Detailed process map Spaghetti map Flow Manufacturing Multi-process Operation Visual Control; 5S Constraint management Level loading; Leveled Production Changeover Manpower Reduction Pull systems; Kanban; Just-in-Time Flexible Process Lot size reduction Autonomation (Jidoka) Quality Assurance (QA) Standard Operation Maintenance and Safety Cost/Benefit analysis

What Tools Are Aligned to Each Step of the Proce ss?

35

Step 4: Eliminate the appropriate waste. Tasks

Candidate Tools and Techniques

Lean strategies implemented

Select and implement • Prioritization matrix, or Lean strategies and Solution Selection supporting implemenmatrix tation plans • Implementation plan • Communication plan • Training plan

Improved state documented

Document standard process and procedures [this should be state and highlight improvements (or changes to current state)]

• Standard Operating Procedures, revised • Should be value stream matrix • Should be process map • Should be Standard Operating Procedures • Control plan • Risk management plan

Step 5: Standardize the process and develop a transition plan. Deliverables

Tasks

Candidate Tools and Techniques

Improved process stabilized

• Work with all the process players to collaborate and gain agreement on how to achieve improvements, including suppliers • Document any changes or modifications to drafted plans

• Implementation plan • Communication plan • Should be value stream matrix • Should be process map • Should be Standard Operating Procedures • Control plan • Risk management plan

Table 2-5 outlines the tool-task-deliverable linkage that address Lean’s Requirement 4.

LSS

Deliverables

36

Lean and Lean Six Sigma Table 2-5: Lean Tools-Tasks-Deliverables for Requirement 4 Requirement: Does the process strive for continuous improvement such that the process players (internal and partners) are empowered to adjust to changes or opportunities for improvement? Step 6: Validate the process and inspect for stabilization. Deliverables

Tasks

Candidate Tools and Techniques

Process improvements validated and stabilized

Create and/or ensure • Control plan a system that automat- • Risk management plan ically adjusts to • Kaizen (See “What Does relevant internal and Kaizen Mean?” later in external environmenthis chapter.) tal changes

What Are Some of the Key Concepts that Characterize this Approach? This next section provides an overview of some key Lean concepts that distinguish it from the other Six Sigma approaches. It covers topics such as velocity (or speed), value-add, wastes, Kaizen (or continuous improvement), and variations on the Lean philosophy.

How Is Value-add Defined? Value is defined as what a customer wants or needs and is willing to pay for. Value-added describes those activities that add value (as defined by the customer) to the product (or service). These activities often are thought of as in-process activities done within the context of producing or manufacturing the final customer offering.

How Do You Know if the Process Activities Are Value-added? Value-added activities should withstand the 3C litmus test: 1. Change—The activity changes the product or service. 2. Customer—The customer actually cares about the activity’s outcome (its impact on the overall offering). 3. Correct—The customer’s perspective determines if an activity is executed, and its outcome is produced correctly the first time—no rework.

What Are Some of the Key Concepts that Characterize this Approach?

37

What Is a Non-value-add Activity? Lean defines work as valued-added activities (designing and building goods and services) for which a customer is willing to pay. Work is made up of those activities that create value from the customer’s perspective. The opposite of work is non-value-added activities, or waste. Waste is often a symptom, rather than a root cause, of a problem. Waste simply leads to the root causes within the system (process, goods, and/or services).

1. Business Non-value-add—Those items required for the business to operate efficiently, such as legal requirements, regulatory requirements, recording financials, and maintaining an intellectual capital management system. 2. Non-value-add—Items that lack value to both the external customer and the business; hence the customer is unwilling to pay for it, and it does not add value to the final product (or outcome). Their various types of waste include Overproduction*—Producing parts ahead of schedule (before a customer requests or needs it) while other items in the process wait; working on unneeded parts (the wrong parts at the wrong time); evidenced by producing too many parts or finished goods or producing them too early Inventory Space*—Excess space consumed by shelving, floor space, excessively wide aisles, bins, filing cabinets, or files that house accumulated in-process or finished goods, including parts waiting for rework or scrap storage Waiting Time*—People unnecessarily waiting, stalled in the workflow due to either shared equipment, unbalanced work activities (operations), or waiting for decisions, approvals, or inspections Transportation*—Moving parts or objects unnecessarily (including papers or files); excess travel distance, which also can waste time and equipment Motion*—Excess, unnecessary, or non-valued people activities such as searching, walking, sitting, choosing, copying, stapling, sorting, climbing, bending over, and lying down

LSS

Are There Different Types of Non-value-added Activities? Non-value-added activities fall into two categories:

38

Lean and Lean Six Sigma

Unnecessary Processing*—Unnecessary operations, steps (including inspection or approvals) and complexity (at times excessive documentation) Defects*—Repairs and rework needed to get something to function properly, which slows the process flow and reduces first-pass throughput yield Materials—Scrap or excess ordering of raw materials Idle Material—Material that waits or sits in inventory Energy—Wasted power or people energy (closely related to motion) Excessive Labor—Too many workers in the process causing inefficient operations. Safety hazards—Unsafe work environments Management Wastes—Failure to orchestrate and unleash the inherent capabilities of the organization (people, process, infrastructure, and technology) A memory-jogger acronym often used in the United States to remind people of the different types of waste is known as DOWNTIME, which stands for: • Defects • Overproduction • Waiting • Non-utilized talent • Transportation • Inventory • Motion • Extra processing

Purists of the Japanese Lean concepts recognize only seven key wastes (defects, overproduction, waiting, transportation, inventory, motion, and excess processing) and drop non-utilized talent. Utilizing talent to its fullest is an ingrained part of the Japanese culture, so they do not need to be reminded of this potential waste. Other cultures, however, benefit from the reminder to appropriately utilize or leverage the people talent. * The first seven categories listed here that are labeled with the asterisk (*) represent the traditional Lean Seven Wastes from Japan.

What Are Some of the Key Concepts that Characterize this Approach?

39

Moreover, Japanese often group things in sets of seven, and DOWNTIME is an eight-letter acronym. Regardless of which set of wastes is used, one can observe these wastes, seeing evidence of them in the workplace. Lean practitioners constantly look for waste to see if it exists. As a result, practitioners have expanded the list of wastes to include managerial wastes, which include • Ownership Waste—Lack of accountability, authority, or empowerment to perform responsibilities in full • Focus Waste—Dispersed energy working on different sets of critical issues due to inconsistent alignment of management and employees

• Discipline Waste—Failure to maintain the process, performance, and behaviors needed to achieve standards (or specifications)

What Are the Key Lean Strategies to Eliminate Waste? The appropriate strategy for waste elimination depends on the scenario in which the waste exists. Customer requirements, the data depicting the waste, the root cause, the work environment, the process players, and the culture represent some of the factors to consider when developing a waste elimination strategy. Lean offers some generally accepted generic waste elimination strategies including • Waste elimination—Rework and loop-backs • Value Stream Mapping—Aids in visualizing the end-to-end flow

and identifies potential sources of waste. (See Also “Value Stream Analysis,” p. 727 in Part II) • Value Stream Analysis—Is a tool used to dissect the process flow and

analyze any waste by categorizing, defining its impact and identifying the countermeasure to it. (See Also “Value Stream Analysis,” p. 727 in Part II) • Work standardization—Tasks organized in the optimal sequence with the most effective conditions (the 5Ms): manpower, materials, methods (motion sequences, technical procedures, administrative procedures), machinery (equipment, tools) and measurements (quality requirements); plus uniform work conditions and workplace arrangements

LSS

• Structure Waste—Lack of infrastructure support (processes, reporting relationship, culture, rewards, and recognition) needed to focus on continuous improvement (Kaizen) or to gain ownership

40

Lean and Lean Six Sigma

• Inventory reduction or elimination—Including Work In Process (WIP) (using a pull system—potentially with a kanbans (Japanese for place card or sign) • Workload balancing—For example, constraint management, level loading, excess resources elimination) • Batch size reduction—Or Lot • Flexible process reconfiguration • Quality control • Complexity reduction—And segregation into similar levels • Linked value-add process steps—Contiguously placed within the workplace • 5S established—For good housekeeping, which are defined as — Sort (or tidiness), known as “Seiri” — Set in order (or orderliness, simplify, sequence), known as

“Seiton” — Shine (or sweep, cleanliness), known as “Seiso” — Standardized clean-up, known as “Seiketsu” — Sustain (or self-discipline), known as “Shitsuke” ■

Visual workplace established

■

Flat organization structure

■

Empowered employees

■

Teamwork established

■

Employee morale improved

What Is the Theory of Constraints? Theory of Constraints (TOC) describes a process for continuous improvement by removing bottlenecks in a process with limited production or throughput. The concept focuses on the identification of the system constraint. A process chart or a value stream map are useful tools to assist in finding bottlenecks. (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522 and “Value Stream Analysis” p. 727 in Part II.) The concept was introduced by E. Goldratt in two books: The Goal (1986) and Theory of Constraints (1990). He proposed three basic measures to be used in the evaluation of a system:

What Are Some of the Key Concepts that Characterize this Approach?

41

• Throughput—The rate which money is generated through sales

(measured as incoming money) • Inventory—The money locked up in the system as investments (or

purchases) that are intended to be sold in the future (measured as money stuck inside the company) • Operational expenses—The money spent to convert inventory into

throughput (measured as money going out of the company). This includes depreciation, scrap, carrying costs

What Does Kaizen Mean? Kaizen is a Japanese philosophy in which one strives for continuous improvement in all aspects of life. Literally, “kai” means change and “zen” means to become good. With respect to the workplace, Kaizen improvements are often thought of as constant, small, incremental improvements that occur as work happens. As you work, if you see something that would improve the process (or product) or make work easier, you are empowered to implement the modification. Kaizen improvements are viewed as a constant initiative to seek an optimal level of operation, rather than a radical change or innovation that would require the efforts of a Six Sigma project. Given the incremental adjustment approach of Kaizen, it may involve short-term projects (one to six-day timeframe), involving only those people involved in working in the process (or product). Often a Kaizen initiative starts with an impromptu conversation among empowered colleagues, who simply brainstorm about the improvements they could make. This conversation evolves into a self-directed work team that focuses on fixing the problem. They may continue to meet during breaks, before or after work, or they could take time out of work to take a more formal examination of how best to optimize the system. An important perspective of Kaizen is that the improvements are documented. The documentation often is in the form of updating Standard Operating Procedures (SOPs). Kaizen thrives in a workplace environment that encourages and supports empowered employees to maintain SOPs as part of an evergreen process that continuously reviews, refreshes, and renews. The approach, tools, and techniques the empowered employees follow could be the DMAIC approach or the classic E. Deming approach of PlanDo-Check-Act (PDCA). Common tools used by Kaizen teams include:

LSS

These measures are said to be more reflective of the true system impact than machine efficiency, equipment utilization, downtime, or balanced plants.

42

Lean and Lean Six Sigma

Affinity diagrams, Tree Diagrams, Process Decision Program Charts, Matrix Diagrams, Interrelationship Diagraphs, Prioritization Matrices, and Activity Network Diagrams.

Are There Any Lean Variations? There are two prevalent variations to the Lean method: 1. PDSA (Plan-Do-Study-Act) or PDCA (Plan-Do-Check-Act)—A mainstay of the Total Quality Management discipline, Edwards Deming popularized the PDSA model (sometimes called the PDCA cycle) to improve a process flow. This often is referred to a simplified foundation of other problem-solving methods. Original writings about this approach originated from Shewhart in the late 1930s when he was describing the power of prediction as an essential ingredient to using experimental data to build knowledge. Hence, Shewhart starts with original data, predicts an outcome, and gains knowledge from observing experiments. Shewhart proposed that since the new findings were based on experimental data, that the knowledge was probable and that the non-experimental circumstances only could be inferred. The approach starts with planning as the important groundwork that establishes the objectives and timeframe for a study (Plan). Next, the approach executes the plan (Do). Third, the study results are collected, checked, and analyzed, evaluating the hypothesis and drawing new conclusions (Study or Check). Last, based on the experimental data that provided the new knowledge, follow-on improvement actions are implemented (Act). Because this is a Plan cyclical model, the learnings from Act feed into another iterative Plan phase, as part of continuous improvement. Act Do Figure 2-3 depicts the cyclical nature of the PDSA model. Plan-Do-Study-Act also inspired early development of the DMAIC approach. Figure 2-4 shows the alignment of the two structures by step.

Study

Figure 2-3: PDSA Model

What Are Some of the Key Concepts that Characterize this Approach? P-D-S-A Continuous Improvement

D-M-A-I-C Problem-Solving and Process Improvement

Do

Plan

Define

Measure

Analyze

Study

43

Act

Improve

Control

Figure 2-4: Comparison of PDSA and DMAIC Models

Six Sigma examines a process to improve quality and reduce variation—variation from target or customer requirements, including variation reduction within a problematic step. Lean concentrates on process timing—overall cycle time, including the timing between process steps by removing non-value-added activities. Combining both Lean and Six Sigma produces a compounding impact on the improved state. Figure 2-5 depicts the impact of integrating the two disciplines on a given process. Step 1

Step 2

Step 3

Step 4

Problematic Process

Process After Six Sigma

Step 1

Step 1

Process After Lean

Step 2

Step 2

Step 3

Step 4

Step 3 (new)

Cycle Time

Figure 2-5: Lean and Six Sigma’s Impact on Cycle Time

Both concepts share similar views on customer-focus, process-centric work, and appropriate tools. Lean simply adds a deeper set of tools to eliminate waste between process steps handoffs. Often DMAIC provides a project the big picture view (what the customer values balanced by business values), process stabilization, and capability, while Lean introduces speed and flow concepts at a more detailed level. The

LSS

2. Lean Six Sigma—Adding Lean concepts—By incorporating Lean concepts into DMAIC, the project adds dimensions such as velocity, value-add, and flow to what the Six Sigma problem-solving method already offers.

44

Lean and Lean Six Sigma

Define-Measure-Analyze-Improve-Control structure still holds true for Lean Six Sigma projects.

Figure 2-6 shows an integrated, high-level, Lean Six Sigma process flow. Define

Measure

Analyze

Improve

Control No

Problem over-time?

No

No

In Control?

No

Remove Known Special Causes

Process Capable? Yes

Yes

Close Project Optimal Value?

Good measurement system? Analyze Non-Value Add

Yes No Close Project Remove Measurement System variation

Standardize and Stabilize Process

No Pulled Production?

No

No

Yes Yes Close Project

No Implement Just-in-Time

Yes

Yes Critical to business?

Map Current Flow

Will customer pay for this?

Eliminate Waste

Process Capable?

Yes

Yes

Label as ValueAdd

Label as NonValue-Add Business

No Label as NonValue-Add

Figure 2-6: High-level Lean Six Sigma Process Flow

Summary Lean concepts are an important part of the Six Sigma arsenal to combat variation. The Lean philosophy introduced a simple but powerful focus on waste elimination and continuous improvement to better meet customer requirements on-demand, sans any lead-time. It is an important cornerstone of the Six Sigma approach and has contributed invaluable tools and techniques to create flow, eliminate waste, and strive for continuous improvement.

Section 3 Design for Six Sigma (DFSS) Design for Six Sigma (DFSS) represents a portfolio of methods and tools that expands Six Sigma concepts to take a preventative approach by designing quality into a product (or process). This section overviews the DFSS concepts and its key methods. Next, the high-level requirements are discussed and the appropriate deliverables are defined, which dictate the tasks and the tool selection to aid the task. While the DFSS category applies to technical design applications, it also features a unique subset of tools applicable to general business applications. This section outlines the DFSS standard toolset, to build an understanding of its tool-task-deliverables linkage and facilitate in the selection of the appropriate business-oriented tools borrowed from DFSS when referencing the “how to” tool articles in Part II of this book.

DFSS Evolution and Approach Overview Design for Six Sigma (DFSS) defines a category of Six Sigma methods that evolved from the DMAIC problem-solving method. Businesses would realize great success in fixing problematic processes and/or product offerings when they first implemented DMAIC. Initial successes often prevail in the manufacturing area and across the board for cost cutting. However, after awhile, they find it increasingly difficult to maintain quality while continuing to reduce defects or variations and take costs out of the business, particularly if the process was poorly designed. This experience of diminishing success occurs more quickly in situations where the process or product is poorly designed. Eventually, modifying the existing process through the DMAIC method alone will fail to meet ever-demanding, often changing customer requirements. For example, additional cost reductions could rob the (process or product) design of its robustness, thereby jeopardizing the quality. Alternatively, a poorly designed process may demand some level of rework or scrap to uphold quality. Such defective processes drain resources of time and money attempting to deliver expected quality levels. In response to this phenomenon, DFSS emerged to design in quality. 45

46

De sign for Six Sigma (DFSS)

DFSS evolved to address the need for a redesign or new design—an innovation in response to a problem. If the process is incapable of meeting the desired customer specifications, it requires a redesign or an all together new design. The new design reduces sensitivity to variation and noise, which yields improved reliability, durability, and quality—a more robust design. The need for a redesign versus new design may be a matter of degree. Note that a redesign modifies an existing design, and a new design starts with a clean sheet. However, a new design also could apply only to a module or sub-component of a current design. The degree of difference between a redesign and new depends on the portion that the sub-assembly represents relative to the entire architecture.

DMAIC Versus DFSS Methods Relative to the classic problem-solving DMAIC method, the DFSS arena of Six Sigma is still maturing. As a result, two consequences have materialized: 1. A fuzzy delineation between DMADV (for reactive redesign) and CDOV (for clean-sheet, proactive) methods, causing the deployment and application to vary from company to company 2. The emergence of other DFSS variants The DMADV (Define-Measure-Analyze-Design-Validate) method aims to redesign a problematic process or product. The approach initially follows the first three steps of DMAIC and then deviates in the last two-steps by introducing Design/Redesign and Validate steps to gain the improvements needed. This approach prevents problems from happening through quality and robust design concepts. Figure 3-1 delineates the alignment of the DMADV and DMAIC steps: D-M-A-D-V Problem-Solving and Redesign of Process and Product

Define

Measure

Analyze

Design

Verify

D-M-A-I-C Problem-Solving and Process Improvement

Define

Measure

Analyze

Improve

Control

Figure 3-1: Comparison of DMADV and DMAIC Methods

If the situation calls for a clean-sheet design and commercialization, wherein the project team is starting from scratch, then the CDOV (Concept-Design-Optimize-Verify) method is preferred. CDOV takes on more

Key DFSS Concepts

47

of a proactive approach by creating a robust design to prevent the need for a redesign. Given its forward-thinking purpose, CDOV is part of a portfolio of Six Sigma methods that drive growth—top-line business growth through new revenue initiatives. Business people and academicians alike have proven repeatedly that fixing issues early on in a process or lifecycle saves money. DFSS, especially CDOV, builds on that notion and starts with a clean-sheet approach, addressing the early stages of an offering’s lifecycle—research, design, and development. DFSS differs from the classic problem-solving (DMAIC) methods that activate once the process moves into production and responds to problems that arise. Hence, the proactive-based DFSS methods cost a business less to deploy in the long-run based on an offering’s lifecycle expenses, as shown in Figure 3-2.

Relative Cost to Make Design Change

“Classic” Six Sigma Focus

DFSS Focus

Design

Development

Production

Product Lifecycle Stages over time Figure 3-2: Cost Comparison of DFSS and Classic Six Sigma Over Product Lifecycle

Key DFSS Concepts DFSS focuses primarily on growth through product development; hence, it relates typically to the technical, engineering, and manufacturing environments. Its main application addresses two main processes 1) the technical product/services development and commercialization and 2) the packaging design and assembly. In the product design world, George Box, Douglas Montgomery, and Genichi Taguchi have contributed crucial concepts to proactively improve both quality and ultimately customer satisfaction. Two such concepts are the Loss Function Model and Robust Design Principle.

DFSS

Research

48

De sign for Six Sigma (DFSS)

Taguchi’s Loss Function Model First, the concept behind the Loss Function model is to avoid economic loss by minimizing deviation of a product’s function (from its desired target) and optimizing its tolerance (often down to the parts-level). The amount of deviation from the target represented the tolerance in the system (or the delta). Calculating the amount that a product’s (or part’s) quality characteristic deviates from its target is called the quality loss. When the quality loss is zero (that is, exactly on target), then the quality characteristic is at the target value, indicating the absence of economic loss to society. Figure 3-3 illustrates how the model’s quality loss increases as the product (or output) moves off the (customer) target and/or as variation increases. As a result of Taguchi’s work, quality metrics aim to prevent potential cost of poor quality that may occur later in the product’s lifecycle. Target (Average and Actual curve)

Increased Drift Quality Loss Increased Variation

Critical Response Customer’s Lower Spec Limits (LSL)

Customer’s Lower Spec Limits (LSL)

Figure 3-3: Taguchi Loss Function Model

Robust Design Principle The second contribution, the robust design principle, promotes stability by minimizing both the impact of noise factors and the cost of failure on a product design. Taguchi incorporated the noise factors into the experimental design to determine the optimal design. This class of experimentation is referred to as Design of Experiments (DOE) techniques, used to select the critical parameters and their optimal levels such that they are insensitive to noise (or stresses). Noise sources fall into three categories: unit-to-unit variation, internal (such as deterioration, contamination, durability, and so on) and external environment. The noise factors can be applied to systems other than tangible manufactured products and apply

DMADV: Define-Measure-Analyze-De sign-Validate

49

to “everyday life.” By using the DOE techniques, Taguchi called attention to the importance of actually “manipulating” variables to witness (or observe) the effects in the critical response factors to demonstrate the cause-and-effect relationships. G. Box and D. Montgomery recommended the use of response surface methodology for robust design and optimization, wherein the experimental design identifies the system response to one or more variables and the response is graphed using geometric concepts.

DMADV: Define-Measure-Analyze-Design-Validate Let us explore in more depth the DMADV method, which is probably the most popular of the DFSS approaches.

What Is the Main Objective of the Approach? The DMADV (Define-Measure-Analyze-Design-Validate) builds off the DMAIC method as a data-driven Six Sigma method fundamentally to redesign a problematic process or product offering.

As part of the Design for Six Sigma (DFSS) branch of Six Sigma, DMADV is the most prevalent method. The objective of the DFSS category of methods is to develop a new or radically redesigned product, process, or service.

Brief Description of Typical Application The 5-step DMADV [pronounced “duh-MAD-vee”] method is often called the process innovation methodology. It was designed to solve a problematic process or product and/or service offering through an innovative redesign when the current design proves to be incapable of meeting customer requirements (or targets). Similar to the DMAIC approach, DMADV is designed to allow for flexibility and iterative-work, if necessary. As more is learned through the 5-step process, assumptions or hypotheses may be disproved, requiring the project team to revisit them and modify or to explore alternative possibilities. It builds on three fundamental principles: • Work is results-focused; driven by data, facts and metrics.

DFSS

The technical community (that is, engineering, manufacturing) was the first discipline to embrace this approach, but it is applicable in other areas of the business. This approach, traditionally, is to be applied to a problem with an existing, steady-state process or product and/or service offering. Sometimes this approach may be also used to design a new process or product when new requirements emerge.

50

De sign for Six Sigma (DFSS) • Work is project-based (short-term in nature, with length depending

on scope and complexity) and project-structured, versus an ongoing process. • Inherent combination of tools-tasks-deliverables linkage that varies

by step in the method.

The DMADV 5-Phase-Gate Structure The DMADV methodology uses a phase-gate structure. Phases generally are sequential; however, some activities may occur concurrently or may be iterative. A formal phase-gate review requires the completed set of respective deliverables prior to approval. Phase-Gate Reviews do occur sequentially. The DMADV five steps are Phase 1.

DEFINE the problem and/or the new requirements (or goals) of the design activity, including external and internal customer requirements. This involves a significant amount of VOC gathering activities to shape the design requirements.

Phase 2.

MEASURE the current process or product performance, gather appropriate data, and determine the customer’s needs and related specifications—what is critical-to-quality. This may involve a comparison between the current process and the new requirements.

Phase 3.

ANALYZE the current performance data to isolate the problem and identify the potential improvement options to better meet the external and internal customer requirements. Identify the cause-and-effect relationships between the problem(s) and key variables.

Phase 4.

DESIGN a new process so the problem is eliminated or new requirements are met. This may involve revamping (or developing a new) process, product, or service offering using predictive models, simulations, experimentation, and pilot testing to identify the optimal design to meet the goals.

Phase 5.

VALIDATE that the new process is capable of meeting the new process requirements. This may involve validating that the process or product design meets (little Y) targets for performance, robustness, and stability to ensure the overall requirement(s) can be met over time in a non-test (real-world) environment.

CDOV: Concept-De sign-Optimize-Verify

51

Four Components of the DMADV Method Similar to the DMAIC method, DMADV is based on the application of statistical process control, quality tools, and process capability analysis wherein the process or product exists and requires a redesign to fix the current problem. Some apply DMADV to a pure clean-sheet design effort; however, the CDOV method is better constructed for this purpose. To effectively implement DMADV, the method requires four components: • A measurement system (a gauge) of the process or product/service

offering in trouble. • Statistical analysis tools to assess samples of data. • An ability to define an adjustment factor(s) to redesign the process

or product/service offering back on target. • A control scheme to maintain the improvement or correction over

time by implementing a control plan with a monitoring system to audit the response performance against statistical control limits and defined action plans if needed.

CDOV: Concept-Design-Optimize-Verify Another DFSS approach, CDOV, utilizes the prevention approach of DMADV but promotes a unique structure that presumes a cross-functional design team.

DFSS

DMADV Duration A DMADV project may run from a relatively short duration (nine to twelve months) to several years depending on the complexity, size, scale, and scope of the product or process. Given the relatively long duration of a Design for Six Sigma (DFSS) project, the method uses a phase- structure to distinguish one step from another. At the completion of each phase, the project sponsor conducts a phase gate review with key project members and other key stakeholders to verify and approve the phase requirements and deliverables and to set (or clarify) expectations for the next phase. Phase deliverables include both “administrative” and “design” outputs. Administrative outputs involve items often associated with project management-type deliverables such as timeline and project life-cycle management, budget (resources and funding), scope (for example, scope creep; what is in and out of scope), communication, risk management, procurement, quality, people resources, and overall integration. Design outputs address the actual process, product, service offering, or information being developed. Albeit the approach may be iterative in nature, the project team should withhold starting a new phase until the sponsor approves the current phase deliverables.

52

De sign for Six Sigma (DFSS)

The structure addresses the design team’s technical community and incorporates a parallel integration with a sister-method uniquely focused on the marketing and business partners. CDOV introduces new concepts that expand the proactive-nature of DFSS to fuel the Six Sigma for Growth momentum.

What Is the Main Objective of this Approach? The CDOV (Concept-Design-Optimize-Verify) structure applies a flexible, adaptive approach to help a technical or engineering project team to design and commercialize a new product or services offering and/or new manufacturing system using a clean sheet design. CDOV can be used to design a new platform, extend product lines, and integrate new modular designs within an existing platform or product family. In addition, CDOV can be applied to a product improvement project. The CDOV structure compliments and integrates with a Six Sigma for Marketing (SSFM) method targeted at the other cross-functional team members, namely marketing and business professionals. This sister SSFM method is called UAPL (pronounced “YOU-apple”)—Understand-Analyze-Plan-Launch. While some companies successfully deploy CDOV as a standalone method, its power is best realized when integrated with the non-technical commercialization method. (Commercialization encompasses preparation for the marketplace and launch.) Together, CDOV and UAPL provide a common structure and language for multiple disciplines to work toward a common goal—the design, development, and commercialization of an offering.

The Commercialization Model Over and above the actual technical design and development of an offering, the technical community should contribute to the entire market-readiness process. The product commercialization process features requirements and deliverables from multiple disciplines of the cross-functional team. Too often companies fail to involve the technical arm of product development team in this portion of work. Commercialization success is defined at minimum as market acceptance and achieving a smooth launch throughout the customer value chain. Successful commercialization needs the technical community to partner with marketing and other functional disciplines to provide critical deliverables and complete the process. The different disciplines within a cross-functional team share some joint requirements and deliverables and also own complimentary requirements and deliverables. CDOV and UAPL orchestrate and integrate the cross-functional team’s work. Each method identifies the mutual dependencies in terms of primary roles, key deliverables, and timing. The deployment of a Stage-Gate process reinforces cross-functional integration and validation of requirements and deliverables. See Also “SSFM,” p. 67 for more details.

CDOV: Concept-De sign-Optimize-Verify

53

Brief Description of Typical Applications The 4-phase CDOV [pronounced “see-DOVE”] method, a product design, development, and commercialization method, integrates the Voice of the Customer with the Product and Manufacturing Process Design to generate top-line growth through new product (or services) launches. This approach relates to not only the technical/design development environment, but also the manufacturing, assembly, materials management, and supply chain development areas.

The Kano and New, Unique, and Difficult (NUD) Principles The Kano principles inspires CDOV in that it pursues the New, Unique, and Difficult (NUD) elements and incorporates them into the design to create customer-pull. Noriaki Kano, a Japanese engineer, developed a model at the product offering level, showing the relationship between customer satisfaction and a company’s ability to fulfill its customer requirements with the offering. If a firm’s offering includes NUD features and functionality, it would tend to “delight” its customers, particularly for those latent requirements. If the offering only included the bare minimum “must have” features and functionality, the customers would be dissatisfied. Finally, if a firm’s offering only included the “expected” features and funcCustomer tionality, Satisfaction declaring it “fit (High) for use,” then Delighters = the customer Fitness to Latent satisfaction Expectations would depend Satisfiers (Linear) on the number = Fitness to Use of items included in the Corporate Corporate offering (where Execution Execution more equates to (Requirement (Requirement Must Haves = Unfulfilled) Fulfilled) better) and Fitness to Standard probably would remain at a “neutral” level. Figure 3-4 is a Customer diagram of this Dissatisfaction three-prong Figure 3-4: The Kano Model Kano Model.

DFSS

CDOV was constructed as a tactical product design engineering process that parallels and integrates with the activities of the marketing and business areas also involved in the offering commercialization process. While the technical team follows the CDOV method, their counterparts in other areas of the business follow the UAPL (Understand-Analyze-Plan-Launch) approach. In concert, the multi-functional effort creates a robust offering ready for successful release to the marketplace. See Also “SSFM,” p. 67 for more details.

54

De sign for Six Sigma (DFSS)

Using the NUD technique, the “New” provides the customer “delighters”—the features, functionality, and services that neither your firm nor competition offers. The “Unique” describes a requirement that the marketplace offers (from either competition or a substitute product) but your firm does not. The “Difficult” may represent something complex in size, scale or scope, or technically challenging, but if provided, creates a competitive advantage and perhaps even a barrier to entry. In addition to the Kano model, CDOV leverages from other principles to be discussed in a subsequent section: robust design, critical parameter management, and systems engineering to prevent problems. CDOV has been successfully deployed as a means to design and improve integrated hardware, firmware, and software for optical-mechanical products, production systems, pharmaceutical, chemical, materials-based systems, and software products.

The CDOV 4-Phase-Gate Structure Similar to DMADV, the CDOV methodology uses a phase-gate structure. The phases generally are sequential; however, some activities may be iterative or may occur concurrently. Formal phase-gate reviews occur sequentially and require the completed set of respective deliverables to gain approval. The four-phase CDOV represents Phase 1.

CONCEPT (at a system-level) is developed to meet specific business goals and customer market segment requirements.

Phase 2.

DESIGN the concept’s subsystem, subassembly, and part-level elements.

Phase 3.

OPTIMIZE the design for robustness at both a sub-component and integrated system level.

Phase 4.

VERIFY the final design, production processes, and customer value chain capabilities are prepared for launch. Verify that 1) the design’s functional performance remains robust and stable and 2) the design, supporting processes, and capabilities all meet requirement(s) in a non-test (real-world) environment.

If the new concept is complex in size, scale, and scope, each of the four CDOV phases may subdivide into sub-phases to correspond with subassemblies, particularly for the Optimize and Verify Phases. As a result, these sub-phases would require “sub-phase-gate” reviews before assembling into an integrated system. Regardless of the phase, each phase-gate review examines the following five topics: 1) the business case (financials and assumptions), 2) technical design and development, 3) manufacturing and distribution preparation (including assembly, materials management, and supply chain development), 4) business and operating

CDOV: Concept-De sign-Optimize-Verify

55

constraints (including regulatory, health, safety, environmental and legal), and 5) post-launch readiness (including shipping, sales, service and support). The CDOV structure supports the PACE® Commercialization Process from PRTM. Figure 3-5 flowcharts the CDOV approach at a high-level view, along with its key requirements. Design

Concept

Optimize No Yes

No No VOC requirements understood and validated?

Design optimal and robust?

NUD Critical Parameters defined?

Yes No

Yes

Yes

Verify No

Sustained performance?

Yes No Yes

Process Capable?

Internal requirements understood?

Launch preparations completed

Close Project

Yes No

Figure 3-5: CDOV Model Process Flow

Comparison Between DMADV and CDOV As part of the DFSS family of methods, Six Sigma experts might argue that DMADV and CDOV share similar objectives but differ in emphasis. As practitioners become exposed to the various DFSS models, the promoted toolset with any one given approach borrows from another. As long as the practitioners comprehend that tool selection follows the understanding of requirements, deliverables, and tasks linkage, then the sharing and changeability of candidate tools only strengthens the execution. The key is that the right tool is selected at the right time to answer the right question. Figure 3-6 compares the DMADV and CDOV methods.

DFSS

Final concept design selected?

56

De sign for Six Sigma (DFSS) D-M-A-D-V Problem-Solving and Redesign of Process and Product

C-D-O-V New Offering and Process Development

Define

Measure

Analyze

Concept

Design

Design

Optimize

Verify

Verify

Figure 3-6: Comparison of DMADV and CDOV Methods

Given the fluidity between DMADV, CDOV and their variants and the fact that the toolset associated with each is similar, this book takes a simplistic approach to describe them. With the purpose of organizing a Six Sigma desk reference of tools, to aid in the appropriate selection and application of tools, the DFSS section next provides an overview of the structure of DMADV and CDOV and their variants and then treats them in aggregate to outline the candidate tools by key questions the tool helps to answer. The remainder of this section presents an integrated DFSS perspective, presuming that regardless of the approach being deployed, its main objective will involve a new design to some extent. Hence, the following information applies to all DFSS methods in a generic way; however, this information most closely aligns with the proactive CDOV approach.

What Key Overall Requirements Define this Approach? Recall that the public delineation among the various DFSS methods has blurred as this evolving field stabilizes. All DFSS methods focus on “design” at some level, either a proactive clean-sheet approach or reactive redesign. For purposes of simplicity, the remainder of this section will generalize and treat DFSS as one category focused on concept design and with a common set of requirements and corresponding tool-task-deliverable combination. This book’s appendix lists several resources to reference if more detail is needed on the various approaches.

CDOV Prerequisites Before getting started on a Product design, development, and commercialization initiative, the technical team will need require several inputs. According to Clyde M. Creveling, contributing author of several books, including most recently Six Sigma in Technical Processes and Design for Six Sigma in Technology and Product Development, the list of CDOV prerequisites entails the following:

What Key Overall Requirements Define this Approach?

57

• Business goals defined and a strategy on how to meet them. • Target market segments identified. • Customer requirements (VOC) and “Voice of Technology.” • Product/Offering portfolio plan and how the specific market segment requirements are fulfilled by the product-line (or product family). • Technical strategy defined to address the product-line strategy. • New platform and subsystem technology developed and certified.

The first four inputs probably come from the Business strategy and/or marketing and Offering Portfolio Renewal process (using perhaps the IDEA method). Research & Technology Development (R&TD), working in part with the Business strategy in the Offering Portfolio Renewal process, typically would provide the last two inputs (using perhaps the I2DOV method). The overall requirements that drive a concept design initiative entail the following: • What is the best concept design based on the complete set of requirements (both market segment and business)? • What are the design’s New, Unique, and Difficult critical parameters?

• Can the performance be sustained in non-test environments (including not only the system design, but also all the supporting postlaunch functions)?

What Requirement Determines the Key Activities in this Approach? Again, for purposes of simplicity, the remainder of this section overviews DFSS as one category using a high-level common set of requirements and corresponding tool-task-deliverable combination. Table 3-1 uses the CDOV phases to tie to the overall requirements and lists the corresponding DMADV phases (for redesign) in the parentheses when they differ.

DFSS

• Are the design and its component-levels robust?

58

De sign for Six Sigma (DFSS) Table 3-1: CDOV Requirements-Phase Linkage (with Corresponding DMADV Phases) Requirements

Resulting High-Level Phase (Corresponding DMADV Phase)

What is the best concept design based on the complete set of requirements (both market segment and business)?

1. CONCEPT (and Define, Measure, Analyze)

What are the design’s New, Unique, and Difficult critical parameters?

2. DESIGN

• Define concept requirements based on the current VOC and VOB requirements. • Identify certified technologies ready for market use. • Design concept as a high-level system.

• Design subsystem, subassembly, and part-level elements of system.

Are the design 3. OPTIMIZE (and Design) and its compon• Develop robust subsystems and subassemblies ent-levels robust? and evaluate performance. • Integrate the robust subsystems and subassemblies as a complete system and evaluate performance. • Conduct initial system reliability assessment. Can the perfor4. VERIFY mance be sustained • Verify the capability of the product design in non-test envifunctional performance. ronments (includ• Verify the capability of production assembly and ing not only the manufacturing processes and the extended supply system design, chain, customer value chain, service, and support but also all the organizations. supporting post-launch functions)?

Which Tools Are Aligned to Each Step of the Process? Given the previous high-level phases, the following set of tables summarizes the subsequent tool-task-deliverables combination associated with each of the four phases in the CDOV approach, which is transferable to the DMADV approach. Table 3-2 summarizes the Concept Phase. Part II of this book overviews the DFSS tools that are commonly applied to the marketing, sales, customer value chain, and supporting infrastructure and business areas. For a more in-depth understanding of

What Key Overall Requirements Define this Approach?

59

the DFSS tools applied to the technical environments, please refer to Part IV, “Appendixes,” for additional DFSS resources. Table 3-2: Concept Phase Tools-Tasks-Deliverables (Including Corresponding DMADV Elements) Phase 1: CONCEPT (and Define, Measure, Analyze)—What is

C

D

O

V

the best concept design based on the complete set of requirements (both market segment and business including the subsystem, subassembly, and

High-level Deliverables

High-level Tasks

Candidate Tools & Techniques

• Voice of Business requirements understood • VOC data documented as New, Unique, Difficult (NUD) requirements

Work with marketing and other business colleagues to categorize and prioritize NUD, easy, common, and old requirements.

• Business case • VOC Gathering techniques • KJ Analysis • Market segmentation analysis • NUD • CTQs

System functional concept models designed

• Create product or system-level HOQ. • Conduct competitive product benchmarking. • Generate product or system-level requirements. • Generate systemlevel concepts. • Test and analyze various concepts.

• Concept generation techniques (e.g. TRIZ, Brainstorming, Brain writing) • Design techniques— architecture, platform, modular. • House of Quality (HOQ) • Functional Flow Diagramming and Critical Parameter Mapping • Competitive benchmarking • Taguchi Noise Diagramming and System Noise Mapping • Taguchi Stress Testing • Highly Accelerated Life and Stress Testing • Reliability modeling • Monte Carlo simulation • Multi-vari studies

continues

DFSS

part-level elements of system design)?

60

De sign for Six Sigma (DFSS) Table 3-2: Continued High-level Deliverables

High-level Tasks

Candidate Tools & Techniques

Final concept selected

Evaluate test data and Pugh’s Concept Evaluation select concept model. Process

Technology transfer control plans documented

• Develop reliability requirements and initial model. • Develop system FMEA.

• Design and Process FMEA • Statistical Process Control and Capabilities Studies (for both design functions and process outputs)

Next steps planned

Create DESIGN Phase project plan and risk analysis.

• Project Management tools, including Monte Carlo simulation of project cycle time • FMEA

Table 3-3 summarizes the tool-task-deliverables linkage for the Define Phase of CDOV. Table 3-3: Design Phase Tools-Tasks-Deliverables (Including Corresponding DMADV Elements) Phase 2: DESIGN—What are the design’s New,

C

D

O

V

Unique, and Difficult critical parameters? (Including 1) Concept requirements based on requirements, 2) Certified market-ready technologies, and 3) high-level concept design.) High-level Deliverables

High-level Tasks

Candidate Tools and Techniques

VOC data validated as New, Unique, Difficult (NUD) details at a sublevel

Work with marketing and other business colleagues to categorize and prioritize NUD, easy, common, and old sub-level requirements.

• VOC Gathering techniques • KJ Analysis • Market segmentation analysis • Business case • NUD

Sub-level models for functionality and reliability designed

• Gather Certified Sub-level technologies from Research and Technology Development process.

• Concept generation techniques (e.g. TRIZ, Brainstorming, Brain writing) • Design techniques— architecture, platform, modular

High-level Deliverables

High-level Tasks

Candidate Tools and Techniques

• Model sub-level alternatives, test and analyze.

• House of Quality (HOQ) • Functional Flow Diagramming and Critical Parameter Mapping • Competitive benchmarking • Taguchi Noise Diagramming and System Noise Mapping • Taguchi Stress Testing • Highly Accelerated Life and Stress Testing • Reliability modeling • Monte Carlo simulation

Final sub-level design selected

Evaluate results and • Pugh’s Concept select sub-level design. Evaluation Process

Baseline design documented under nominal conditions, including Critical Adjustment Parameters (CAPs) and reliability plans

• Develop sub-level reliability plan. • Develop sub-level FMEA.

Next steps planned

Create OPTIMIZE • Project Management Phase project plan and tools, including Monte risk analysis. Carlo simulation of project cycle time • FMEA

• Design and Process FMEA • Statistical Process Control and Capabilities Studies (for both design functions and process outputs) • Design for manufacture and assembly • Value engineering and analysis • MSA • CPM • Engineering methods and math modeling • DOE • Descriptive and inferential statistical analysis • ANOVA data analysis • Regression and empirical modeling • Design capability studies

61

DFSS

What Key Overall Requirements Define this Approach?

62

De sign for Six Sigma (DFSS)

Table 3-4 summarizes the tool-task-deliverables linkage for the Optimize Phase of CDOV. Table 3-4: Optimize Phase Tools-Tasks-Deliverables (Including Corresponding DMADV Elements) Phase 3: OPTIMIZE (and Design)—Are the design and its component-levels robust? (Including 1) Robust subsystems

C

D

O

V

and subassemblies, 2) Integrates the robust system, 3) Initial system reliability assessment.) High-level Deliverables

High-level Tasks

Candidate Tools and Techniques

Robustness, adjustability, • Review and finalize and reliability documenthe critical functional ted for each sub-level responses. and integrated system • Develop noise diagram and map. • Conduct robustness experiments. • Conduct and analyze reliability/ capability evaluations.

• Noise diagramming and mapping • Measurement System Analysis (MSA) • Design of Experiments (screening and modeling) • Analysis of Means (ANOM) • ANOVA • Baseline signal-to-noise characterizations • Taguchi methods for Robust Design • Analysis and Empirical Tolerance Analysis • Additive experiments and modeling • Regression • Capability studies

Risk profile and assessment documented for each sub-level and integrated system

Analyze data, build predictive model and run verification experiments.

• Statistical process control (SPC) • Reliability analysis life testing • Design FMEA

Critical parameters documented at the sub- and integrated system level

Document critical functional parameter nominal set points.

• Design capability studies • Critical Parameter Management (CPM)

What Key Overall Requirements Define this Approach? High-level Deliverables

High-level Tasks

Candidate Tools and Techniques

Next steps planned

Create VERIFY Phase project plan and risk analysis.

• Project Management tools, including Monte Carlo simulation of project cycle time • FMEA

63

Table 3-5 summarizes the tool-task-deliverables linkage for the last CDOV phase, Verify. Table 3-5: Verify Phase Tools-Tasks-Deliverables (Including Corresponding DMADV Elements) Phase 4: VERIFY—Can the performance be sustained in

C

D

O

V

non-test environments (including not only the system design, but also all the supporting post-launch functions)? This includes 1) the capability of the product design functional performance and 2) the capability of production assembly and manufacturing processes, as well as the extended supply chain, cus-

High-level Deliverables

High-level Tasks

Candidate Tools and Techniques

Sub-level and system capabilities documented

• Conduct final tolerance design and document. • Verify product design meets all requirements.

• MSA • Analysis and Empirical Tolerance Design • Worst case analysis • Root Sum of Squares analysis • Monte Carlo simulation • DOE • ANOVA • Regression • Multi-vari studies • Design capability studies • System noise mapping • System-level sensitivity testing • Nominal system CFR • Stress case system

Sub-level and system design reliability growth

Evaluate system performance and reliability under nominal conditions.

Reliability assessments

continues

DFSS

tomer value chain, and service and support organizations.

64

De sign for Six Sigma (DFSS) Table 3-5: Continued High-level Deliverables

High-level Tasks

Candidate Tools and Techniques

Sub-level and system design risk profiles and assessment

Complete corrective actions on problems.

Design FMEA

Manufacturing, assembly, and supply chain capability

• Establish statistical process control (SPC) for critical-tofunction components and critical functional responses. • Build product design verification using production parts

• • • • • • • • • • • • • • • • • •

SPC CPM MSA Analysis and Empirical Tolerance Design Worst case analysis Root Sum of Squares analysis Monte Carlo simulation DOE ANOVA Regression Multi-vari studies Design capability studies System noise mapping System-level sensitivity testing Nominal system CFR Stress case system Kaizen DMAIC and Lean

Critical parameter management database

Develop transfer plan • CPM for the critical • Competitive benchmarkparameter database ing studies for production, supply • SPC chain, and service organizations.

Next steps planned

Create launch plan and risk analysis.

• SPC • CPM • Project Management tools • FMEA

What Are the DFSS Variations? Design for Six Sigma encompasses a portfolio of design methods. The origin of each approach stems from the key business requirement that it addresses and the timing of its evolution. Some grew out of

Summar y

65

DMAIC-based projects, some from a deep understanding of engineering techniques and principles. What they all have in common is the objective of designing quality into the design of a product or process. Other DFSS variations of note over and above DMADV and CDOV include the following four models: • DMEDI (Define-Measure-Explore-Develop-Implement)

Phase 1.

DEFINE the problem or new requirements.

Phase 2.

MEASURE the process and gather the data that is associated with the problem or new requirements.

Phase 3.

EXPLORE the data to identify cause-and-effect relationships between key variables.

Phase 4.

DEVELOP (or DESIGN) a new process so the problem is eliminated; measured results meet new requirements.

Phase 5.

IMPLEMENT the new process under a control plan.

• PIDOV (Plan-Identify-Design-Optimize-Validate)

[Reference: Design for Six Sigma Statistics, by Andrew Sleeper, McGraw-Hill, New York, 2006; ISBN 0-07-145162-5.] • ICOV (Identify-Characterize-Optimize-Verify) • IIDOV or I2DOV (Invent/Innovate-Develop-Optimize-Verify) A technology development process used in a Research and Development environment, wherein the goal is to prepare technology for commercialization.

Summary Design for Six Sigma (DFSS) refers to a category of methods and tools that designs quality into a product (or process). While DFSS generally applies to technical design applications, it also features a unique subset of tools applicable to general business applications. The DFSS standard toolset contains invaluable tools applicable to non-technical scenarios as

DFSS

This method closely aligns with the DMADV approach as a redesign technique and incorporates elements of Lean Six Sigma. It uses a slightly different vocabulary from DMADV and adds tools from the Lean methodology to ensure efficiency or “speed.” DMEDI phase-gates include

66

De sign for Six Sigma (DFSS)

well. General Six Sigma practitioners should include the business-oriented DFSS tools and techniques in their portfolio of candidate tools; hence, many of the applicable tools can be found in Part II of this book, the “How To” utilize tools section.

Section 4 Six Sigma for Marketing (SSFM) Six Sigma for Marketing (SSFM) represents a portfolio of methods and tools that expands Six Sigma concepts to take a preventative approach primarily for the non-technical community. It complements the DFSS design methods, but also encompasses the business processes before and after product design and development. SSFM spans across an organization’s processes, starting with its strategic planning front-end to deciding what to offer, through its offering design and development, through its marketing, selling, and supporting of offerings and ending up with the discontinuance of an offering. SSFM methods benefit an organization taken individually or in its entirety. This section overviews the three SSFM methods and each of their high-level requirements and appropriate deliverables, which dictate the tasks and the tool selection to aid the task. This section outlines the SSFM standard toolset to build an understanding of its tool-task-deliverables linkage and facilitate in the selection of the appropriate tools when referencing the “How To” tool articles in Part II of this book.

What Is the Main Objective of the Approach? Six Sigma for Marketing (SSFM) defines a category of Six Sigma methods applied to marketing and sales to drive growth. In addition, SSFM concepts can be applied to other general business functions (such as finance and customer service and support). SSFM is the newest category of Six Sigma. It builds off the fundamental principles of both Lean Six Sigma and Design for Six Sigma (DFSS) and their respective tools. Its focus is proactive in nature, to drive growth in a balanced way, rather than simply cutting costs. SSFM concepts concentrate on lead indicators of issues, to incorporate preventative mechanisms as part of ongoing operations. SSFM fills a Six Sigma void that both Lean Six Sigma and DFSS have had difficulty filling. The marketing and sales portion of a business traditionally tend to have unique work approaches and language that did not 67

68

Six Sigma for Marketing (SSFM)

adapt well to the conventional cost cutting and revenue growth approaches. SSFM, albeit in its infancy, has begun to yield benefits for companies such as 3M. At a recent Six Sigma conference, 3M claimed that its SSFM efforts have taken longer to produce the benefits but have positively impacted the bottom-line profit far greater than the classic DMAIC initiatives. In fact, its benefits continue given the ongoing nature of the revenue annuities.

Brief Description of Typical Application Recall that SSFM is the newest category of Six Sigma. Initially, SSFM addresses the three core business processes that marketing, as a functional discipline, typically supports: • Strategic Planning—to develop, manage, and refresh a portfolio of offerings. This activity defines the initial product concept, position in the portfolio, and high-level pricing targets. • Offering Development—to design, develop, commercialize and launch a new offering (including products, services, and information) often as part of a cross-functional team. This activity refines the product, its position and pricing, and begins to define launch promotions. • Operations Management—to manage a launched offering throughout the customer value chain, involving the go-to-market, customerfacing functional partners such as marketing, sales, telemarketing, services (for example, consulting, field engineering, or customer training), service and maintenance, customer administration, customer support centers, and customer financing services. This set of activities includes continued refinement of product positioning and pricing, as well as new product promotions. SSFM, when applied to each of these three areas, applies a unique method for each process—IDEA ((Identify-Define-Evaluate-Activate) for strategic planning, UAPL (Understand-Analyze-Plan-Launch) for offering development, and LMAD (Launch-Manage-Adapt-Discontinue) for operational management. The three SSFM methods define the work that marketing performs in each area and define how that work links together in a progressive closed-loop fashion, as shown in Figure 4-1.

Brief De scription of Typical Application

69

Strategic Marketing Processes

Operational

Tactical

Figure 4-1: The Strategic-Tactical-Operational Triangle

The Tactical Planning and Portfolio Renewal Process A business’ tactical practices entail its Product and/or Services Commercialization process. This process defines, develops, and “readies” a business’ offering for the marketplace. The industry, market segment, and size/scale/complexity of the offering dictate the number of the functional disciplines involved in this process and the amount of time it spans. The timeframe ranges from several months to several years. A business usually manages this process by establishing a unique project team to develop a single product or service from the portfolio of opportunities. At a minimum, two types of disciplines are needed—the “technical” functions to drive content and the “customer-facing” functions. The “technical” experts develop the offering and may include engineering, research, and manufacturing. The “customer-facing” disciplines represent those roles along the value chain that interface with a business’ customer or client, such as marketing, sales, services, and customer support. In the Commercialization process, marketing may represent the “customer-facing” touchpoints throughout the process, and may bring in the other functional areas toward the conclusion of the process in preparation for “hand-off” to ongoing operations.

SSFM

The Strategic Planning and Portfolio Renewal Process The Strategic Planning and Portfolio Renewal process defines a business’ set of marketplace offerings. This fundamental process refreshes an enterprise’s offerings to sustain its existence over time. Multiple functional disciplines may be involved in this process, or the enterprise may limit this work to a small set of corporate officers depending on the size of the enterprise and scope of its offerings. This process generally calls for a cross-functional team comprised of finance, strategic planning, marketing, and sometimes research, engineering, sales, service and customer support. Some businesses, with a unique strategic planning department, may serve as a surrogate for the other various functional areas. If this is the case, the strategy office typically staffs people with various backgrounds (research, finance, and marketing). This process can span a year and should get refreshed on a regular basis. Portfolio planning and management are the foundation from which to build and grow a business. Our experience tells us that successful businesses have marketing play a key role in the Strategic Planning and Portfolio process.

70

Six Sigma for Marketing (SSFM)

The Post-Launch Operational Management Process The Post-Launch Operational Management process unifies the operational dimensions of a business along its customer supply chain. This process manages and coordinates a business’ offerings of products and services and all its supporting activities. The offering and go-to-market strategy dictates the variety of functional disciplines involved across the value chain. The various functions may include sales, service, support, customer financing, customer administration, customer training, other services, and third party partners. Again, marketing may serve a representative role, integrating the multiple functional areas as it manages the product-line (or offering) throughout its lifecycle. This process represents long timeframes (often years) depending on the lifecycle of a given offering (product or service). The natural flow of marketing work starts with strategic renewal of the offering portfolios to the tactical work of commercializing new offerings and finally to the operational work of managing the product and services lines in the post-launch sales, support, and service environment. Marketing professionals frequently overlook the fact that their contributions are part of a process (or set of related processes). They view their work as part of a program or project. However, marketing work is repeatable. Though the timeframe for repetitiveness may extend over a year or more, the work is nonetheless procedural in nature.

Note The American Society for Quality (ASQ) defines a process as “a set of interrelated work activities characterized by a set of specific inputs and value added tasks that make up a procedure for a set of specific outputs.” Most marketers would agree that strategic planning and launching a product meet this definition of process. The Six Sigma approach embraces a process view to communicate its structure and flow of interrelated tasks. Although it may seem unnatural to marketing professionals, the best way to describe Six Sigma for Growth is through a process lens.

The Strategic and Tactical process areas are internally focused; hence, we refer to them as inbound marketing areas. Yes, external data are critical to a successful portfolio definition and development and product commercialization. However, those processes’ outputs are earmarked for internal use. These processes’ outputs are not yet ready for external consumption. The outputs ready for prime-time market exposure are part of outbound marketing. The operational processes involving post-launch

Brief De scription of Typical Application

71

product marketing, sales, services, and support are customer-facing activities. Given the different customers of inbound and outbound marketing, the requirements for each differ. These requirements ultimately define the success (or failure) of the deliverables. Figure 4-2 depicts how these three marketing areas interrelate. The inner-circle shows the circular connection, and the outer-circle depicts the demarcation between inbound and outbound marketing.

Inbound Marketing Strategic Post-Launch Line Management and Sales

Portfolio Renewal

Marketing Processes

Operational

Tactical

Inbound Marketing Commercialization

Figure 4-2: Interconnectivity of Marketing and the Process Triangle

Outbound Marketing Outbound marketing is focused on customer-facing operations. It encompasses post-launch product line management across the value chain (sales and services—including customer support). Outbound marketing can create problems and waste by failing to develop the right data to make key decisions about managing, adapting, and discontinuing the various elements of the existing product and service lines. Outbound marketing

SSFM

Inbound Marketing How does SSFM apply problem-prevention to marketing processes? Inbound marketing focuses on strategic product portfolio definition and development and tactical product commercialization. Inbound marketing can cause problems by under-developing the right data needed to renew product portfolios. The data are needed to define specific, new product requirements, thereby directing commercialization activities. Moreover, inbound marketing data define launch plans, which determine downstream operational success. Marketing can design and launch an inappropriate mix of products. As a result, a business could miss the growth targets promised in the business cases that were supposed to support the long-term financial goals of the company.

72

Six Sigma for Marketing (SSFM)

also could fail to get the right information back upstream to the product portfolio renewal teams. They need to renew the portfolio based upon real, up-to-date data and lessons learned from customer feedback and the marketing and sales experts in the field. The importance of the comprehensive, closed-loop Strategic-TacticalOperational scope provided the structural underpinnings used to create the unique Six Sigma methods for marketing. Each of these arenas has a flow of repeatable work—a process context that is quite different from the steps found in the traditional Six Sigma methods. However, the Six Sigma fundamental elements from the classic approaches have been maintained: tool-task linkage, project structure, and result metrics. This new work is made up of specific tasks that are enabled by flexible, designable sets of tools, methods, and best practices. The Strategic, Tactical, and Operational processes within an enterprise align with phases that can be designed to prevent problems—to limit the accrual of risk and enable the right kind and amount of data to help make key decisions. The traditional methods help get your processes improved, re-designed and under control. If the objective is to renew portfolios, commercialize products, or manage product lines, a different approach is required, which employs a different set of steps or “phases.” Each of these processes feature distinct phases in which sets of tasks are completed. Each task can be enabled by one or more tools, methods, or best practices that give high confidence that the marketing team develops the right data to meet the task requirements for each phase of work. A gate review at the end of a phase is commonly used to assess the results and define potential risks. Marketing executives and professionals find phase-gate reviews an important part of risk management and decisionmaking. In the post-launch environment, “gates” are replaced by key milestone reviews because you are in an ongoing process arena unlike portfolio renewal or commercialization processes that have a very strictly defined end date. Tackling the job of adding process discipline and metrics to the functions within a portfolio renewal process is by far the most difficult task when integrating Six Sigma into the three marketing processes. Adding process metrics and discipline to marketing functions in any environment is not a trivial matter. It must be done carefully, taking into account the unique culture that exists in marketing organizations. There is art and science here. Attention must be paid to how marketing teams interact with other teams within adjacent business processes—such as Research and Development (R&D), engineering, and the customer-facing partners. SSFM attempts the balance between marketing creativity with the necessary process discipline and metrics to add rigor and robustness to the deliverables, thereby improving the likelihood for marketplace success.

IDEA (Identify-Define-Evaluate-Activate)

73

A unique Six Sigma method for marketing was created for each of the three areas: Strategic, Tactical, and Operational. The method to guide marketing’s strategic work is called IDEA. A second approach was developed for its tactical work, called UAPL. A third method to direct marketing’s operational work is called LMAD. A brief description of each can be found in the following sections. An organization gains the biggest benefit from implementing all SSFM methods because they interrelate. However, depending on the organization’s maturity, size, complexity, needs, and resources, it could be quite an undertaking to deploy all three approaches at once. A staged deployment may start with the greatest need or the one easiest to establish. Any one SSFM method can be used successfully by itself, but the true power is in the aggregate of the three-part suite. An overall change management plan is helpful to define which method is appropriate and the execution approach that will ensure success.

IDEA (Identify-Define-Evaluate-Activate) Brief Description of Typical Applications In the Strategic marketing process environment, the unique SSFM method is known as the IDEA Process to renew and refresh the portfolio of offerings. This approach [pronounced “Idea” or “I-’dE”] works well at either a corporate, division, or department level to identify, develop, and fund an offering concept. It may trigger an initiative that lasts a few months to a year or more before funding an offering development and commercialization team to make it real. Hence, a portfolio renewal process can be viewed as either a project-based or an operationally-based process. Regardless of the time it takes to complete the renewal and refresh process, the IDEA method contains four distinct phases: Phase 1. IDENTIFY—markets, their segments, and the opportunities they offer

Phase 3. EVALUATE—portfolio alternatives against competitive portfolios, by offering Phase 4. ACTIVATE—ranked and resourced individual commercialization projects

SSFM

Phase 2. DEFINE—portfolio requirements and product portfolio architectural alternatives

74

Six Sigma for Marketing (SSFM)

Figure 4-3 diagrams the high-level flow of the IDEA approach and its key requirements.

Define

Identify No

VOC translated into NU D requirements?

Markets and segments of interest defined?

Yes

Yes No

Opportunities identified?

Evaluate No

Activate No

No Final concept project (s) selected and funded?

Portfolio Mix identified?

No

Yes No

Yes Candidate architectures identified?

Funding requirements determined?

Yes Yes

Activate an Offerings Development and Commercialization project

Close Project

Figure 4-3: High-level IDEA Model Process Flow

The design of IDEA is to be flexible, such that it can be tailored to serve as additional foundational structure to any existing strategic portfolio renewal and refresh process at a company-wide or division level. Its tools-tasks-deliverables combination can be incorporated to make the current approach more structured and robust. Its structure provides clear and concise deliverables that link strategic direction to organizational capabilities. Moreover, the IDEA approach reinforces an evergreen process that continually renews and fuels the tactical development and commercialization process, which flows through the post-launch operations environment back into the strategic portfolio renewal planning process.

What Key Overall Requirements Define this Approach? Prior to entering the IDENTIFY Phase; the business strategy must be clearly defined and documented. The business strategy breaks down into several general areas:

IDEA (Identify-Define-Evaluate-Activate)

75

• Financial Growth Goals • Core Competencies and Capabilities • Innovation Strategy (including both marketing and technology) These goals and capabilities serve as critical criteria to evaluate, prioritize, and select potential opportunities. As with all projects, the appropriate team members must participate in the project. The right mix of disciplines are needed to understand not only what the customer wants and what the business goals are, but also what is possible and what is feasible. What is possible and feasible refers not necessarily to technological hurdles, although that is an important ingredient for the concept under consideration and any related infrastructure support, but for the technology acceptability in the marketplace. Services and information concepts may require as careful a consideration around technological feasibility as a tangible product-offering concept when comprehending the activities along the full customer value chain. For example, customer support requirements or a system to manage intellectual capital or complex projects all require some level of technological consideration, often beyond the given offering. The appropriate amount of funding (resources) is required for the team to adequately complete their deliverables. Recall that though this Portfolio Renewal represents an inbound marketing function, it requires the proper funding for solid benchmarking and customer requirements gathering.

What Requirement Determines the Key Activities in this Approach? The requirements that define the method’s respective key activities (or phases) are found in Table 4-1.

SSFM

The appropriate mix of renewal and commercialization projects partly determines whether the right people are available to work on the critical projects and if the available funding exists. Companies (and/or divisions) often take on too many projects in the hopes that enough of them will deliver financial results that will meet the growth requirements of the business strategy. They overload the workforce—particularly understaffed marketing organizations. Applying Six Sigma thinking can minimize this dilemma.

76

Six Sigma for Marketing (SSFM) SSFM Portfolio Renewal Table 4-1: IDEA Requirements-Phase Linkage Requirements

I

D

E

A

Resulting High-Level Phase

What market holds what opportunity?

1. IDENTIFY

What are the customer requirements, and how do they translate into an offering?

2. DEFINE • Translate and document the opportunities as NUD requirements (e.g. New, Unique, and Difficult) • Document the requirements by segment. • Identify candidate portfolio architectures or platforms.

What is the recommended mix of portfolio offerings and funding requirements?

3. EVALUATE • Summarize the candidate offering concepts (product and services) and position them against the current offerings. • Document Growth Potential of the balanced offering portfolio architecture (including financial targets and tolerances). • Select the best portfolio architecture from among the candidates based on market potential and business case analysis. • Document product portfolio architecture. • Summarize the financial requirements and potential revenue and entitlement from the selected portfolio architecture as a comparison to the overall growth goals for the business. • Document Risk Assessment for the Portfolio that shows a balance across multiple dimensions of risk, with a Portfolio FMEA (Failure Modes and Effects Analysis) to identify possible failures, respective frequencies, and potential impact of those failures. • Document Real-Win-Worth (RWW) Analysis across the elements of the portfolio.

• Define the key markets of interest and describe their key characteristics. • Determine if they contain any segments and describe their key characteristics. • Identify the opportunities within each.

IDEA (Identify-Define-Evaluate-Activate) Requirements

77

Resulting High-Level Phase

Which offering concept 4. ACTIVATE projects will be activated • Rank the potential “activation” projects, as a development and requesting an offering development and commercialization commercialization team. project(s)? • Define the timing and headcount requirements for activation of projects based on a risk-balanced commercialization project control plan.

What Tools Are Aligned to Each Phase of the Process? Given the preceding high-level phases, the following series of tables summarize the subsequent tool-task-deliverables combination associated with each individual phase within the four-phase IDEA approach. The detail behind how to use each tool can be found in Part II of this book, “The Six Sigma Encyclopedia of Business Tools and Techniques: Choosing the Right Tool to Answer the Right Question at the Right Time.” Table 4-2 outlines the Identify Phase into its tool-task-deliverables linkages. Table 4-2: Identify Phase Tools-Tasks-Deliverables Phase 1: IDENTIFY—What market holds what opportunity

SSFM Portfolio Renewal

(including 1) General market defined, 2) Specific segments

I

D

E

A

Deliverables

Tasks

Documented Growth Goals and Core Competencies

Define Business Project Charter Strategy and Financial Goals.

Documented Innovation Strategy

Define Innovation Strategy.

Idea capture and database development tools

Documented Markets of interest and respective market segmentation, if any • Market and Segment

Define Markets and Market Segments.

• Market Identification and Segmentation Analysis • Porter’s 5 Forces Analysis and segmented risk profile • Market Behavioral Dynamics Map Methods • Value chain diagrams • Key event timelines

Behavioral Dynamics Map • Porter’s 5 Forces Chart

Candidate Tools and Techniques

continues

SSFM

identified, and 3) Specific opportunities determined)?

78

Six Sigma for Marketing (SSFM) Table 4-2: Continued Deliverables

Tasks

Candidate Tools and Techniques

Opportunities • Design and conduct Documented—Support VOC validation ing deliverables include: surveys. • Competitive Benchmarking • Document VOC Data and Trend Charts findings. • Market Perceived Quality • Conduct competitive Profile and Gap Matrix benchmarking (for • SWOT Matrix marketing, sales channel, and technical disciplines). • Define Opportunities across Markets and within segments. • Create a database of internal ideas based on opportunity categories.

• VOC Gathering Methods • Competitive Benchmarking and Best Practice Analysis • SWOT analysis matrix • Market Perceived Quality Profile Method • Idea (or Concept) capture database

Next steps planned

• Project Management tools • FMEA

Create DEFINE Phase project plan and risk analysis.

Table 4-3 organizes the Define Phase of IDEA into its tool-task-deliverables linkages. Table 4-3: Define Phase Tools-Tasks-Deliverables Phase 2: DEFINE—What are the customer requirements,

SSFM Portfolio Renewal

and how do they translate into an offering (including 1) Documented NUD portfolio requirements, 2) Documented

I

D

E

A

requirements by segment, and 3) Identified candidate portfolio architectures)? Deliverables

Tasks

Opportunities translated and documented as NUD (New, Unique, Difficult) requirements • Segmentation statistics

• Analyze VOC data to • Statistical survey design find common versus and analysis differentiated require- • Statistical data analysis ments for use in time and circumstance. • Conduct statistical analysis on VOC data to refine segment identification.

summary

• VOC-based requirements data

Candidate Tools and Techniques

IDEA (Identify-Define-Evaluate-Activate) Deliverables

Tasks

79

Candidate Tools and Techniques

• Common requirements • Translate NUDs into across segments Portfolio Require• Differentiated require- ments (including ments across segments metrics). • Customer Survey • Document NUD Results customer needs across segments. Requirements by segment Construct and/or documented refine market and segment Behavioral Dynamics Map.

Quality Function Deployment (QFD)

Candidate portfolio Define candidate port-. • Portfolio House of architectures or platforms folio architectures Quality (HOQ) identified • Product Portfolio Architecting methods Next steps planned

Create EVALUATE • Project Management Phase project plan and tools risk analysis. • FMEA

Table 4-4 presents the Evaluate Phase of IDEA into its tool-task-deliverables linkages. Table 4-4: Evaluate Phase Tools-Tasks-Deliverables Phase 3: EVALUATE—What is the recommended mix of

SSFM Portfolio Renewal

portfolio offerings and funding requirements (including

I

1) Summary candidate offering portfolio architectures,

D

E

A

2) Best candidates selected based on market potential and business case analysis, and 3) Summary financials)? Tasks

Candidate Tools and Techniques

Documented Growth Potential

• Create Preliminary • Business Case. • Develop Portfolio Financials [including NPV (Net Present Value), ECV (Expected • Commercial Value), ROI (Return on Investment)].

Financial Modeling and Forecasting tools such as NPV, ECV, ROI analysis and Monte Carlo simulation Business case development and valuation methods

continues

SSFM

Deliverables

80

Six Sigma for Marketing (SSFM) Table 4-4: Continued Deliverables

Tasks

Candidate Tools and Techniques

Documented Product Portfolio Architecture

• Develop a Portfolio Portfolio balancing Evaluation Criteria methods and benchmark portfolio architecture • Evaluate and Select the best portfolio architecture from the candidates. (It may be a hybrid of the candidates.)

• Documented Risk Assessment • Documented RealWin-Worth (RWW) Analysis

• Assess market dynamics and fit with trend lines. • Assess technical risk profiles. • Assess Portfolio Financials.

• Real-Win-Worth analysis. • Pugh Concept Evaluation and Selection Method

Next steps planned

• Evaluate concept risk.

• Create ACTIVATE Phase project plan. • Project Management tools • FMEA

Table 4-5 organizes the last phase of IDEA, the Activate Phase, into its tool-task-deliverables linkages. Table 4-5: Activate Phase Tools-Tasks-Deliverables Phase 4: ACTIVATE—Which offering concept projects will be activated as a development and commercialization project(s) (including the 1) Rank order of activation

SSFM Portfolio Renewal

I

D

E

A

project and 2) Timing for activation of projects, based upon a risk balanced commercialization project control plan)? Deliverables

Tasks

Documented availability, • Rank projects for readiness, and deployactivation priority ment of Core Competenand strategic value. cies and Resources • Determine required core competencies and resources for top projects.

Candidate Tools and Techniques

• Dynamic Rank Ordering methods • Pareto process • Resource planning • RACI

U-A-P-L (Understand-Analyze-Plan-Launch) Deliverables

Tasks

81

Candidate Tools and Techniques

• Document availability of resources. Documented Project Activation Plan and Control Plan

Create project activation Control Plan timing schedule and control plan.

Documented risk response plan

Conduct a risk analysis FMEA on the activation plan and the portfolio for financial performance against growth goals.

Documented enabling technologies maturity and readiness

Define stability of enabling technologies and document and balance resources across the project activation plan.

Next steps planned

• Create high-level Project Management Development and tools Commercialization project plan. • Document Lessons Learned. • Close Portfolio Renewal project.

Project Management tools for resource planning and budgeting

U-A-P-L (Understand-Analyze-Plan-Launch)

The project team may be comprised of multiple functional teams or one integrated team made up of the required technical and business resources to convert the strategic concept into a viable, market-ready offering. This team bridges the strategic initiative with the operational management of a launched offering; hence, this team performs the tactical activities.

SSFM

Brief Description of Typical Applications The successful completion of an offering concept passing the IDEA Activation Phase triggers the need to initiate a project team to design, develop, and commercialize the concept. The new project team takes the strategic planning processes deliverables as key inputs to start the new process.

82

Six Sigma for Marketing (SSFM)

The marketing and business project team members develop the inbound marketing deliverables in parallel with their technical counterparts, who design the offering. Using a Six Sigma approach, the technical discipline follows a DFSS method (DMADV (Define-Measure-AnalyzeDesign-Verify) or CDOV (Concept-Design-Optimize-Verify)) to develop and stabilize the offering design. In turn, the marketing and business professionals should follow a complimentary Six Sigma approach, namely UAPL [pronounced “YOU-apple”]. This tactical marketing process recognizes the interdisciplinary dependency for design and commercialization deliverables, regardless if the offering is a product, service, or combination. The UAPL method reinforces the bi-directional communication between the two core disciplines and aids in the translation of functional requirements and specifications. The integrated flow between CDOV and UAPL with their complimentary approach is depicted in Figure 4-4. C-D-O-V (Technical) U-A-P-L (Marketing and Business)

Concept

nderstand U

Design

Analyze

Optimize

Plan

Verify

Launch

Figure 4-4: Comparison of CDOV and UAPL Methods

The design of UAPL is to be flexible, such that it can be tailored to serve as additional foundational structure to any existing tactical offering development and commercialization method. Its tools-tasks-deliverables combination can be incorporated to make the current approach more robust. For example, if a Product Development Process (PDP) process is well established and successful, critical elements of UAPL can strengthen the design and development in two key ways. First, UAPL reinforces a proactive perspective to the process. Second, UAPL reinforces a bi-directional communication through a structured translation approach between the customer-supplier relationships for both external and internal process players across any functional language or other perspective barriers. The standard UAPL approach is segmented into four distinct phases, defined as follows: Phase 1. UNDERSTAND—the market opportunity and specific customer requirements translated into product (or service) requirements. Phase 2. ANALYZE—customer preferences against the value proposition.

U-A-P-L (Understand-Analyze-Plan-Launch)

83

Phase 3. PLAN—the linkage between the value chain process details (including marketing and sales) to successfully communicate and launch the product (or service) concept as defined in a maturing business case. Phase 4. LAUNCH—preparation of the new product (or service) under a rigorously defined launch control plan. Figure 4-5 provides a diagram of the high-level flow of the UAPL approach and its key requirements: Understand No

Customers’ needs and preferences understood?

Plan

Analyze No

Yes

Needs translated into viable NU D requirements?

Launch No

Yes

Complete and comprehensive tactical plans documented?

No

Key launch elements (inputs, outputs, metrics and key suppliers)

Yes

defined?

Yes Launch preparations completed

Close Project

Figure 4-5: High-level UAPL Model Process Flow

What Key Overall Requirements Define this Approach? Prior to entering the UNDERSTAND Phase of UAPL; the offering development/commercialization team must fully comprehend the concept strategy from the offerings portfolio and renewal (IDEA) process. The concept strategy inputs should include

• Offering’s value proposition to its target market segments (including environmental influences, such as marketplace, political, regulatory, competitive, and technological assumptions) • Financial growth goals and assumptions for the offering. • Core competencies and capabilities required for the offering • Innovation strategy and assumptions (impacting technology, marketing, and business across the customer value chain and any infrastructure)

SSFM

• Offering positioning within the portfolio mix

84

Six Sigma for Marketing (SSFM)

When the tactical project team has these inputs in hand, it can embark on the specific design, development, and commercialization efforts of the specific offering concept. The overall requirements that direct the activities within UAPL incorporate the efforts of marketing, general business, and the customer value chain players. These UAPL activities compliment any required technical activities needed to design, develop, and commercialize the offering concept. The UAPL approach answers the follow overarching questions: • What are the customers’ specific offering needs and preferences? • What is the best way to translate those customer needs into concept design and deployment requirements that take into consideration the NUD (New, Unique, and Difficult)? • What are the detailed tactical plans to design, develop, and prepare the offering for commercial success? • What are the key elements (inputs, outputs, metrics, and key suppliers) to introduce and manage the offering in the marketplace and to ensure a high-quality launch? These four requirements define the UAPL approach and define the phasegates and the respective key activities within each phase. The breakdown of the requirement-phase linkage is detailed in the next section.

What Requirement Determines the Key Activities in this Approach? The requirements that shape the UAPL method into its four phases can be found in Table 4-6. Table 4-6: UAPL Requirements-Phase Linkage

SSFM Commercialization

U Requirements

What are the customers’ specific offering needs and preferences?

A

P

L

Resulting High-Level Phase

1. UNDERSTAND • Document customer requirements. • Document offering requirements and superior offering concepts. • Refine and update Real-Win-Worth (RWW) Analysis. • Refine and update business case.

U-A-P-L (Understand-Analyze-Plan-Launch) Resulting High-Level Phase

What is the best way to translate those needs into concept design and deployment requirements that take into consideration the NUD (New, Unique, and Difficult)?

2. ANALYZE • Finalize offering requirements from customer testing. • Document value proposition linked to brand. • Refine price model and sales forecasts. • Update critical parameters for marketing, sales channel, customer value chain participants, and supporting infrastructure and business areas. • Refine and update Real-Win-Worth (RWW) Analysis. • Refine and update business case. • Draft a high-level marketing and business plan, including a preliminary risk mitigation plan (defined in concert with the technical development plan, as appropriate).

What are the detailed tactical plans to design, develop and prepare the offering for commercial success?

3. PLAN • Develop a robust marketing and business plan, including a refined risk mitigation plan (defined in concert with the technical development plan, as appropriate). • Develop and/or coordinate the development of a customer value chain plan (including the sales channel(s), customer administration and financing, customer services and training (preand post-purchase), customer support and maintenance, supplies and parts, and other supporting infrastructure and business). • Develop marketing collaterals, advertising, promotional, and other customer support materials. • Document channel management plan for marketing, sales channel, and other customer value chain participants and include both risk and Critical Parameter Management. • Update critical parameters appropriate for marketing, sales channel, customer value chain participants (including any third-party partners), and supporting business and infrastructure. • Refine and update Real-Win-Worth (RWW) Analysis. • Refine and update business case. • Develop project management plan for launch.

continues

SSFM

Requirements

85

86

Six Sigma for Marketing (SSFM) Table 4-6: Continued Requirements

Resulting High-Level Phase

What are the key 4. LAUNCH elements (inputs, • Document the launch plan, including critical outputs, metrics, and inputs, respective suppliers, key outputs, and key suppliers) to introperformance metrics. duce and manage the • Document post-launch customer value chain offering in the marketprocesses and control plans including marketplace and to ensure a ing, business, selling, support, and services. high-quality launch? • Document customer relationship management process and control plan. • Finalize marketing, sales, services, and support collaterals, advertising, and promotional materials. • Update critical parameters appropriate for marketing, sales channel, customer value chain participants (including any third-party partners), and supporting business and infrastructure. • Update Real-Win-Worth (RWW) Analysis, as appropriate. • Update business case and sales forecast, as appropriate.

Which Tools Are Aligned to Each Step of the Process? Given these high-level phases, the series of tables that follow summarize the subsequent tool-task-deliverables combination associated with each individual phase within the four-phase UAPL approach. The detail behind how to use each tool can be found in Part II of this book. Table 4-7 organizes the Understand Phase into its tool-task-deliverables linkages.

U-A-P-L (Understand-Analyze-Plan-Launch)

87

Table 4-7: Understand Phase Tools-Tasks-Deliverables Phase 1: UNDERSTAND—What are the customer’s specific

SSFM Commercialization

offering needs and preferences (including 1) Customer

U

A

P

L

requirements, 2) Offering requirements and superior offering concepts, 3) Real-Win-Worth (RWW) Analysis, and 4) The offering’s business case)? Deliverables

Tasks

Candidate Tools and Techniques

Business case goals documented

• Gather offering GOSPA Analysis financial goals, positioning, and assumptions from the strategic plan. • Conduct a Goals, Objectives, Strategies, Plans, and Actions (GOSPA) analysis.

Segmented markets identified

Verify markets and their segments and specific customer types and characteristics for the offering concept.

Customer requirements documented

• Create a customer • interview guide. • • Gather and translate specific VOC data. • • Document the • specific NUD offering requirements. • • Construct specific customer behavioral dynamics maps. • • Conduct competitive benchmarking against the NUD customer requirements.

VOC gathering methods Customer Interview Guide Customer Value Map KJ diagrams (images and requirements) Competitive benchmarking data and trend analysis. NUD: What is New, which no one fulfills today; What is Unique to what you offer today but what competitors fulfill; What is Difficult, that may give you a competitive advantage and compliment current competencies and business model? • Porter’s 5 Forces Analysis • Offering category SWOT Analysis

continues

SSFM

• Market Perceived Quality Profile • Customer Behavioral Dynamics Map

88

Six Sigma for Marketing (SSFM) Table 4-7: Continued Deliverables

Tasks

Offering requirements documented

Translate the VOC into • Value chain analysis the value proposition • CTQ tree or matrix and help translate them into technical requirements.

Superior offering concept documented

• Help generate and evaluate concepts that are candidates to fulfill NUD. • Generate and/or revise competitive position analysis.

Innovation strategy documented

• Define the innovation strategy (s) and portfolio fit for marketing, business, customer value chain and infrastructure. • Generate the qualitative and quantitative elements of value that the concept provides to substantiate the specific offering opportunity.

Core competencies documented

• Conduct customer Value chain mapping value chain analysis. • Determine availability and readiness at time of launch.

Critical parameters • Update market perdefined for marketing, ceived quality profiles. sales channel, customer • Develop offering value chain, and support- category specific key ing business and infraevent timeline. structure

Candidate Tools and Techniques

• Product Category Data Mining • Offering category SWOT • Concept generation tools • Pugh concept • QFD/HOQ methods

• Market Perceived Quality Profile • Value Chain analysis • Key events timeline • CTQ tree or matrix

U-A-P-L (Understand-Analyze-Plan-Launch) Deliverables

Tasks

Candidate Tools and Techniques

Real-Win-Worth Analysis updated

• Conduct Won-Lost analysis. • Revise the offering Real-Win-Worth Analysis.

• Real-Win-Worth Analysis • Won/Lost Analysis

Business case updated

Develop the offering • Business case modeling business case and document assumptions.

Next steps planned

Create ANALYZE Phase project plan and risk analysis.

89

• Project Management tools • FMEA

The Analyze Phase of UAPL can be characterized as a robust phase due to the many requirements directing this phase’s activities. Several crucial elements are needed to answer the critical question, “What is the best way to translate those needs into concept design and deployment requirements that take into consideration the NUD (New, Unique, and Difficult)?” For purposes of emphasis, these additional phase requirements are listed here as follows: 1. Finalized offering requirements from customer testing. 2. Documented value proposition linked to brand. 3. Refined price model with Monte Carlo simulations. 4. Refined sales forecasts with Monte Carlo simulations. 5. Updated critical parameters (and respective database(s)) for marketing, sales channels, customer value chain, and supporting infrastructure and businesses. 6. Refined and updated Real-Win-Worth (RWW) Analysis. 7. Refined and updated business case.

Table 4-8 presents the Analyze Phase of UAPL into its tool-task-deliverables linkages.

SSFM

8. PLAN Phase project plan, including a preliminary risk mitigation plan for marketing and supporting business areas.

90

Six Sigma for Marketing (SSFM) Table 4-8: Analyze Phase Tools-Tasks-Deliverables Phase 2: ANALYZE—What is the best way to translate those needs into concept design and deployment requirements that take into consideration the NUD

SSFM Commercialization

U

A

P

L

(including 1) Finalized offering requirements, 2) Documented value proposition, 3) Refined price model, 4) Refined sales forecast, 5) Updated critical parameters, 6) RWW Analysis, 7) Refined business case, and 8) Project plan for the PLAN Phase)? Deliverables

Tasks

Candidate Tools and Techniques

Critical Parameter Management defined

Assess strategic fit of evolving customer value chain support and services.

• Critical Parameter Management (CPM) • Multi-vari studies

Organizational capability Examine offering assessment completed and/or Product Development Process (PDP) capability to meet launch date.

Statistical Process Control (SPC)

Customer preferences defined

• Define customer • Customer-based concept preferences. testing • Conduct a Conjoint • DOE (Design of Analysis if offering’s Experiment) complexity supports • Descriptive and inferenpotential “bundling” tial statistical data of options. analysis. • ANOVA (Analysis of Variance) • Regression and empirical modeling methods • Conjoint Analysis • Demand Elasticity

Price model finalized

Simulate price model and select optimal configuration.

Marketing, selling, and customer value chain support strategy defined

Evaluate capabilities of • Sales channel analysis marketing, sales chan- • Marketing plan template nels, customer value chain, and supporting infrastructure and business areas and develop plan to close any gaps and meet customer needs.

Monte Carlo simulation

U-A-P-L (Understand-Analyze-Plan-Launch) Deliverables

Tasks

Candidate Tools and Techniques

Sales forecasts finalized

Simulate sales forecasts by channel, and select optimal configuration.

Monte Carlo simulation

Marketing plan drafted

• Develop a marketing plan. • Refine RWW Analysis

• RWW Analysis • Marketing Plan template

Business case refined

Update the business case and financials.

Business case analysis

Next steps planned

Create PLAN Phase project plan and risk analysis.

• Project Management tools • FMEA

91

Similar to the UAPL Analyze Phase, the Plan Phase has many requirements directing its activities. Several critical elements are needed to answer the critical question, “What are the detailed tactical plans to design, develop, and prepare the offering for commercial success?” For purposes of emphasis, these additional phase requirements are listed here as follows: 1. Finalized robust plans for both marketing and the sales channels. 2. Developed marketing collateral, advertising, promotional, training, and customer support materials. 3. Documented management plan for the post-launch operations, including the sales channels, customer value chain components, and supporting infrastructure and business areas. 4. Updated critical parameters (and respective database) for marketing, the sales channels, customer value chain components, and supporting infrastructure and business areas. 5. Refined and updated Real-Win-Worth (RWW) analysis.

7. LAUNCH Phase project plan, including a preliminary risk mitigation plan, for marketing and supporting business areas. Table 4-9 arranges the Plan Phase of UAPL into its tool-task-deliverables linkages.

SSFM

6. Refined and updated business case.

92

Six Sigma for Marketing (SSFM) Table 4-9: Plan Phase Tools-Tasks-Deliverables Phase 3: PLAN—What are the detailed tactical plans to

SSFM Commercialization

design, develop and prepare the offering for commercial

U

success (including 1) Marketing and the sales channels

A

P

L

plans, 2) Marketing collateral, advertising, promotional, training, and customer support materials, 3) Post-launch operations management plans, 4) Post-launch operations critical parameters, 5) Real-Win-Worth (RWW) analysis updated, 6) Business case updated, and 7) LAUNCH Phase project plan)? Deliverables

Tasks

Candidate Tools and Techniques

Updated Critical Parameter Management documentation and database

• Update, revise and • CPM implement the Critical • CTQs Parameter Manage- • RACI ment plan for the functional areas involved in post-launch operations. • Develop Who/What matrices to map responsibility for delivery of all CTQs.

Post-Launch Process Maps

• Develop detailed • Process Mapping process maps for • Kaizen on current marketing and sales processes processes. • Lean on current • Develop high-level processes (or appropriate level) process maps for customer value chain organizations and supporting infrastructure and business areas. • Document process maps depicting how to fulfill the CTQ requirements.

U-A-P-L (Understand-Analyze-Plan-Launch) Tasks

Candidate Tools and Techniques

Plans: Preliminary • Audit and update • Marketing, Sales Channel the marketing plan. Management, Market • Establish Market • Communications, CustAccountability Plan. omer Relationship Man- • Create and impleagement, and Brand ment internal com- • Positioning munications plan to the company • employees and across the value • chain including the sales force, customer • support, services, and suppliers. • Develop external communications strategy (comprehending multiple venues and format options such as collaterals, the Internet/ Web). • Develop sales support plan.

Plan Development Methods Marketing and Sales Process Noise Diagramming Customer Relationship Management methods Marketing Communications Planning Brand Positioning and Management RACI

Preliminary designs for marketing and sales collateral, advertising, and promotional materials

Develop preliminary • Marketing Collateral designs for marketing and Promotional and sales collateral, Material Planning and advertising, and Development promotional materials.

Refined and updated RWW Analysis

Test effectiveness of offering positioning.

• Real-Win-Worth (RWW) Analysis

Refined and updated business case

• Develop Pricing Plan by segment. • Update Business Case and Sales Forecast.

• Business case analysis • The Hybrid Grid • Monte Carlo simulation

Next steps planned

• Create LAUNCH • Project Management Phase project plan tools and risk analysis. • FMEA • Document critical risks, problems, and assumptions.

SSFM

Deliverables

93

94

Six Sigma for Marketing (SSFM)

The UAPL Launch Phase contains a robust set of requirements (the key phase-gate question plus seven supporting requirements), like its two preceding phases. Its list of deliverables reaches 11 items. The tasks to produce those 11 outputs can be consolidated into only three core categories: 1) Conducting an audit of the marketing plan, 2) Assessing the plan completeness and readiness of both internal and external communications, and 3) Assessing the plan completeness and readiness of postlaunch operations. The brevity of the task list can deceive project teams into a false sense of control and excess available time. However, this concise three-task list represents a great deal of required work effort to produce the 11 Launch Phase-Gate deliverables. Table 4-10 organizes the last UAPL phase, Launch, into its tool-taskdeliverables linkages. Table 4-10: Launch Phase Tools-Tasks-Deliverables Phase 4: LAUNCH—What are the key elements (inputs, outputs, metrics and key suppliers) to introduce and

SSFM Commercialization

U

A

P

L

manage the offering in the marketplace and to ensure a high-quality launch (including 1) Launch plan, 2) Post-launch process maps and control plans with Discontinuance criteria, 3) CRM process plan, 4) Final marketing, sales, customer support collaterals, advertising and promotional materials, 5) Post-launch CPM, 6) Refined RWW Analysis, and 7) Final Business case and sales forecast)? Deliverables

Tasks

Candidate Tools and Techniques

Integrated Value Chain Plan

• Develop an integrated customer value chain plan that encompasses the marketing coordination, services, support, supply chain, customer financing and administration, and supporting infrastructure and business areas. • Assess Implement ation Plan readiness of the External Communications Strategy Implementation.

• Value chain development methods • Process mapping • RACI • Kaisen of currently operating processes • Lean and DMAIC of currently operating processes

U-A-P-L (Understand-Analyze-Plan-Launch) Deliverables

Tasks

Risk management plan

• Assessment of post- • Process Noise Diagrams launch operations • FMEA risk (sales, support, • Market plan audit and so on) • Post-launch operations Process Noise Diagrams and Failure Modes and Effects Analysis. • Conduct final marketing plan audit.

Launch Plan

• Finalize Launch Plan. • Launch planning • Assess Implementamethods tion Plan readiness • Kaizen of the Communications Plan for both external and internal (employees and value chain players such as the sales force, customer support and service, partners, suppliers).

Control plans and process maps of postlaunch processes

• Finalize the post• launch operations process maps and • control plans for marketing, sales, entire • value chain, and sup- • porting infrastructure and business areas (including marketing, advertising, promotion, public relations, selling, CRM, customer training and support, supporting alliances and partnerships, supplies management, customer financing and administration, and so on). • Establish discontinuance criteria to determine appropriate conditions wherein the offering needs to be discontinued.

95

Candidate Tools and Techniques

SSFM

Control plan development methods Process Mapping and analysis RACI Customer relationship management methods

continues

96

Six Sigma for Marketing (SSFM) Table 4-10: Continued Deliverables

Tasks

Candidate Tools and Techniques

Statistical Process Control Charts on critical data

• Finalize Statistical • SPC Process Control • Statistical data mining Charts for marketing, and analysis tools sales, and post• Market plan audit launch operations.

Critical Parameter Management databases

• Finalize the post• CPM launch operations CPM database with requirements, metrics, and controls.

Final launch materials

• Finalize marketing • Marketing communicaand sales collaterals, tions planning advertising, promotional, training, and customer support materials.

Final documentation for brand alignment with value proposition

• Conduct final marketing plan audit.

• Brand positioning and management • Marketing plan audit

Updated and refined Real-Win-Worth (RWW) Analysis

• Conduct final marketing plan audit.

• RWW Analysis • Market plan audit

Updated and refined business case and sales forecast

• Conduct final marketing plan audit.

• Business case analysis • Market plan audit

Next steps planned

• Prepare to implement • Project Management Launch Plan. tools • Document Lessons • FMEA Learned. • Close Development and Commercialization project.

L-M-A-D (Launch-Manage-Adapt-Discontinue)

97

L-M-A-D (Launch-Manage-Adapt-Discontinue) Brief Description of Typical Applications In the Operational marketing process environment, the LMAD method [pronounced “Elle-mad”] is used for managing the portfolio of launched products and/or services across the customer value chain. This approach solely applies to ongoing field operations, with the middle two phases specifically describing the cyclical lifecycle management process until the offering reaches end-of-life, triggering a withdrawal of commercial support. This approach has the following four distinct phases: Phase 1. LAUNCH—the offering through its introductory period into the market according to the Launch Control Plan of the prior process. Phase 2. MANAGE—the offering in the steady-state marketing and sales processes, while integrating the partner go-tomarket value chain functions. Phase 3. ADAPT—the marketing and sales tasks and tools, integrating the cross-functional go-to-market partners when marketplace events and noises require a course correction, modification, or any other kind of change. Phase 4. DISCONTINUE—the offering with discipline to sustain brand loyalty. Recall that as with the other SSFM methods, the design of LMAD supports any existing operational method such that its tools-tasks-deliverables can be incorporated to help make the current approach more robust. For example, within sales, standard processes for account management and selling may be well established and successful. If this is the case, LMAD should be tailored to fit as an underpinning to these current account management and selling methods.

The Phase-Gate Review approach applied to the strategic IDEA and tactical UAPL processes transform with LMAD because of the post-launch environment. The post-launch operational processes behave differently than portfolio renewal or commercialization processes. Like steady-state manufacturing, operational environments lack formal phase gates. Hence, operations require timely, periodic reviews of progress against a plan. (Recall that the post-launch operations plans developed during the tactical UAPL commercialization project cover marketing, sales channels,

SSFM

The four-phased LMAD approach ebbs and flows in and out of the MANAGE and ADAPT phases to stay “on plan” throughout the life of an offering. Figure 4-6 shows the need for a course correction, across a generic LMAD lifecycle.

98

Six Sigma for Marketing (SSFM)

customer service and Adapt Manage Launch Manage support, customer financing and administration, supForecast Delta (or difference) plies, and between Actual and Forecast supporting infrastructure and business Actuals areas.) These periodic operations reviews should be Regression Analysis of Actual sales thought of as data mapped against the Forecast key milestones in the Figure 4-6: Generic Revenue Versus Forecast Sales Data continuum of executing the appropriate plan—not as gate reviews. Data and results can be gathered and summarized on an hourly, daily, weekly, monthly, quarterly, and yearly basis. A balance between proactive and reactive performance measures should be maintained—with an emphasis on leading indicators when possible to stay on plan and under control. Most of the tools, methods, and best practices of outbound marketing are the same as inbound marketing. LMAD focuses on the tools and techniques that add value to staying on a given operational plan. The marketing, sales, and customer value chain professionals in an operational environment use these tools to refine forecasts, adjust estimates and to analyze data streams. They respond to change—change in the environment, customer needs, competition, technology, regulation, and so on. They adapt particularly when the change jeopardizes meeting their goals. Given that in a post-launch steady-state environment, a business will fluidly cycle in and out of the MANAGE and ADAPT phases as necessary. The LMAD approach also applies to the production engineering and manufacturing arena with relatively minor adaptations; however, this book focuses on the customer-facing value chain processes and activities. Figure 4-7 depicts the high-level flow of the LMAD approach and its key requirements for the post-launch operations of the customer value chain environment.

L-M-A-D (Launch-Manage-Adapt-Discontinue) Manage

Launch No

Adapt

Do conditions fit Discontinuance criteria?

On plan?

No

No

Yes

No No

Discontinue Yes

Yes

Launch activities completed per the Plan?

Do conditions fit Discontinuance criteria? No

Yes No

Yes

Launch metrics fulfill Plan requirements?

99

Assignable causes of variation from plan detected?

Critical adjustment parameters identified to return to statistical control? Yes Make appropriate adjustments and communicate to value chain players

Communicate data, discontinuance plan and timeline to value chain partners, marketplace and portfolio renewal team

Commence discontinuance process

Offering Discontinued

Figure 4-7: High-level LMAD Model Process Flow

What Key Overall Requirements Define this Approach? Prior to entering the LAUNCH Phase of LMAD, the tactical offering development/commercialization project team must provide the following inputs: • Integrated value chain plan across each of the various post-launch processes areas, including the following components: • Critical Parameter Management (CPM) requirements, metrics,

and controls • Readiness assessment • Control plan and Discontinuance criteria • Risk management plans

• Launch plan addressing market readiness activities, communication, and deployment to the external marketplace (customers, industry analysts, media/public relations, investors), employees, and each of the various post-launch processes areas. • Launch control and change management plan, which has built-in robustness to handle nonrandom sources of variation and sensitivities to change issues.

SSFM

• Process maps

100

Six Sigma for Marketing (SSFM)

• Metrics to monitor both launch process and performance results. • Risk management, response, and mitigation plan. • Final documentation on the offering’s value proposition and brand alignment. • Final support materials, such as marketing and sales collaterals, advertising, promotional materials, training, and customer support. • Updated Real-Win-Worth (RWW) Analysis. • Updated business case and sales forecast. Once the launch project team has received and understood these inputs, the offering launch team can begin its activities. The overall requirements that direct the activities within LMAD incorporate the efforts of marketing, supporting infrastructure and business and the customer value chain players. These LMAD activities compliment any current operational activities already in place to manage and adapt other offerings to be successful in the marketplace. The LMAD approach answers the follow overarching questions: • Has the Launch activities completed per the Launch Plan and fully satisfy the Launch Plan metrics? Including: • Communication and training activities completed to internal and external parties. • Launch deliverables fully distributed and available. • The post-launch customer value chain process players understand their respective plans and how they affect one another (marketing, sales, services, customer financing and administration, support and maintenance, and supporting infrastructure and businesses). • Can the post-launch marketing, sales, customer value chain, and supporting infrastructure and businesses detect assignable causes of variation that indicate that adaptive actions are required to stay on plan? • What are the critical adjustment parameters needed to return to a state of statistical control? In addition, what are the leading indicators (assignable causes of variation) that signal the post-launch customer value chain players to adjust their critical adjustment parameters? • Do the current conditions fit the discontinuance criteria?

L-M-A-D (Launch-Manage-Adapt-Discontinue)

101

These four high-level requirements define the LMAD approach and define the key milestones and the respective key activities within each phase. The breakdown of the requirement-milestone linkage is detailed in the next section.

What Requirement Determines the Key Activities in this Approach? The requirements that shape the LMAD method into its four milestones can be found in Table 4-11. Table 4-11: LMAD Requirements-Milestone Linkage SSFM

L

A

D

Resulting High-Level Milestones

Have the launch activities completed per the Launch Plan and fully satisfied the Launch Plan metrics?

1. LAUNCH

Can the post-launch operations players detect assignable causes of variation that indicate that adaptive actions are required to stay on plan?

2. MANAGE • Manage the ongoing, steady-state commercialization process throughout the lifecycle of the offering. • Depending on the offering’s lifecycle, both the Manage and Adapt phases can repeat several times and last years, or the timeframe could be very brief, before the final Discontinuance phase is reached.

• Deploy and stabilize offering’s launch activities, per the Launch Plan, across the postlaunch customer value chain processes. • Communicate and train internal and external parties. • Distribute launch deliverables. • Ensure that the post-launch customer value chain process players understand their respective plans and how they affect one another.

continues

SSFM

Requirements

M

Post-Launch

102

Six Sigma for Marketing (SSFM) Table 4-11: Continued Requirements

Resulting High-Level Milestones

What are the critical adjustment parameters needed to return to a state of statistical control? In addition, what are the leading indicators (assignable causes of variation) that signal the postlaunch customer value chain players to adjust their critical adjustment parameters?

3. ADAPT • Determine when to react or be proactive to events to stay on plan. (Can be in response to changes in the marketplace, economy, technology, regulatory, or simply the performance of the value chain processes.)

Do the current conditions fit the discontinuance criteria?

4. DISCONTINUE • Discontinue an offering based on its planned or forced end-of-life criteria. • Develop a preplanned set of deliverables and tasks to increase the likelihood of an efficient, more cost-effective transition to a replacement offering that renews the portfolio.

Which Tools Are Aligned to Each Step of the Process? Given these high-level milestones, the following series of tables summarize the subsequent tool-task-deliverables combination associated with each individual phase within the four-phase LMAD approach. Because of the operational environment within which LMAD fits, a common set of tools repeatedly gets utilized across the suite of LMAD phases. Hence, a single LMAD scorecard for performance evaluation featuring a common tool set the can be adapted on an “as needed” basis as part of the annual and/or operations review process. The tool set aligns within three categories: 1) Process Definition, 2) Risk Management, and 3) Operational Models, Data Analysis and Controls. In total, the tool set enables the required post-launch operational tasks throughout the lifecycle of a commercialized offering for marketing, sales, and customer value chain teams and supporting businesses. Process definition: There are five major types of tools, methods, and best practices used to define the Outbound Marketing and Sales Process. 1. Process Requirements Development—Including Interviewing and Requirements Data Gathering; Requirements Structuring, Ranking and Prioritization (for example, KJ Analysis); and Quality Function Deployment (to define detailed metrics).

L-M-A-D (Launch-Manage-Adapt-Discontinue)

103

2. Process Mapping—This captures the marketing and post-launch functions Inputs, Outputs, and Constrains in a process map, along with Process Noise Mapping and Failure Modes and Effects Analysis (FMEA). 3. Concept Generation—For marketing and post-launch operations professionals (including any third-party partners). 4. Pugh Concept Evaluation and Selection Process and the Concept Innovation Process. 5. Critical Parameter Management—This entails Critical Parameter Identification and Metrics for Product Promotion; Advertising; Channel and Distribution; Customer Relationship and Support; and Marketing Communications. Process Risk Management: This includes three types of major types of tools, methods, and best practices used to define the outbound marketing, sales, customer value chain, and supporting infrastructure and business processes. 1. Offering Management Scorecard to drive risk analysis, risk management, and decision-making 2. Failure Modes and Effects Analysis (FMEA) to delve further into Risk Analysis, Risk Management, and decision-making 3. SWOT Analysis to evaluate Strengths, Weaknesses, Opportunities, and Threats Operational Models, Data Analysis, and Controls: These have several appropriate Six Sigma for Growth tools, methods, and best practices available, including the following: 1. Line Management Control Planning for the offering 2. Project Management Methods, including two key Six Sigma tools: 1) Cycle-time Monte Carlo Simulation and 2) Critical Path Task FMEA 3. Cost Modeling for outbound marketing, sales, customer value chain, and supporting infrastructure and business processes

5. Sales Forecasting models using the Monte Carlo Simulation 6. Market Perceived Quality Profiles (MPQP) 7. Customer Value Chain Mapping

SSFM

4. Price Modeling for the offering, services, support, and supplies using the Monte Carlo Simulation

104

Six Sigma for Marketing (SSFM)

8. Surveys and Questionnaires design 9. Post-launch data structures, analysis, and management utilizing four types of tools: a. Descriptive and Inferential Statistics—such as Graphical Data Mining, Multi-vari Studies, Hypothesis Testing, Confidence Intervals, t-Tests, Data Sample Sizing, and last Regression and Model Building b. Capability Studies c. Statistical Process Control d. Design of Experiments (DOE)[md]on the critical parameters and Conjoint Analysis for the outbound marketing, sales, customer value chain, and supporting infrastructure and business processes 10. Data feedback structures for strategic portfolio renewal process and “advanced” or variant offering (product/services) planning. 11. Offering discontinuance planning. Recall that the candidate LMAD tool set is common across the four phases. The detail behind how to use each tool can be found in Part II of this book. Table 4-12 organizes the Launch Milestone into its tool-task-deliverables linkages. Table 4-12: Launch Milestone Tools-Tasks-Deliverables Milestone 1: LAUNCH—Have the launch activities

SSFM

been completed per the Launch Plan and fully satisfied the

L

M

Post-Launch

A

D

Launch Plan metrics? Deliverables

Tasks

• Capability Studies • Gather critical data. • Customer Identification • Analyze the perforand Qualification mance data: the actual Metrics versus forecast, the • Customer Purchase initial trend assessExperience and Bement, and growth havioral Analysis rate versus the plan. • Customer Satisfaction • Generate statistical Assessment process control charts and capability indices for key metrics.

Candidate Tools and Techniques

• Process Requirements Development • Process Mapping • RACI • Concept Generation • Pugh Concept Evaluation and Selection Process and the Concept Innovation Process • CPM (Critical Parameter Management)

L-M-A-D (Launch-Manage-Adapt-Discontinue) Tasks

Candidate Tools and Techniques

• Refine applicable forecast models.

• Offering Management Scorecard • Line Management Control Planning • Market Perceived Quality Profiles (MPQP) • Customer Value Chain Mapping • Surveys and Questionnaires • Post-launch data • Data feedback structures • Kaizen

Risk assessment and management

Conduct a Failure Modes and Effects (FMEA) Analysis, take appropriate action, and document Lessons Learned.

• FMEA • Offering Management Scorecard

Competitive Assessment

Conduct a competitive • Offering Management assessment. Scorecard • SWOT

Advertising and Promo tion Effectiveness Evaluation

Evaluate any adver. tising and promotion effectiveness

Critical parameters updated

Update CPM and SPC • CPM charts. • SPC • Offering Management Scorecard • Offering discontinuance planning

• Offering Management Scorecard • SWOT

Business Case Fulfillment Assess performance Assessment against the offering business case.

• Offering Management Scorecard • Cost modeling • Price modeling • Forecast modeling

Next steps planned

• Project Management tools • FMEA • Offering discontinuance planning

Create MANAGE Phase project plan and risk analysis.

SSFM

Deliverables

105

106

Six Sigma for Marketing (SSFM)

To govern over the ongoing operational marketing, sales, and customer value chain processes, the same set of tasks and resulting deliverables found in the Launch Milestone also are used in the Manage portion of LMAD. The difference is that they continue to get updated and modified as the environment and information change over time. Table 4-13 portrays the LMAD Manage Milestone into its tool-task-deliverables linkages. Table 4-13: Manage Milestone Tools-Tasks-Deliverables Milestone 2: MANAGE—Can the post-launch operations players detect assignable causes of variation that indicate that adaptive actions are required to stay

SSFM

L

M

Post-Launch

A

D

on plan? Deliverables

Tasks

Candidate Tools and Techniques

• Capability Studies • Gather critical data. • Customer Identification • Analyze the performand Qualification ance data: the actual Metrics versus forecast, the • Customer Purchase initial trend assessExperience and ment, and growth Behavioral Analysis rate versus the plan. • Customer Satisfaction • Generate statistical Assessment process control charts and capability indices for key metrics. • Refine applicable forecast models.

• Process Requirements Development • Process Mapping • RACI • Concept Generation • Pugh Concept Evaluation and Selection Process and the Concept Innovation Process • CPM (Critical Parameter Management) • Offering Management Scorecard • Line Management Control Planning • Market Perceived Quality Profiles (MPQP) • Customer Value Chain Mapping • Surveys and Questionnaires • Post-Launch data • Data feedback structures • Kaizen

Risk assessment and management

• FMEA • Offering Management Scorecard

Conduct a Failure Modes and Effects (FMEA) Analysis, take appropriate action, and document Lessons Learned.

L-M-A-D (Launch-Manage-Adapt-Discontinue) Deliverables

Tasks

Competitive Assessment

Conduct a competitive • Offering Management assessment. Scorecard • SWOT

Advertising and Promotion Effectiveness Evaluation

Evaluate any advertising and promotion effectiveness

• Offering Management Scorecard • SWOT

Critical parameters updated

Update CPM and SPC charts.

• CPM • SPC • Offering Management Scorecard • Offering discontinuance planning

Candidate Tools and Techniques

Business Case Fulfillment Assess performance Assessment against the offering business case.

• Offering Management Scorecard • Cost modeling • Price modeling • Forecast modeling

Next steps planned

• Project Management tools • FMEA • Offering discontinuance planning

Create ADAPT Phase project plan and risk analysis.

107

Table 4-14 arranges the Adapt Milestone of UAPL into its tool-task-deliverables linkages. Table 4-14: Adapt Milestone Tools-Tasks-Deliverables Milestone 3: ADAPT—What are the critical adjustment parameters needed to return to a state of statistical control? In addition, what are the leading indicators

SSFM

L

M

Post-Launch

A

D

(assignable causes of variation) that signal the post-launch customer value chain play-

Deliverables

Tasks

Candidate Tools and Techniques

• Updated critical parameters • Updated Process Noise Maps

• Apply critical adjust- • Process Requirements ment parameters to Development get results back on • Process Mapping target. • RACI

continues

SSFM

ers to adjust their critical adjustment parameters?

108

Six Sigma for Marketing (SSFM) Table 4-14: Continued Deliverables

Tasks

Candidate Tools and Techniques

• Conduct Designed Experiments as necessary to improve effectiveness of Critical Adjustment Parameters for current conditions. • Generate Statistical Process Control Charts and Capability Indices after adjustments. • Refine advertising, promotion, customer support services, and channel management plans. • Refine ADAPT Phase Control Plan. • Develop Noise Maps.

• Concept Generation • Pugh Concept Evaluation and Selection Process and the Concept Innovation Process • CPM (Critical Parameter Management) • Offering Management Scorecard • Line Management Control Planning • Market Perceived Quality Profiles (MPQP) • Customer Value Chain Mapping • Surveys and Questionnaires • Post-launch data • Data feedback structures • Offering Management Scorecard • SPC • Offering discontinuance planning • Kaizen • Lean and DMAIC

Updated Process FMEA

Analyze FMEAs, respond accordingly, and document actions.

• FMEA • Competitive situation analysis • SWOT • Market Perceived Quality Profile and Gap Analysis • Offering Management Scorecard

Update Annual Operating Plan

• Refine price models. • Refine sales forecast models. • Assess Business Case against current performance.

• Offering Management Scorecard • Cost modeling • Price modeling • Forecast modeling

L-M-A-D (Launch-Manage-Adapt-Discontinue) Deliverables

Tasks

Documented DISCONTINUE Phase Plan approved and ready for use

Refine DISCONTINUE • Offering Management Phase Control Plan. Scorecard • RACI

Next steps planned

• Create LAUNCH Phase project plan and risk analysis. • Document Critical Risks, Problems, and Assumptions.

109

Candidate Tools and Techniques

• Project Management tools • FMEA • Offering Management Scorecard

Table 4-15 organizes the last LMAD Milestone, Discontinue, into its tooltask-deliverables linkages. Table 4-15: Discontinue Milestone Tools-Tasks-Deliverables Milestone 4: DISCONTINUE—Do the current conditions

SSFM

fit the discontinuance criteria?

L

M

Post-Launch

A

D

Deliverables

Tasks

Candidate Tools and Techniques

Refined discontinuance criteria

• Apply CPM to adjust marketing, sales, and customer value chain functions and results to control the discontinuance, protect the brand, and maximize the business case. • Document discontinuance criteria based on CPM data to assess the business case versus current performance. • May conduct Designed Experiments as necessary to improve the effectiveness of critical adjustment parameters for discontinuance conditions.

• • • •

SSFM

CPM SPC charts Capability studies Trend analysis

continues

110

Six Sigma for Marketing (SSFM) Table 4-15: Discontinue Milestone Tools-Tasks-Deliverables Deliverables

Tasks

Candidate Tools and Techniques

Discontinuance Phase Project Plan

• Develop Discontin- • uance Plan and risk • assessment. • • Update FMEA. • • Update Noise Maps. • • Update competitive assessment. • • Refine price models • for discontinuance. • • Refine advertising, • promotion, and channel management plans for discontinuance. • Generate discontinuance forecast models for sales, supplies, support. • Document final business case assessment against goals.

Documented discontinuance data sent to portfolio renewal team

• Gather and summarize • discontinuance data. • • Send performance report to portfolio renewal team, including recommendations • for sustaining the brand through nextgeneration offerings. • •

Project Plan template FMEA Noise Maps Competitive assessment Project Management tools RACI Pricing models Forecast models Business case template

VOC Gathering survey data from marketing, sales, and customer value chain players Market Perceived Quality Profile (MPQP) and Gaps SWOT Porter’s 5 Forces Analysis • FMEA • Lessons Learned

Do DMAIC and Lean Six Sigma Apply to Marketing? Yes; if a marketing process is broken, ineffective, or out of control, then use one of the traditional Six Sigma approaches to improve or re-design it. SSFM presumes that the strategic, tactical, and operational marketing processes have been designed to function properly. The three SSFM methods define what to do and when to do it within structured marketing processes.

Summar y

111

Where to Get More Information on SSFM For more in-depth understanding of the three Six Sigma for marketing and business processes (IDEA, UAPL, and LMAD), please refer to Six Sigma for Marketing Processes, An Overview for Marketing Executives, Leaders, and Managers, by Clyde M. Creveling, Lynne Hambleton, and Burke McCarthy, published by Prentice-Hall, New Jersey, 2006; ISBN 0-13199008-X.

Summary

SSFM

Six Sigma for Marketing (SSFM), though a bit of a misnomer, refers to a category of proactive growth methods and tools for the marketing, sales, and other business professionals involved in strategic planning, offerings development, and ongoing operations management of the launched offering portfolio. The SSFM standard toolset contains invaluable candidate tools applicable to several business processes throughout an organization and can be found in Part II of this book, the “How To” utilize tools section.

This page intentionally left blank

Pa r t I I Six Sigma Tools and Techniques: Choosing the Right Tool to Answer the Right Question at the Right Time Encyclopedia The Six Sigma Encyclopedia of Business Tools and Techniques Summary Tool Matrix A Activity Network Diagram (AND)— 7M Tool

Checklists—7QC Tool Communication Plan Conjoint Analysis Control Charts—7QC Tool Control Plan

Affinity Diagram—7M Tool

Cost/Benefit Analysis

Analysis of Variance (ANOVA)

Critical Path Method (CPM)

Arrow Diagram

Critical-to-Quality (CTQ)

B

Benchmarking

D

Data Collection Matrix

Box Plots—Graphical Tool

Design of Experiment (DOE)

Brainstorming Technique

Dotplot

C

Capability Analysis

Cause and Effect Diagram—7QC Tool Cause and Effect Prioritization Matrix

F Failure Modes and Effects Analysis (FMEA) 5-Whys Fault Tree Analysis (FTA) Fishbone Diagram—7QC Tool

Cause and Prevention Diagram 113

114

Flowchart—7QC Tool G

Gantt Chart

GOSPA (Goals, Objectives, Strategies, Plans and Actions) Graphical Methods H

Histogram—7QC Tool

R RACI Matrix (Responsible, Accountable, Consulted, Informed) Real-Win-Worth (RWW) Analysis Regression Analysis Risk Mitigation Plan Rolled Throughput Yield

House of Quality (HOQ)

Run Chart - 7QC Tool

Hypothesis Testing

S Tool

I Interrelationship Diagram—7M Tool K

KJ Analysis

M Market Perceived Quality Profile (MPQP) Matrix Diagrams -7M Tool Measurement System Analysis (MSA)

7M—Seven Management

7QC—Seven Quality Control Tool Sampling Scatter Diagram—7QC Tool Scorecards SIPOC (Supplier-Input-ProcessOutput-Customer)

Monte Carlo Simulation

SMART Problem & Goal Statements for a Project Charter

Multi-vari Chart

Solution Selection Matrix

N

Normal Probability Plot

Stakeholder Analysis

P

Pareto Chart—7QC Tool

Statistical Tools

PERT (Program Evaluation and Review Technique) Chart

Stratification—7QC Tool

Poka-Yoke

SWOT (Strengths-WeaknessesOpportunities-Threats)

Porter’s 5 Forces

T

Prioritization Matrices— 7M Tool

TRIZ

Process Capability Analysis

V

Process Decision Program Charts (PDPC) - 7M Tool

Voice of Customer Gathering Techniques

Process Map (or Flowchart)—7QC Tool

W Work Breakdown Structure (WBS)

Pugh Concept Evaluation

Y

Q Quality Function Deployment (QFD)

Tree Diagram— 7M Tool Value Stream Analysis

Y = f (X)

Encyclopedia The Six Sigma Encyclopedia of Business Tools and Techniques Summary Tool Matrix For easy navigation, use the Jump Table that follows, which lists the different tools and techniques featured in this book and organizes them by purpose and the key questions they answer. The statistical and graphical tools include additional descriptors about their primary applications—describe, compare, and predict. In addition, the project management tools also include appropriate descriptors— scheduling, planning, scope, human resources (HR), deliverables, and tasks. The Encyclopedia Summary Tool Matrix Jump Table Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

Benchmarking

p. 160

Competitive Marketplace and Positioning

What are the experts doing, and how do they do it?

How does the market Market Perceived perceive the quality of Quality Profile the product and/or (MPQP) services offerings versus competition?

p. 390

continues 115

116

Encyclopedia Continued Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

How are we performing relative to the competitive business threats?

Porter’s 5 Forces

p. 464

How would a potential Real-Win-Worth offering be valued in (RWW) Analysis the marketplace, and positioned well against competition? Is it worth the investment to develop the idea? Would it be successful? Would the concept outpace competition?

p. 560

How do the organization’s strengths and weaknesses compare with the competitive opportunities and threats?

SWOT (StrengthsWeaknessesOpportunitiesThreats)

p. 699

What is the payback time period for an investment?

Cost/Benefit Analysis

p. 238

What are the probabil- Monte Carlo ities and risk associated Simulation with several financial possibilities? (that is, pricing, forecasting)

p. 435

What does the distriBoxplot— bution of a set of data Graphical Tool look like? Alternatively, what do the distributions look like and compare for multiple sample groups?

p. 164

Financial

Predict

Graphing

Describe

Summar y Tool Matrix Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Describe

How is the data of a process distributed; what does the data set look like?

Dotplot

p. 280

Describe

How is the data of a process distributed; what does the data set look like? Are the data normally distributed?

Histogram— 7QC Tool

p. 330

Describe & Compare

Across multiple sources of variability, which one contributes the most?

Multi-vari Chart

p. 439

Compare & Predict

Are the data distributed normally?

Normal Probability p. 444 Plot (See Also “Control Charts”—Normal Versus Non-normal Data section p. 227)

Describe & Compare

What are the vital few items with the biggest impact? Which 20% of items produce 80% of the impact (the 80/20 rule)?

Pareto Chart— 7QC Tool

p. 445

Describe

How does the data look over time? Are the data randomly distributed over time? Does the process look stable and random over time?

Run Chart— 7QC Tool

p. 610

Describe & Compare

Are these two factors Scatter Diagram— correlated? What is the 7QC Tool relationship between these two factors?

p. 640

Describe & Compare

Are there any patterns in the data? Is the data set homogeneous?

p. 697

Stratification— 7QC Tool

Page

continues

117

118

Encyclopedia Continued Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

What is the progress of the project and project team or the process and process players?

Scorecards

p. 653

How can you elicit new ideas in a short period of time?

Brainstorming Technique

p. 168

Which technical specifications for a product or services offering best meet a specific set of customer requirements?

House of Quality (HOQ) (See Also Quality Functional Deployment (QFD),” p. 543

p. 335

Which design or potential solution option is best?

Pugh Concept Evaluation

p. 534

Governance

Idea/Solution Generation and Selection

Which technical spec- Quality Function ifications for a product Deployment (QFD) or services offering best meet a specific set of customer requirements?

p. 543

Which solution option best meets requirements?

p. 672

Solution Selection Matrix

What is the best TRIZ solution to address this problem to create a competitive advantage?

p. 715

Summar y Tool Matrix Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

Organization & Planning

What activities or Checklists— deliverables are required 7QC Tool to meet requirements? Alternatively, what was observed?

p. 204

What information is available about the current process, product, or services offering?

Data Collection Matrix

p. 248

How do the activities of a function, program, or project align with the overall organizational strategy?

GOSPA (Goals, Objectives, Strategies, Plans, and Actions)

p. 320

How do these two Matrix Diagrams— (or more) groups relate 7M Tool to one another?

p. 399

What are the details Tree Diagram–— behind this general 7M Tool topic breakdown into smaller components? How does it break down into its piece-parts?

p. 712

Which key variables Cause-and-Effect (process steps or process Prioritization inputs) best meet cus- Matrix tomer requirements (or the key process output variables)?

p. 188

What is the best option Prioritization among the several pos- Matrices—7M Tool sibilities for a crucial (often mission-critical) decision that carries risk of significant consequences if wrong?

p. 470

Prioritization

continues

119

120

Encyclopedia Continued Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

Process Evaluation (Also see Variation Analysys)

Describe & Compare

Is the process able to to meet customer requirements?

Capability p. 173 Analysis (See Also “Process Capability Analysis,” p. 486)

Describe & Compare

How is the process behaving; is it in control?

Control Charts— 7QC Tool

Describe

What are the compon- Flowchart — ents of the process; 7QC Tool (See what is involved? Also “Process Map,” p. 522)

p. 316

Describe & Compare

Is the process able to meet customer requirements?

p. 486

Describe

What are the compon- Process Map— ents of the process; 7QC Tool what is involved?

Describe

What is the yield of a process?

Describe

Where is value being Value Stream added in the process? Analysis Conversely, where does waste exist in the process?

Describe

Where is value being added in the process? Conversely, where does waste exist in the process?

Process Capability Analysis

p. 217

p. 522

Rolled Throughput p. 610 Yield (See Also “Process Capability Analysis,” p. 486)

Value Stream Mapping (See Also “Value Stream Analysis,” p. 727)

p. 727

Summar y Tool Matrix Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

Schedule

What is the most efficient way to complete this process or project?

Activity Network Diagram (AND)— 7M Tool

p. 136

Schedule

What is the most efficient way to complete this process or project?

Arrow Diagrams (See “AND”)

p. 159

Schedule

What is the most efficient way to complete this process or project?

Critical Path Method (CPM) (See “AND”)

p. 242

Schedule

How long will the Gantt Chart project take? What are the key milestones, and when should you expect them?

p. 317

Schedule

What are the probabil- Monte Carlo ities and risks associated Simulation with several schedule possibilities?

p. 435

Schedule

What is the most efficient way to complete this process or project?

PERT (Program Evaluation and Review Technique) Chart

p. 453

Planning

What might go wrong during the planning of this complex, significant project?

Process Decision Program Charts (PDPC)—7QC Tool

p. 515

Scope

What are the project scope, boundary conditions, deliverables, budget and schedule?

Project Charter (See SMART Problem and Goal Statemants for Project Charter p. 665)

Project Management

continues

121

122

Encyclopedia Continued Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

HR, Deliverables & Tasks

Who is responsible for what?

RACI Matrix (Responsible, Accountable, Consulted, Informed)

p. 554

Scope

What is the scope of SIPOC (Supplierthe project or process? Input-ProcessOutput-Customer)

p. 663

Scope

What is the most succinct description of the project’s goal and problem statements?

SMART Problem & Goal Statement for a Project Charter

p. 665

Risk

Who supports the project, who doesn’t, and why?

Stakeholder Analysis

p. 681

Deliverables & Tasks

What are the project Work Breakdown deliverables, and what Structure (WBS) activities will produce them?

p. 753

Risk Management

What are the potential Cause and risk response strategies Prevention Diagram to a potential risk or failure? What can go wrong? FMEA: Failure What can be done to Modes and prevent or minimize it? Effects Analysis What are the potential failures that could occur, and what is the best response (action) plan to minimize its impact if it does happen?

p. 198

Summar y Tool Matrix Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

What are the potential Fault Tree Analysis root causes of a single (FTA) problem or problematic outcome and how can they be prevented or mitigated? Typically applied to a design of a system, product, or process that involves human interactions.

p. 309

What are the probabil- Monte Carlo ities and risk associated Simulation with several financial possibilities? (pricing and forecasting, for example)

p. 435

How best to prevent or correct in-process errors (often human mistakes)?

p. 462

Poka-Yoke

What might go wrong Process Decision during the planning of Program Charts this complex, signifi(PDPC)—7M Tools cant project?

p. 515

How best to plan for, manage, and mitigate unforeseen risk?

Risk Mitigation Plan

p. 601

Who supports the project and who doesn’t, and why?

Stakeholder Analysis

p. 681

What differences exist between two groups, if any?

Analysis of Variance p. 142 (ANOVA)—7M Tool

Root Cause

Compare & Predict

continues

123

124

Encyclopedia Continued Generic Hypothesis Testing Schematic Purpose

Compare & Predict

Key Question

Tool Name

Page

What are the potential causals of a problem or problematic outcome?

Cause-and-Effect p. 173 Diagram—7QC Tool (a.k.a. Fishbone, Ishikawa Diagram)

Which variables or Design of combination of Experiment (DOE) variables proves as the best in producing the desired results, based on experimental data?

p. 250

What is the root cause 5-Whys of this problem or problematic outcome?

p. 305

What are the potential Fault Tree Analysis root causes of a single (FTA) problem or problematic outcome and how can they be prevented or mitigated? Typically applied to a design of a system, product, or process that involves human interactions.

p. 309

What are the potential Fishbone Diagram— p. 316 causals of a problem or 7QC Tool (See also problematic outcome? Cause-and-Effect Diagram) Compare & Predict

Is this population (or Hypothesis Testing sample) of data different from another by chance alone, or because of an outside influence?

p. 335

How do the various Interrelationship cause-and-effect ideas Diagram—7M Tool relate to one another in a complex situation?

p. 369

Summar y Tool Matrix Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

Describe & Predict

What is the causeand-effect model that describes the process and its critical variables?

Regression Analysis

p. 571

Describe & Compare

Are these two factors correlated? What is the relationship between these two factors?

Correlation Analysis (See also Scatter Diagram— 7QC Tool, p. 640)

Describe

What are the critical parameters in the process?

Y = f(x)

p. 758

How best to collect a representative sample of the population?

Sampling

p. 618

Sampling

Predict

Variation Analysis and Process Evaluation

Describe & Compare

Is the process able to to meet customer requirements?

Capability Analysis p. 173 (See also Process Capability Analysis, p. 486)

Describe & Compare

How is the process behaving; is it in control?

Control Charts— 7QC Tool

p. 217

Describe & Compare

How accurate is the measurement system? Is the process truly performing the way the data seems to be reporting, or is the measurement system inaccurate?

Measurement System Analysis (MSA)

p. 412

continues

125

126

Encyclopedia Continued Generic Hypothesis Testing Schematic Purpose

Key Question

Tool Name

Page

Describe, Compare & Predict

Are the data distributed normally?

Normal Probability Plot (See also Control Charts— Normal Versus Non-normal Data section, p. 227)

p. 444

Describe & Compare

Is the process able to meet customer requirements?

Process Capability Analysis

p. 486

Voice of the Customer (VOC); Voice of the Business (VOB)

What are the major themes of ideas, opinions, issues, and so on found in large amounts of language data?

Affinity Diagram— p. 136 7M Tool

What product and/or services features and functionality do customers prefer and are willing to buy?

Conjoint Analysis

p. 207

How does my work Critical to Quality relate to the customer (CTQ) Matrix requirements and how do I know when I have fulfilled them?

p. 242

What matters to your KJ Analysis customers? What are the natural groupings, categories, or affinities of a large number of topics, ideas, and quotations, and how best to translate the verbal input into requirements?

p. 375

How best to capture the customer requirements?

p. 737

Voice of Customer Gathering Techniques

Activity Network Diagram (AND)—7M Tool

127

A Activity Network Diagram (AND)—7M Tool A

What Question(s) Does the Tool or Technique Answer? What is the most efficient way to complete this process or project?

B

An AND helps you to

D

• Manage the timing of activities and overall completion of a project

or process

C E F G

• Graphically organize process steps into the most efficient sequence

H

• Show the most critical path and any parallel paths

I

• Evaluate and reorganize the step sequence. It identifies any simulta-

J

neous tasks and tasks that will take the longest to complete • Identify any slack time—that is, the amount of time a non-critical

path task can be delayed without delaying the project • Manage resources and understand upstream and downstream

dependencies

K L M N O P

Alternative Names and Variations This tool is also known as

Q R S

• Activity chart

T

• Activity-on-Arrow (AOA)

U

• Arrow diagram or Arrow Diagramming Method (ADM) • Activity on Node (AON)

V W X

• Critical Path Method (CPM) or CPM chart

Y

• Network diagram

Z

• Node diagram • Precedence Diagram (PDM)

Variations on the tool include • Program Evaluation and Review Technique (PERT) chart. (See Also

“PERT (Program Evaluation and Review Technique) Chart,” p. 453)

128

A B C D E F G H I J K L M

Encyclopedia

When Best to Use the Tool or Technique The Activity Network diagram (AND) is an important Project Management tool. It helps to sequence the steps of the project into the most efficient order such that some steps are completed simultaneously. Most project managers will use the AND technique with the project team prior to developing the project schedule that commits the team to a target project completion date. The AND technique is helpful when managing a project whose deadline is critical and for which any delay has significant consequences—and conversely whose acceleration could yield great benefits. Another application of the AND tool is from the perspective of the process of interest. Similar to a process map, the AND technique evaluates the existing steps in a process to identify the critical path and any parallel activities. The critical path is comprised of a set of tasks that if delayed will delay the outcome of a process or the completion of the project. Even if one critical path activity is delayed, the timing of the overall process or project is negatively impacted. As a result, the critical path identifies the least flexible activities critical to maintaining the schedule and the float activities. The critical path also calculates slack time associated with the non-critical path activities to understand the flexibility of their start and finish without impacting the overall schedule determined by the critical path.

N O P Q R S T U V W

Brief Description The Activity Network diagram (AND) is a simple graphical representation of the process or project steps. It depicts which steps must be completed when to complete a project and in what sequence. It can be done either manually or aided by computer software. A typical AND includes the following symbols, shown in Figure A-1: • Rectangles, called nodes, represent the

process or project’s activities (hence, the technique is sometimes called Activity on Node).

X

• Arrows connect the activities and show

Y

dependencies between two activities.

Z

Activity A

Activity B

“From”

“To”

Figure A-1: AND Using Activity on Node Symbols

There are four types of dependencies. These dependencies describe the logical relationship between two activities. Sometimes they also are referred to as precedence relationships; hence, the technique is sometimes called the Precedence Diagramming Method (PDM). Figure A-2 illustrates the various types of dependencies.

Activity Network Diagram (AND)—7M Tool

129

1. Finish-to-Start—Where the “from” activity must finish before the

“to” activity can start. This is the most common dependency. 2. Finish-to-Finish—Where the “from” activity must finish before the

“to” activity can finish. 3. Start-to-Start—Where the “from” activity must start before the “to”

activity can start. 4. Start-to-Finish—Where the “from” activity must start before the

“to” activity can finish. This type of dependency is rarely used and is usually only used in the engineering discipline.

A B C D E

12-day total duration

7-day total duration

F

B

A

G

A

B

H

Finish-Finish

I

Finish-Start

Activity A

Activity B

5-day duration

7-day duration

Activity A

J

5-day duration

Activity B 7-day duration

7-day total duration

B Activity A Start-Start

A

M O P

Activity A

Q

5-day duration

R

Start-Finish

S

5-day duration

Activity B

L N

12-day total duration

B

A

K

Activity B

T

7-day duration

U

7-day duration

Figure A-2: AND Four Different Dependencies

V W

A less common set of symbols includes circles rather than rectangles, as illustrated in Figure A-3. However, this less common approach has the added nuance of using the arrows to represent the activities, rather than the shape, which is reversed from the more traditional approach. Both methods work equally well, and it is a matter of preference as to which technique to use. • Arrows represent activities (hence, the technique is sometimes called

Activity-on-Arrow). The tail of the arrow represents the beginning of the activity, and the head of the arrow represents its completion.

X Y Z

130

Encyclopedia • Connected at nodes showing the

A B C D E F G H I J K L M N O P Q R S T U V W

dependencies between two activities. The nodes often contain lettering in the center of the circle to identify it. As shown in Figure A-3, the “Prepare Spec” task also may be referred to as the AB activity. The Activity on Arrow technique, by design, uses the Finish-to-Start dependency.

Activity Node

A

Prepare Spec

B

Node Purchase Supplies

C

Design Document

D

Figure A-3: AND Using Activity-onArrow Symbols

The AND technique is a member of the 7M Tools, attributed in part to Dr. Shewhart, as seven management tools, sometimes referred to as the 7MP (or seven management and planning tools). These 7M Tools make up the set of traditional quality tools used to analyze quantitative data. The 7M Toolset includes 1) Activity Network diagrams or Arrow diagrams, 2) Affinity diagrams, 3) Interrelationship digraphs or Relations diagrams, 4) Matrix diagrams, 5) Prioritization matrices, often replacing the more complex Matrix data analysis, 6) Process decision program charts (PDPC), and 7) Tree diagrams. The Quality Toolbox by Nancy Tague presents the 7M Tools ranked from those used for abstract analysis to detailed planning: Affinity diagram, Relations diagram, Tree diagram, Matrix diagram, Matrix data analysis (commonly replaced by a more simple Prioritization matrix), Arrow diagram, and Process Decision Program Chart (PDPC). Technically speaking, the Critical Path Method (CPM) and the Program Evaluation and Review Technique (PERT) chart are estimating techniques applied to a completed Activity Network diagram, rather than a different type of AND. A critical path identifies the project path with the longest duration, which also determines the shortest time to complete the project. Hence, the critical path lacks any slack time; it represents the minimum amount of time to complete the project. Slack time is the amount of time a task can be delayed without delaying a project. In contrast, if parallel paths exist, they will contain slack time or a shorter duration than the critical path. How to calculate the critical path is described in this encyclopedia, in the PERT technique section under the Ps. (See Also “PERT (Program Evaluation and Review Technique) Chart,” p. 453)

X Y Z

How to Use the Tool or Technique This technique can be done individually or in a small group. When using the simple approach to develop an AND, the procedure involves the following steps: Step 1.

Identify the topic and then collect any data about the task. Examples include a. Brainstorming

Activity Network Diagram (AND)—7M Tool

131

b. Process maps c. Procedure manuals Step 2.

List the necessary tasks. One technique used is to create a master list and then to put one activity on either index cards or sticky notes to make it easy to sequence the tasks.

A B

Warning

C

Be sure you only assign one activity step per index card or sticky

D

note!

E F G

Place the tasks into a logical order sequence. Identify which tasks must precede others, and arrange them accordingly. If some tasks can occur simultaneously, put them on the same vertical plane (parallel to one another). If using sticky notes, follow the same procedure and place the activities on a flipchart sheet or section of the wall.

H

Continue this step until all the individual steps are placed in a logical flow. Be sure to leave space in-between each step to allow additional notes to be added.

M

Step 4.

Add times to the steps (usually above the activity card or note).

P

Step 5.

Determine the critical path.

Step 3.

I J K L N O Q

a. Identify all the possible path combinations from start to finish.

R

b. Sum the times of each activity in each of the possible path combinations and record the path and its respective duration.

T

S U V

c. Identify the longest path and highlight by bolding (highlighting or darkening) the path arrows—this is the critical path.

W

d. Calculate the Earliest Start (ES) times and Earliest Finish (EF) times based on how long the preceding task(s) take. Hence, the first activity has zero ES time, and its EF time equals the total duration for the task. Preceding tasks have a compounding (or cumulative) effect on subsequent tasks’ ES and EF times.

Z

i. ES = the latest EF from the preceding task ii. EF = the ES time minus the actual task time

X Y

132

Encyclopedia

e. Calculate the Latest Start (LS) times and Latest Finish (LF) times by starting at the end of the network (project or process completion) and working backward toward the start activity. i. LF = the shortest LS of the subsequent task A

ii. LS = the LF time minus the actual task time

B

f. Calculate the slack times for each task and for the project. Total slack is the time an activity can be delayed without affecting the project schedule. Free slack is the time an activity can be delayed without affecting the early start of any subsequent task.

C D E F

i. Total slack = LS minus ES = LF minus EF

G H

ii. Free slack = the earliest ES of all subsequent tasks minus EF

I J

Step 6.

Examine and adjust the diagram as needed.

K L M N

How to Analyze and Apply the Tool’s Output After the Activity Network diagram is complete, use the output to

O

• Reveal the interdependencies of activities.

P

• Determine what elements are critical to meeting the completion date

Q R S T U V

(the critical path), including necessary resources and time requirements. • Determine projected completion date. • Facilitate “what if” discussions (or alternatives) as to what could be

shortened or rearranged; consider any trade-offs. • Highlight critical activities that will require monitoring and provide

W

a baseline to evaluate the actual process or project performance. Acquire a tool for planning and predicting.

X

• Help to manage uncertainties, particularly where slack time exists.

Y Z

Examples Next is an example of an AND depicting six activities with a Finish-Start dependency. The list of activities includes the following: • Activity A @ 3-day duration • Activity B @ 5-day duration

Activity Network Diagram (AND)—7M Tool

133

• Activity C @ 7-day duration • Activity D @ 4-day duration • Activity E @ 9-day duration • Activity F @ 3-day duration

This 6-step scenario with its critical path and slack time calculations can be found in Figure A-4.

A

Slack: 0

Slack: 0

Slack: 12

ES: 0

EF: 3

ES: 3

EF: 8

ES: 8

EF: 12

ES: 12

EF: 15

LS: 0

LF: 3

LS: 3

LF: 8

LS: 20

LF: 24

LS: 24

LF: 27

Activity A

Activity B

Activity D

Activity F

3 days

5 days

4 days

3 days

Slack: 0

C D E

Slack: 0

G H

ES: 8

EF: 15

ES: 15

EF: 24

LS: 8

LF: 15

LS: 15

LF: 24

I

Activity C

Activity E

J

7 days

9 days

• A-B-D-F = 15

days days

B

F

This example’s path calculations include

• A-B-C-F = 18

Slack: 12

Figure A-4: 6-Step AND Example

• A-B-C-E-F = 27 days

K L M N

The longest path is A-B-C-E-F, at 27 days; therefore, it is the critical path.

O

The calculations for the Earliest and Latest times are found in Figure A-4, and the formulas are as follows:

P

• ES = the latest EF from the preceding task

Q R S

• EF = the ES time—the actual task time

T

• LF = the shortest LS of the subsequent task

U

• LS = the LF time—the actual task time

The calculations for the slack times are • Total slack = LS—ES = LF—EF; • Free slack = the earliest ES of all subsequent tasks—EF

Note There is zero slack time along the critical path.

V W X Y Z

134

Encyclopedia

Hints and Tips A summary activity, called a hammock, represents a group of related activities as one to show the lapsed time for the collection of tasks. A B

The hammock spans between two activities that may be separated by several other activities, as illustrated in Figure A-5.

C D E F G H I J K L M N O P Q R S T U V X Y Z

gramming method [or Activity-onArrow (AOA)]:

Activity A

Activity B

Activity D

Activity C

Activity E

Activity F

• No activity

can start before its predecessors have been completed.

Activity H

Figure A-5: AND Hammock (as Activity H)

• The arrow

length has no meaning; it simply represents logical precedence. • Each activity must have a unique identifier (a number or alpha);

duplicates are not permitted. • If an Arrow Diagramming method is used, only finish-to-start

dependencies are appropriate. • AOA may use dummy activities (with Prepare Purchase Design zero time durations) to help comSpec Supplies Document A B C D plete the logic in the network diagram by showing the flow of Dummy Activity activities. Dummy activities are idenE tified with hashed or dotted arrows, as shown in Figure A-6. They typiFigure A-6: Activity-on-Arrow with cally show relationships between Dummy Activity activities that have more than one predecessor by keeping the sequence correct.

tify No am Te

W

• Arrow Dia-

• Non-sequential steps, such as loops or conditional branches using deci-

sion-diamonds are not used in any of the AND techniques. These

Activity Network Diagram (AND)—7M Tool

135

special conditions are found in process mapping and more specialized diagramming techniques called Graphical Evaluation and Review Technique (GERT) and Systems Dynamics modeling.

Supporting or Linked Tools Supporting tools that might provide input when developing an AND include

B

• VOC Data Gathering Tools and Techniques, such as surveys and

D

interviews (See Also “Voice of Customer Gathering Techniques,” p. 737)

A C E F

• Written reports

G

• Brainstorming sessions (See Also “Brainstorming Technique,” p. 168)

H I

A completed AND provides input to tools such as

J

• Tree diagram and matrix (See Also “Tree Diagrams,” p. 712)

K

• Fishbone (See Also “Cause-and-Effect Diagrams,” p. 173)

L

• Simple matrix (See Also “Matrix Diagrams,” p. 399)

Figure A-7 illustrates the link between the AND and its related tools and techniques.

M N O P Q

Brainstorm

R S

Activity list

Process Map

T U

Process Map

Activity Network Node (AND)

FMEA

V W X

Brainstorming

Cause-Effect Diagram

Y Z

Value Stream Activity Matrix

Figure A-7: Activity Network Diagram Tool Linkage

136

Encyclopedia

Affinity Diagram—7M Tool

A

What Question(s) Does the Tool or Technique Answer? What are the major themes of ideas, opinions, issues, and so on found in large amounts of language data?

B

An Affinity diagram helps you to

C D E F G

• Organize large volumes of language-based data (for example, text or

verbal input) into related categories (natural groupings, categories, or affinities of a large number of topics, ideas, and quotations). • Show a parent-child relationship between more detailed ideas to

their higher-order themes, also called affinity groupings.

H I J K L M N O P Q

Alternative Names and Variations This tool is also known as • Affinity chart • KJ Method or analysis (See Also “KJ Analysis,” p. 375) • Idea Map

Variations on the tool include

R

• Fishbone diagram (See Also “Cause-and-Effect Diagram,” p. 173)

S

• Tree diagram or matrix (See Also “Tree Diagram,” p. 712)

T U V W X Y Z

When Best to Use the Tool or Technique The tool organizes a large number of ideas, quotations, facts into themes. This technique is also helpful when the topic is complex and lacking apparent themes or categories. It may be applied to arrange brainstorming ideas into categories or analyze customer interview data or survey results. When facts or thoughts are in disarray, this approach organizes the data. When the main ideas of the issues seem too large or complex to grasp, the Affinity diagram helps to reveal the key themes. Often this tool is used as an interim step prior to putting data into a Fishbone or Tree diagram. This tool often accompanies a brainstorming session to organize and sort through the ideas generated.

Affinity Diagram—7M Tool

Brief Description A Japanese anthropologist, Jiro Kawakita, developed the Affinity diagram in the early 1960s; hence, it is also referred to as the KJ Method. [Recall: A Japanese name reverses the order of the given and sir name relative to an English name.] Dr. Kawakita was the founder of the Kawayoshida Research Center. Today’s use of Affinity diagrams is less complex than the original KJ Analysis approach developed by Jiro Kawakita. The reference to this tool often is used hand-in-hand with brainstorming as a means to organize the brainstormed ideas. Hence, this entry discusses the common application of the Affinity diagram. In a separate entry, the more in-depth KJ approach can be found in the section under the Ks. (See Also, “KJ Analysis,” p. 375) The Affinity diagram tool is a member of the 7M Tools, attributed in part to Dr. Shewhart, as seven management tools, or sometimes referred to as the 7MP, for seven management and planning tools. These 7M Tools make up the set of traditional quality tools used to analyze quantitative data. The 7M Toolset includes 1) Activity Network diagrams or Arrow diagrams, 2) Affinity diagrams, 3) Interrelationship digraphs or Relations diagrams, 4) Matrix diagrams, 5) Prioritization matrices, often replacing the more complex Matrix data analysis, 6) Process decision program charts (PDPC), and 7) Tree diagrams. The Quality Toolbox, by Nancy Tague, presents the 7M Tools ranked from those used for abstract analysis to detailed planning: Affinity diagram, Relations diagram, Tree diagram, Matrix diagram, Matrix data analysis (commonly replaced by a more simple Prioritization matrix), Arrow diagram, and Process Decision Program Chart (PDPC).

137

A B C D E F G H I J K L M N O P Q R S

How to Use the Tool or Technique This technique can be done individually or in a small group. The procedure to develop an Affinity diagram involves the following steps:

U

Step 1.

V

Identify the topic and then collect and/or capture the text or verbal data. Examples of different means of gathering such input include

W X

a. Interviews

Y

b. Brainstorming

Z

c. Surveys Step 2.

T

Prepare the data by putting one idea, thought, or suggestion (either written or verbal input) on either

138

Encyclopedia

a. Single piece of paper (or index cards) or b. Sticky note (Only one idea per piece of paper (index card or sticky note). A B C D E

Note If the data already exists in a text document, a short cut may be to simply cut the document into strips, such that one strip of paper contains only one written idea.

F G H

Step 3.

I J

Sort the ideas. Review each individual piece of paper and begin to place them in piles whose subjects relate to one another (those that have a similarity or an “affinity” for one another). If using sticky notes, follow the same procedure, but place the ideas with a common theme on a single flip chart sheet or section of the wall. Continue this step until all the individual ideas are placed in a group.

K L M N

a. A few ideas may not fit well into the groupings or themes; they may be independent ideas or “loners.” If so, place them in a “Miscellaneous” grouping.

O P Q R S T U V W X Y Z

Step 4.

Organize the ideas by reviewing each within a grouping or affinity

Preparing data

Sorting data

Organizing data

Naming data Topic

Figure A-8: Affinity Technique Illustration of Preparing, Sorting, Organizing, and Naming Steps

Affinity Diagram—7M Tool

139

and then sort them again into further subsets. Identify those that are similar and group them. a. Look across groups and determine if any groups are related to one another. Step 5.

Name the themes by first identifying a category heading or the highest-level of commonality in a given group. Next, name any of the subsets within a group, if possible, until all the groupings have a label. a. Reexamine if the “Miscellaneous” ideas relate to any of the named groupings. If so, move them to that category.

Figure A-8 illustrates at a high level the flow of the Preparing, Sorting, Organizing, and Naming steps of the Affinity technique.

A B C D E F G H

How to Analyze and Apply the Tool’s Output After the Affinity diagram is complete, its contents (categories and their respective detail) can be transferred to a format that better communicates the themes and respective linkages, such as one of the following formats: • Matrix—Either one category or group per matrix or one per column,

depending on how complex the topic or diverse the data • Tree diagram—Either one category or group per tree or one major

grouping per major branch, depending on how complex the topic or diverse the data • Fishbone diagram—Either one category or group per tree or one

major grouping per major branch, depending on how complex the topic or diverse the data

I J K L M N O P Q R S T U

Examples An example of an Affinity diagram depicting the selection criteria for choosing a restaurant can be found in Figure A-9. Figure A-10 illustrates the same “Criteria for Selecting a Restaurant” example but as a Tree diagram, using Quality Companion by MINITAB.

V W X Y Z

140

Encyclopedia

Cuisine

Atmosphere

Service

Location

Ethnic

Dress Code

Quick Food

Proximity to Home

Vegetarian

Quiet versus Noisy

Professional Wait Staff

Safe Neighborhood

Kids Menu

Music (live or piped)

Wine List

Table Settings (cloths, candles, flowers)

A B C D E

Variety of Desserts

F

Plenty of Parking

Reservations Required

Landscape / Setting Prompt Seating

Lighting (Soft / Dim versus Bright)

Prompt Service

Patrons (Adults versus Families

Able to Linger

“Fast Food”

G H

Cost

Daily Specials

I J

Figure A-9: Example Affinity Diagram—Criteria for Selecting a Restaurant

K

Selecting a Restaurant

L M N O P Q R S T U V

Cuisine _

Atmosphere _

Y Z

Location _

Cost

Ethnic

Dress Code

Quick Food

Proximity to Home

Vegetarian

Quiet versus Noisy

Professional Wait Staff

Safe Neighborhood

Kids Menu

Music (live or piped)

Reservations Required

Plenty of Parking

Wine List

Table Settings (cloths, candles, flowers)

Variety of Desserts Fast Food

W X

Service _

Daily Specials

Prompt Seating

Landscape / Setting

Prompt Service Lighting (soft versus bright)

Able to Linger

Patrons (Adults versus Families

Figure A-10: Example Affinity Tree Diagram—Criteria for Selecting a Restaurant

Affinity Diagram—7M Tool

141

Hints and Tips • Take time to complete the diagram. Initially start it in a concentrated team session but allow the team members to reflect on it and edit it with any additional ideas. • Multiple perspectives are important. Ensure that individuals with different perspectives (that is, different roles, different levels, different functional responsibilities) have an opportunity to contribute ideas. • If a perspective is missing from the initial creation of the Affinity diagram, have the individuals contribute as “editors.” • The naming of the Affinity categories is often best done at the end of the building process, rather than at the beginning. • If an idea can fit in more than one grouping, create a duplicate note(s) and place it in all of the applicable places. This may trigger additional ideas or show unique relationships that should not be lost. During the organizing step, a stronger affinity with one group versus the others may become evident and outweigh the other placements. If so, the team may choose to reduce the idea to one representation.

A B C D E F G H I J K L M N O P Q

Supporting or Linked Tools Supporting tools that might provide input when developing an Affinity diagram include • VOC Data Gathering Tools and Techniques, such as surveys and

interviews (See Also “Voice of the Customer Gathering Techniques,” p. 737)

R S T U V W

• Written reports

X

• Brainstorming sessions (See Also “Brainstorming Technique,” p. 168)

Y

A completed Affinity diagram provides input to tools such as • Tree diagram and matrix (See Also “Tree Diagram,” p. 712) • Fishbone (See Also “Cause-and-Effect Diagram,” p. 173) • Simple matrix (See Also “Matrix Diagrams,” p. 399)

Z

142

Encyclopedia

Figure A-11 illustrates the link between the Affinity diagram and its related tools and techniques.

A B C

VOC Data Gathering (Survey, Interviews)

Brainstorming

Matrix

Affinity Diagram

Tree Diagram

D E

Written Report

Fishbone

F G

Figure A-11: Affinity Diagram Tool Linkage

H I J K L M N O

Variations KJ Analysis The KJ Analysis involves more detail and evaluation than what is described in this section. (See Also “KJ Analysis,” p. 375)

Analysis of Variance (ANOVA)—7M Tool

P Q R

What Question(s) Does the Tool or Technique Answer? What differences exist between two groups, if any?

S

The ANOVA helps you to

T U V W X Y Z

• Make comparisons among many items. Are their means different?

Are their variances different? • Determine the relationship between groups, within some degree

of confidence, using statistical analysis to analyze and compare variables. • Make inferences as to whether an “improvement initiative” shows

statistically significant differences compared with a “non-intervention” group.

Analysis of Variance (ANOVA)—7M Tool

143

Alternative Names and Variations This tool is also known as • A type of hypothesis testing • F-test

A

• One-way ANOVA

B

• Two-way ANOVA without replicates

C

• Two-way ANOVA with replicates

D

• Analysis of Means (ANOM), a type of ANOVA when comparing

three or more means • Test for Equal Variances, a type of ANOVA when comparing three or

more variances

E F G H I J

When Best to Use the Tool or Technique Use ANOVA when comparing among many items. It determines if the average of a group of data is statistically different from the average of other (multiple) groups of data. It helps draw conclusions about a hypothesis that determines if two or more populations are different. This technique can be used when analyzing either of the following: • The current environment to determine if the variation is due to natu-

ral (common) variation or a special cause variation • An intervention (or improvement) that was newly implemented

ANOVA tests for a difference between two or more means and compares and analyzes the data from group to group and the variation within a group. For example, it could analyze the output of multiple machines, examining the output from each machine and from machine to machine. A similar analysis could be achieved by running multiple two-sample t-Tests. However, ANOVA is more efficient by essentially running the required two-way comparisons simultaneously. ANOVA is a statistical technique sometimes called a “poor man’s” Design of Experiment (DOE). A DOE is a structured, organized method for determining the relationship between factors (Xs) affecting a process and the output of that process.

Brief Description Ronald Fisher developed the ANOVA tool in 1920. He designed this statistical method to conduct group tests that determined if differences were

K L M N O P Q R S T U V W X Y Z

144

Encyclopedia

statistically significant. Group tests involve multiple data sets and multiple independent and dependent variables, similar to Regression Analysis. Fisher’s work was important because it examined the impact of interventions and whether any variance was because of the intervention or random chance. A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Part of the technique uses a statistical distribution to which the calculations are compared, called an F-statistic, named for its inventor, Fisher. The F-statistic is the ratio of the explained variation to the unexplained variation and is compared with the F-distribution. The technique relies on the assumption that the data comes from a normal distribution. The formula for F-statistic is Factor Variance divided by the Error Variance [F=Factor Variance/Error Variance]. The null hypothesis (H0) would be that the F-distribution would not be a good fit for the model; therefore, the p-value would be greater than the significance level. The p-value is the probability that the calculated Fstatistic would be large if the group means were equal. The hypothesis would be written as the following mathematical sentences: • Null hypothesis: H0: means of all tested factors are equal • Alternative hypothesis: Ha: at least two tested factors have different

means

Typically, the default significance level is 5% [also called “alpha” (α)]. If the calculated ratio (called F-calculated) is greater than the F-critical for the specified significance level found in the F-distribution table, then the null hypothesis is rejected. The F-statistic has a negative correlation to the p-value. As the F-statistic increases, the explained variance grows, and the distance between the two means increases as the p-value decreases to show that there is greater confidence in two means being statistically different. If the p-value is less than or equal to the significance level (α), then at least one source of explained variation is significant in helping the distribution fit the F-distribution. That is if the p-value is less than 0.05, for example, then at least one group of data is different from at least one other group. This paragraph may become clearer after reviewing the subsequent procedure section, “How to Use this Tool or Technique.” For more discussion on p-value, reference the Glossary and “Hypothesis Testing,” p. 335. ANOVA is a type of hypothesis testing. It determines the source of variation, whether due to a technician (or operator), the part (or process), the interaction, and/or the error. The following terms describe some of the sources of variance:

Analysis of Variance (ANOVA)—7M Tool

145

• Repeatability—When the variation comes from repeated measure-

ments of the same item, by the same person, with the same measuring device. This is variation within a factor and is also called Error or Error Variance. • Reproducibility—When the variation source is the averages from

repeated measurements (or factors) made by different people (or machines) on the same item. This is variation between factors and is also called Operator or Technician Error.

A B C

ANOVA is a flexible technique. It is used to identify a statistically significant difference between means of three or more levels of a factor. It is used when two or more means (a single factor with three or more levels) must be compared with each other. The technique falls into three different categorizes. Why and when to use each different type of ANOVA is outlined as follows.

D

One-Way ANOVA (sometimes referred to as the “fixed effects” ANOVA) When to Use—Compares two components or sources of variance, treatments, and experimental error.

H

Example—Analyzes one factor from one source of variation and one measurer.

Two-Way ANOVA without Replicates When to Use—Compares three components of variance (Factor A treatments, Factor B treatments, and experimental error). Example—Analyzes two factors and only one technician (or measurer) unless the measurer is one of the factors being analyzed.

Two-Way ANOVA with Replicates When to Use—Compares three components of variance (Factor A treatments, Factor B treatments, and experimental error), with multiple repetitions. Example—Analyzes two factors with multiple repetitions of each combination and only one technician (or measurer) unless the measurer is one of the factors being analyzed. In preparation for conducting an ANOVA, collect the data to be analyzed. This technique can use any number of data sets, which can be of unequal size. Within each data set, the sample size can be as small as three or four measurements or as large as necessary and economically feasible. The structure of an ANOVA test has an independent variable (a category of interest) and a dependent variable(s), which is hypothesized to be affected by differing levels of treatments (or interventions). For example, if the dissolution time of different types of tablets were of interest, tablets would be the category (independent variable), and the dependent variable would be the time it takes for the different tablets to dissolve. Typically, the sample size is 20 or more per group.

E F G I J K L M N O P Q R S T U V W X Y Z

146

A B C D E F G H I J K L M N

Encyclopedia

How to Use the Tool or Technique This technique can be done manually, computing the statistical calculations by hand and referencing an F-statistic table. First run an ANOVA by hand before using the computer programs, to help understand where the mechanics and source of the calculations. Afterward, it is wise to use a computer for assistance. Excel (using statistical add-ins) can run an ANOVA. The current software packages that specialize in statistical analysis simplify the technique procedures and reduce any calculation errors. Two procedures are provided: 1) the manual calculation to understand the mechanics behind generating the ANOVA output and 2) the computer aided approach. While there are several good statistical application software packages available, this book uses MINITAB. ANOVA tests are best done with a minimum sample size of 20 per group. To simplify the explanation of the process, the following examples use less than 20.

One-Way ANOVA Manual Process Step 1. State the Null (H0) and Alternative (Ha) hypotheses. – a. Null hypothesis: H0: means (X ) of all 3 factors are equal – – – – i. X A =X B =X C; where (X ) “x-bar” is the mean b. Alternative hypothesis: Ha: at least two factors have different means

O P Q

Step 2.

a. Default is α = 0.5

R S

Choose the significance level or risk level (α)

Step 3.

Randomly select the parts or processes.

Step 4.

Identify the factors (that is, parts, processes, and technicians).

V

Step 5.

Collect the data.

W

Step 6.

Calculate the F-statistic (or F-calculated) using the sample data.

T U

X Y Z

a. Sum up the values for each factor. b. Calculate the average value (sum divided by the number of trials). c. Calculate the variance (s2) (sum of the squared difference – of each value x) from the average (X ), then divide by the sample size minus 1. s2
(Σ( x   x )2 )  (n  1)

Analysis of Variance (ANOVA)—7M Tool

147

d. Calculate the various Sum of Squares (SS): i. Total Sum of Squares (Total SS): Total SS
Σ( x)2  CM Where CM stands for the “Correction for the Mean” and “N” is the total population, determined by CM
( Σx )2  N ii. Part Sum of Squares (Part SS), sometimes called the Machine SS, and “n” is the sample size: Part SS
[( Σx A )2  n A]  [( Σx B )2  n B]  [( Σx C )2  n C]  CM iii. Error Sum of Squares (Error SS) Error SS
Total SS  Part SS Where the “Part Sum of Squares” represents all “Other” Sum of Squares that are calculated. e. Determine the degrees of freedom (df) = (n—1). f. Calculate the Mean Squares (MS) = [Sum of Squares/df]. g. Calculate the F-statistic = [MS for Part/MS for Error].

A B C D E F G H I J K L M N

Step 7.

Find the F-critical (or F-α) in the F-distribution table.

O

Step 8.

Compare the F-statistic to the F-critical.

P

One-Way ANOVA Manual Example—Dissolving Rate of Different Types of Tablet Step 1. State the Null (H0) and Alternative (Ha) hypotheses.

Q R S T

a. Null hypothesis: H0: means of all three factors are equal— Average time of A = Average time of B = Average time of C.

U

b. Alternative hypothesis: Ha: at least one average time is different from the other means.

W

Step 2.

Choose the significance level or risk level (α)— α =0.5, or 95% confidence.

Step 3.

Randomly select the parts or processes—Dissolving three different types of tablet (Brand A, Brand B, and a generic C).

Step 4.

Identify the factors (that is, parts, processes, and technicians)— Dissolution rate in seconds; one person stirring and measuring the time (in seconds) for the different types of tablet to dissolve in room temperature water.

V X Y Z

148

Encyclopedia

Step 5.

Collect the data in seconds. Tablet Type

Trial 1

Trial 2

Trial 3

Trial 4

Trial 5

Brand A

34

39

42

47

40

A

Brand B

55

48

61

49

54

B

Generic C

45

41

47

51

49

C D

Step 6.

Calculate the F-statistic (or F-calculated) using the sample data.

E F

Variance (S2)

Sum of each X2

Tablet Type

Trial 1

Trial Trial Trial 2 3 4

Trial 5

n

Brand A

34

39

42

47

40

5 202

40.4 4.72

8250

J

Brand B

55

48

61

49

54

5 267

53.4 5.22

14367

K

Generic C 45

41

47

51

49

5 233

46.6 3.84

10917

L

Totals

15 702

234

G H I

M

Sum

Avg

32.51 33534

N

CM = (702)2/15 = 32853.6

O

Total SS = Sum of each X2–CM = 33534–32853.6 = 680.4

P

Part SS = (202)2/5 + (267)2/5 + (233)2/5–32853.6 = 422.8

Q

Sum of Squares

680.40

Error SS = 680.4–422.8 = 257.6

R S

Mean Square = Sum of Square/Degrees of Freedom, per row.

T

Mean Square for Part = 422.8/2 = 211.4

U

Mean Square for Error = 257.6/12 = 21.5

V

F-statistic = Mean Square for Part/Mean Square for Error

W

F-statistic = 211.4/21.5

X Y Z

Source

Sum of Squares

Degrees of Freedom

Mean Square

F-statistic

F-critical

Part

422.8

2

211.4

9.8

3.89

Error

257.6

12

21.5

Total

680.4

14

Analysis of Variance (ANOVA)—7M Tool

149

Degrees of Freedom: • Total df = 15 total samples–1 = 14 • Parts df = 3 tablet types–1 = 2 • Error df = Total df–Parts df = 14–2 = 12 Step 7.

Find the F-critical (or F-α) in the F-distribution table. Look up F-critical for 5% significance, and the numerator df of 2 (Parts) and denominator df of 12 (Error) = 3.89

Step 8.

A B C D

Compare the F-statistic to the F-critical.

E

F-statistic of 9.8 > F-critical of 3.89; therefore Reject the Null Hypothesis.

F

This means that the dissolving time for at least one type of tablet differs from another.

One-Way ANOVA Automated Process Using MINITAB for the Same Tablet Example Step 1. State the Null (H0) and Alternative (Ha) hypotheses. – a. Null hypothesis: H0: means (X ) of all 3 factors are equal – – – – i. X A =X B =X C; where (X ) “x-bar” is the mean. b. Alternative hypothesis: Ha: at least two factors have different means.

G H I J K L M N O P Q

Step 2.

Choose the significance level or risk level (α); select α =0.5.

Step 3.

Randomly select the parts or processes.

S

Step 4.

Identify the factors (that is, parts, processes, and technicians).

T

Step 5.

Collect the data and record in MINITAB Worksheet, as illustrated in Figure A-12.

Step 6.

Calculate the F-statistic (or Fcalculated) using the sample data. a. Select the following commands: Stat > ANOVA > One-Way (Unstacked)…… (if the recorded data is in columns).

R

U V W X Y Z

Figure A-12: Data Recorded in MINITAB Worksheet

b. Put data into the Responses field and select acceptable confidence level in the ANOVA dialog box, as illustrated in Figure A-13.

150

Encyclopedia

A B C D E F G H I J K

Figure A-13: Tablet Example in MINITAB ANOVA Dialog Box

c. Select OK, and the following output is displayed in the session window, as illustrated in Figure A-14.

L M N O P Q R S T U

Figure A-14: Tablet Example—ANOVA Output in MINITAB Session Window

V W X Y Z

Two-Way ANOVA Without Replicates Recall: Two-way ANOVA has three components of variance (two factors and experimental error) but no replicates. Because this is similar to the Two-Way ANOVA With Replicates, this book discusses the more complex of the two approaches. Either way, Two-Way ANOVA adds a level of complexity that is best addressed using application software for statistical hypothesis testing.

Two-Way ANOVA With Replicates—Manual Process Recall: Two-way ANOVA has three components of variance (two factors and experimental error) and multiple replicates of the factor combinations.

Analysis of Variance (ANOVA)—7M Tool

Setup: The procedure is essentially the same as the One-Way ANOVA; however, two factors are being studied. In the case of the tablet test, there are the same three types of tablet used in the OneWay ANOVA, but this test involves varying the temperature of the water in which the different tablets will be dissolved by five degrees. Per the water set-up procedure, the lab analyst dissolved each type of tablet five times (replicates). The sample data is described next.

Example Dissolving Rate of Different Types of Tablet in Different Temperatures of Water Step 1. State the Null (H0) and Alternative (Ha) hypotheses. a. Null hypothesis: H0: there is no difference between the means of one type of tablet in the different water temperatures (the interaction) and no difference between the means of the different tablets; all are equal. b. Alternative hypothesis: Ha: at least one factor causes the tablet to have different means, either the variation among the different types or the interaction between the tablet and the different water setups.

151

A B C D E F G H I J K L

Step 2.

Choose the significance level or risk level (α)— α =0.5, or 95% confidence.

M

Step 3.

Randomly select the parts or processes—Dissolving three different types of tablet (Brand A, Brand B, and a generic C).

O

Step 4.

Identify the factors (that is, parts, processes, and technicians)— Dissolution rate in seconds; one person stirring and measuring the time (in seconds) for the different types of tablet to dissolve in the different water setups, varying by five degrees.

Step 5.

Collect the data—five runs per tablet-type, per water setup, in seconds.

N P Q R S T U V

Tablet Type

Hot

Warmer

Room Temp Cooler

Cold

Brand A

34

39

42

47

40

33

35

39

48

48

Y

33

37

38

47

47

Z

33

35

42

45

45

32

37

40

47

48

W X

continues

152

Encyclopedia Tablet Type

Hot

Warmer

Room Temp Cooler

Cold

Brand B

55

48

61

49

54

50

52

60

50

50

A

55

53

55

52

54

B

54

52

59

49

50

53

52

60

50

55

45

41

47

51

49

44

43

49

49

53

H

42

40

49

53

49

I

45

43

49

51

53

J

43

44

46

51

55

C D E

Generic C

F G

K L M

Step 6.

Calculate the F-statistic (or F-calculated) using the sample data.

Tablet Type

Summary

Hot

Warmer

Room Temp

Cooler

Cold

Total

Brand A

Count

5

5

5

5

5

25

Sum

165

183

201

234

228

1011

Average

33

36.6

40.2

46.8

45.6

40.44

Variance

0.50

2.79

3.20

1.20

11.89 31.75

Count

5

5

5

5

5

25

Sum

267

257

295

250

263

1332

W

Average

53.4

51.4

59

50

52.6

53.28

X

Variance

4.28

3.80

5.48

1.49

5.81

13.37

Count

5

5

5

5

5

25

Sum

219

211

240

255

259

1184

Average

43.8

42.2

48

51

51.8

47.36

Variance

1.69

2.69

1.99

1.99

7.18

17.72

N O P Q R S T

Brand B

U V

Y Z

Generic C

Analysis of Variance (ANOVA)—7M Tool

153

Tablet Type

Summary

Hot

Warmer

Room Temp

Cooler

Cold

Total

Total

Count

15

15

15

15

15

75

Sum

651

651

763

739

750

Average

43.4

43.4

49.1

49.27

50

A

Variance

76.21

42.5

66.75

4.80

17.39

B C

CM = (3527)2/75 = 165,863 Total SS = Sum of each X2–CM = 169,437–165,863 = 3574 Tablet SS = (Sum of Row Square–CM) = (1011)2/25 + (1332)2/25 + (1184)2/25–165,863 = 2065

D E F G

Water SS = (Sum of Column Square–CM) = 665

H

Interaction SS = (Interaction Square–CM–SS Column–SS Row) = 622; where the Interaction Square = Sum of all the Cell Square = (Sum Cell)2/k; where k = #replicates per cell.

I

Error SS = Total SS–All other SS = 222 Mean Square = Sum of Square/Degrees of Freedom, per row.

J K L M

Mean Square for Tablets = 2065/2 = 1032.5

N

Mean Square for Water Temp = 665/4 = 166.2

O

Mean Square for Interaction = 622/8 = 77.8

P Q

Mean Square for Error = 222/60 = 3.7

R

F-statistic = Mean Square/Mean Square for Error

S

F-statistic for Tablets = 1032.5/3.7

T

F-statistic for Water Temp = 166.2/3.7

U

F-statistic for Interaction = 77.8/3.7

V W

Source

Sum of Squares

Degrees of Freedom

Mean Square

F-statistic

F-critical

Tablets

2065

2

1032.5

279.05

3.15

Water Temp

665

4

166.2

44.92

2.53

Interaction

622

8

77.8

21.02

2.10

Error (or Within)

222

60

3.7

Total

3574

74

X Y Z

154

Encyclopedia

Degrees of Freedom: • Total df = 75 total samples–1 = 74 • Tablets (rows) df = 3 tablet types–1 = 2 • Water Temp (columns) df = 5–1 = 4

A B

• Interaction (column df times row df) = 2 x 4 = 8

C

• Error df = (Total df–Rows df–Column df–Interaction df )=

14–2 = 12

D E

Step 7.

F

Find the F-critical (or F-α) in the F-distribution table. Look up F-critical for 5% significance.

G

Tablets: numerator df of 2 and denominator df of 60 (Error) = 3.15

H I

Water: numerator df of 4 and denominator df of 60 (Error) = 2.53

J

Interaction: numerator df of 8 and denominator df of 60 (Error) = 2.10

K L M N O P Q R S T U V W X Y Z

Step 8.

Compare the F-statistic to the F-critical. F-statistic of each factor > F-critical; therefore Reject the Null Hypothesis. This means that each of the factors has different means, including the interaction between the tablet and the different water setups.

Two-Way ANOVA With Replicates—Automated Process Compare with the ANOVA Two-Way output for the same example. Select the following commands: Stat > ANOVA > One-Way to bring up the dialog box as illustrated in Figure A-15. Select OK, and the following output will be displayed in the session window, as illustrated in Figure A-16. Figure A-15: Tablet Example in MINITAB ANOVA Twoway Dialog Box

Analysis of Variance (ANOVA)—7M Tool

155

A Figure A-16: Tablet Example—ANOVA Two-Way Output in MINITAB Session Window

B C D

Note

E

All three p-values are less than 0.05. Hence, there is sufficient data to

F

reject the Null Hypothesis, meaning that at each of the factors have

G

different means, including the interaction between the tablet and the

H

different water setups.

I J

At this level of complexity, clearly the application software approach is more efficient and minimizes errors than calculating the results manually.

K L M

How to Analyze and Apply the Tool’s Output For the manual process, compare the calculated F-statistic to the F-critical to determine if the Null hypothesis was correct or not.

N O P Q

• If the F-statistic is greater than the F-critical, then the Null Hypothe-

sis of all factors’ averages being equal can be rejected. This means that at least one factor’s mean was different from another’s. • In the tablet example, the calculated F-statistic was 9.8, which was greater than the F-critical of 3.89; therefore Reject the Null Hypothesis. [Note that the F-critical requires an F-Distribution table for the selected significance (for example, 5%). Look up the F-critical at the intersection of the column with the numerator df of 2 (Parts) and row with the denominator df of 12 (Error). This intersection is 3.89.] For the automated process, there are two pieces of output used to indicate if the Null hypothesis was correct or not. Examine the output in the session window (shown in Figure A-14). The following data determines the results: • The “P” column in the ANOVA table connotes the p-value. If the pvalue < 0.05 (for a 5% significance level), then at least one factor is significant and differs from another; therefore, reject the Null

R S T U V W X Y Z

156

Encyclopedia

hypothesis. The converse is true if the p-value > 0.05, wherein there would be insufficient evidence to reject the Null. • In the tablet example, the p-value = 0.003, which is less than 0.05 (for a 5% significance level), so the Null hypothesis is rejected. A B C D E

• The graph in the lower portion of the session window shows the

confidence interval as a range for each tablet type. Notice that Brand A and Brand B’s ranges do not overlap. Hence, with 95% confidence, Brand A can be said to be statistically faster (on average 40.4 seconds) than Brand B (with an average of 53.4 seconds). Figure A-17 highlights the points of interest in the ANOVA session window:

F G H I J K L M N O P Q

Figure A-17: Highlighted Tablet Example—ANOVA Output in MINITAB Session Window

R S T U V W X Y Z

Examples If the questions of interest were to study and compare the variances among multiple items (or the standard deviation), then the ANOVA analysis would follow the same procedure but substitute the test data means (seconds) with variances (s2). The manual procedure would follow the same steps. The automated procedure also would follow the same steps, except utilizing a different MINITAB ANOVA test called “Test for Equal Variances.” The procedure steps for the same tablet example would include Step 1.

State the Null (H0) and Alternative (Ha) hypotheses. a. Null hypothesis: H0: the variances (s2) of all three factors are equal i. s2A = s2B = s2C; where s2 is the variance

Analysis of Variance (ANOVA)—7M Tool

157

b. Alternative hypothesis: Ha: at least two factors have different variances Step 2.

Choose the significance level or risk level (α); α = 0.5.

Step 3.

Randomly select the parts or processes.

Step 4.

Identify the factors (that is, parts, processes, and technicians).

Step 5.

Collect the data and record in MINITAB Worksheet. For this test, the data is best represented in a stacked fashion, which lists the trial times (called responses) in sequence for each factor (that is, type of tablet), and in a second column lists the corresponding factor alongside the response time, as illustrated in Figure A-18.

Step 6.

B C D E F G H

Calculate the F-statistic (or F-calculated) using the sample data.

I J

a. Select the following commands: Stat > ANOVA > Test for Equal Variances. b. Put the response time data into the “Responses” field and put the corresponding Tablet Type in the factors field. Select acceptable confidence level in the ANOVA Dialog box, as illustrated in Figure A-19.

A

K L Figure A-18: Stacked Tablet Data Recorded in MINITAB Worksheet

M N O P Q R S T U V W X Y Z

Figure A-19: Tablet Example in MINITAB ANOVA Dialog Box

c. Select OK, and the following output will be displayed in the session window, as illustrated in Figure A-20.

158

Encyclopedia

A B C D E F G H I J K L M N O P Q R S T

Figure A-20: Tablet Example—ANOVA Output in MINITAB Session Window and Graph

The graph in Figure A-20 provides the essential output information. Notice that the confidence intervals overlap, showing that it is difficult to distinguish one range from another, thereby supporting the Null hypothesis that the variances in dissolution time for the three different types of tablet are the same. The session window information on the left of Figure A-20 supports the graph. The p-value for both the Bartlett’s and Levene’s tests are greater than 0.05 (the selected significance level); hence, there is insufficient evidence to reject the Null hypothesis. Therefore, the variability in the time to dissolve the different types of tablet is the same. The Figure A-20 session window output mentions two different tests, Bartlett and Levene. The Bartlett test applies to normal distributions, while the Levene test applies to skewed (non-normal) distributions. This ANOVA test shows that although at least two of the tablet means are different with a 95% certainty, they come from populations and have the same variance. The MINITAB ANOVA Help states, “The effect of unequal variances upon inferences depends in part upon whether your model includes fixed or random effects, disparities in sample sizes, and the choice of multiple comparison procedure.”

U V W X Y Z

Hints and Tips • Understand how to establish and analyze a statistical hypothesis test (See Also “Hypothesis Testing”, Part II, p. 335 for more details). • Use statistical application software to help with the calculations. • If an “interaction” effect is suspected to be present, a replicate of the experiment should be run to estimate its effect. [Interaction is defined as when the response of one variable is dependent on another. Hence, an Interaction Plot would depict perpendicular lines for two factors where the output of one (input) factor depends on the level of a second (input) factor.]

Arrow Diagram

159

Supporting or Linked Tools Supporting tools that might provide input when developing an ANOVA study include • Data Gathering Plan (See Also “Data Collection Matrix,” p. 248) • Cause-and-Effect diagram or Fishbone (See Also “Cause-and-Effect

Diagram,” p. 173)

A B

• Process map (See Also “Process Map,” p. 522

C

• Hypothesis Testing (See Also “Hypothesis Testing,” p. 335)

D

A completed ANOVA provides input to tools such as the following:

E

• Potential solution brainstorming (See Also “Brainstorming Tech-

nique,” p. 168)

F G H

• Solution SelectionMatrix(SeeAlso“Solution Selection Matrix,”p. 672)

I

• QFD (See Also “Quality Function Deployment (QFD),” p. 543)

J

• FMEA (See Also “FMEA,” p. 287)

K

• Control plan (See Also “Matrix Diagrams,” p. 399 for a brief discus-

sion on control plans.) Figure A-21 illustrates the link between ANOVA and its related tools and techniques. Brainstorm Data Gathering Plan Solution Selection Matrix Cause-Effect Diagram ANOVA

QFD

L M N O P Q R S T U V

Process Map FMEA Hypothesis

W X Y

Control Plan

Figure A-21: ANOVA Tool Linkage

Arrow Diagram See Also “Activity Network Diagram (AND),” p. 127.

Z

160

Encyclopedia

B Benchmarking A B C

What Question(s) Does the Tool or Technique Answer? What are the experts doing, and how do they do it?

D

Benchmarking helps you to

E F G H I J

• Identify “best practices” in the marketplace for both what organiza-

tions are doing and how they are doing it. • Innovate, think creatively, about solving a problem or doing some-

thing new to establish a competitive advantage. • Focus people on continuous improvement.

K L M N O P

When Best to Use the Tool or Technique Benchmarking is best used when new ideas are needed to improve or solve a situation. It can be conducted formally or informally; however, having an organizational impact should involve the company’s leaders and/or decision-makers. Benchmarking results that have been selected to act on should be integrated into the strategic direction and goals of an organization. That takes the commitment of the organization’s leadership team.

Q R S T U V W X Y Z

Brief Description Benchmarking compares your organization’s processes and practices with an external organization. Often the best marketplace “expert” may not be part of your industry. Benchmarking another industry and figuring out how to apply its best practices within your organization creates a competitive advantage for your enterprise. Even if the Benchmarking visit is an informal event, it can spawn creativity and innovation. Regardless if the initiative is formal or informal, the technique requires some structure to yield the best benefits. Xerox Corporation in the mid- to late-1970s was one of the first companies to document and popularize Benchmarking under the leadership of David Kearns, then CEO and President. Xerox chartered Sy Zivan to improve its parts distribution process. In turn, Sy and his team (including Bob Camp) stepped outside the office products industry and sought to benchmark retailer L.L. Bean. L.L. Bean was leading the way and enjoying great success with e-commerce as measured by customer satisfaction and overall financial growth. [See S. Zivan’s article, “Benchmarking— Avoid Arrogance and Lethargy,” in Part III, p. 789, of this book.]

Benchmarking

How to Use the Tool or Technique This technique can be done individually or in a small group but requires the support and active commitment from the organization’s leadership team. Regardless if the Benchmarking activity is formal or informal, the technique requires some structure. Successful Benchmarking often is conducted as a project in and of itself and involves these steps at a high level: Step 1.

Plan the benchmark activity. All successful projects require good planning.

A B C

a. Identify the key topic of interest (what).

D

b. List organizations thought to be “experts” (learn from the best practices) or “not-experts” (learn from their mistakes) (why).

E

Experience shows that business people tend to skimp on the appropriate time it takes to research the best organizations to benchmark. Some clients assign outside reading to the Benchmarking and/or strategy team, only to reconvene and do book reports on key points learned. Other clients hire market research and Benchmarking firms for advice. Thinking outside of the box often reveals industry leaders that otherwise would be overlooked. The most successful mental model often entails unrelated industries at first glance. c. Define the data collection plan. i. Who will collect the data, and who is the benchmark firm’s key contact? ii. Where and how will the data collection take place? Will it take place at the firm’s location, or will it be by phone, written survey, or a combination thereof? iii. When will this data collection take place, and what is the target schedule? Make certain to have a defined beginning and end and avoid scope creep and schedule slippage. iv. How will this Benchmarking occur? What is the budget, and what resources are needed? Is the appropriate level of organization involved? Are the resources properly trained in what to do and how? Step 2.

161

Design the data collection approach. This step answers the questions, “What specific data will be collected?” and “How will the data be collected and communicated?”

F G H I J K L M N O P Q R S T U V W X Y Z

162

Encyclopedia

a. Who is the target audience (consumer) of the benchmark data? b. What is purpose—the gap or opportunity to be addressed? c. How will the data be collected? Develop an interview guide to collect the data.

A B

i. Identify the process (or practice) and performance metrics to be studied. What does success look like? Anticipate how success might be measured by the benchmark organizations, compared with how your organization does today. Note: Make certain that process and cultural dimensions are studied; avoid limiting it to critical parameters that are easy to measure.

C D E F G H

ii. Develop the key questions to be asked (limit it to a small set of questions, that is, no more than about 10).

I J

iii. Test the questions with the Benchmarking Stakeholders.

K

iv. Determine the format or combination of approaches to best collect the desired data.

L M

If possible, allow for time to observe the benchmark firm. Often seeing the benchmark practice in operation triggers additional questions, helps put the success in context, and identifies the implicit elements to success.

N O P Q

v. Develop the final document.

R S

Step 3.

T U V

Implement the plan by conducting the Benchmarking activities, gathering the data, and analyzing it. Analysis tools (which are detailed in their own entry within the encyclopedia section of the book) include

W

a. Affinity diagrams

X

b. KJ Analysis

Y

c. Cause-and-Effect diagrams

Z

d. QFD Step 4.

Summarize and recommend a Go Forward strategy. Determine the key messages, the lessons learned from the Benchmarking initiative, and then develop and document a Go Forward plan. The documentation should include the goals, a recommended action plan to achieve these goals, and a control plan to monitor performance.

Benchmarking

163

Hints and Tips • Confirm proper sponsorship within the context of the organization’s strategic initiative. Benchmarking produces creative ideas for change and improvement. To ensure that the changes will be supported (properly funded and resourced), the leadership needs to buy into the Go Forward strategy and lead its implementation, thereby making it part of the organization’s strategy and goals. • Take time to research and develop a list of potential Benchmarking candidates. Research who is an industry leader outside of your own industry. • If possible, conduct Benchmarking at the benchmarked organization’s location to observe the work in process. • Encompass the full context of the benchmark’s success beyond the simple metrics and include process and cultural elements that are probably main contributors and critical parameters. • Offer reciprocal Benchmarking opportunities to establish a winwin scenario for the benchmarked organizations.

A B C D E F G H I J K L M N O P

Supporting or Linked Tools Supporting tools that might provide input when developing Benchmarking include • VOC Data Gathering Tools and Techniques, such as surveys and

interviews (See Also, “Voice of Customer Gathering Techniques,” p. 737)

Q R S T U V

• Written reports

W

• Brainstorming sessions (See Also “Brainstorming Techniques,”

X

p. 168) A completed benchmark imitative provides input to tools such as • Affinity diagram (See Also “Affinity diagrams,” p. 136) • KJ Analysis (See Also “KJ Analysis,” p. 375) • Cause-and-Effect diagram or Fishbone (See Also “Cause-and-Effect

Diagram,” p. 173)

Y Z

164

Encyclopedia • Conjoint Analysis (See Also “Conjoint Analysis,” p. 207) • QFD (See Also “Quality Function Deployment (QFD),” p. 543) • Implementation (See Also “Matrix Diagrams,” p. 399 for a brief dis-

cussion on transition and launch plans) A B C D

• Control plan (See Also “Matrix Diagrams,” p. 399 for a brief discus-

sion on control plans) Figure B-1 illustrates the link between Benchmarking and its related tools and techniques.

E F

Affinity Diagrams

G H I J

KJ Analysis VOC Data Gathering

K L

Written Reports

Cause-Effect Diagram Benchmarking

M N O

QFD Brainstorming

P

Implementation Plan

Q R

Control Plan

S T

Figure B-1: Benchmarking Tool Linkage

U V W

Boxplots—Graphical Tool

X Y Z

What Question(s) Does the Tool or Technique Answer? What does the distribution of a set of data look like? Alternatively, what do the distributions look like, and how do they compare with multiple sample groups? Boxplots help you to • Visualize what the distribution looks like for a sample set of data.

Boxplots—Graphical Tool

165

• Display the data distribution over its range (the spread) and indicate its quartiles and median. • Graphically summarize and compare frequency distribution infor-

mation about one or more sets of data. A

Alternative Names and Variations This tool is also known as • Box-and-whisker plot

B C D E

Variations on the Tool • Dotplot—displaying the data detail versus a summary, therefore used for a small sample size of about 15. (See Also “Dotplot,” p. 280)

F G H I J

When Best to Use the Tool or Technique The old adage holds true, “A picture is worth a thousand words.” The graphical method is the picture that visually displays the distribution of a set of data or that compares multiple sets of data. A Boxplot is best used when the individual data points are not of interest. It specifically is a good summary graphical tool when the sample size is too small to build a meaningful histogram (less than 50 pieces of data). A Boxplot works well when a Dotplot is too busy with over 15 data points.

K L M N O P Q

Brief Description The Boxplot is a simple graphical tool containing a great deal of information about a continuous, quantitative data set’s distribution, its range (indicating the minimum and maximum), its median, and its quartiles. This tool was developed by John W. Tukey to display five summary data characteristics: 1. the data median

R S T U V W X

2. the upper quartile of data

Y

3. the lower quartile of data

Z

4. the minimum data point 5. the maximum data point Figure B-2 illustrates the basic shape and features of a Boxplot graph.

166

A

Encyclopedia

Notice how the diagram divides the data into quarters (25% segments), where the first quartile (Q1) demarcation is the bottom of the box, the second quartile (identifying 50% of the data on either side) is marked by the median line dissecting the box, and the third quartile demarcation is the top of the box. The lines on either end of the box, referred to as “whiskers,” depict data on either end of the range within a limit that are calculated as

B

• Lower Limit = Q1–1.5 (Q3–Q1)

C

• Upper Limit = Q3 + 1.5 (Q3–Q1)

D E F G H I

If data exceeds the whisker limits, an asterisks (*) represents the unusual observation as an “outlier” data point. Outliers are defined as data beyond 1.5 times the inter-quartile distance from each quartile. In Figure B-2, the outlier indicates an unusually large value.

* 25%

“Whisker” 25% Median

Minimum

25% 25%

Figure B-2: Basic Boxplot Illustration

J K

Maximum (“Outlier”)

N

How to Use the Tool or Technique This graphical technique can be used on a single data set or to compare multiple data sets. A Boxplot can be created by hand or by using statistical software. The procedural steps and illustrations are based on the MINITAB software for a simple Boxplot for one variable.

O

Step 1.

Collect the data and enter the variable in a single column in the MINITAB Worksheet.

Step 2.

From the tool bar, select Graph > Boxplot….

Step 3.

Select One Y Simple; click the OK button as illustrated in the left portion of Figure B-3.

Step 4.

Enter the column of data in the Graph variables window of the dialog box, and click the OK button, as illustrated in the left portion of Figure B-3.

L M

P Q R S T U V W X

4a

3a

Y Z

3b

Figure B-3: MINITAB Boxplot Dialog Windows

4b

Boxplots—Graphical Tool

167

The output will resemble the illustration in Figure B-2. To examine two groups of data, follow the same procedure, but in the first dialog box, select the appropriate graph type depending on how many Y variables there are. Figure B-4 depicts a two-group “simple” Boxplot example with multiple “Ys” comparing “Yield2” and “Yield3” variables. A B C D E F G H I J K L Figure B-4: Simple Boxplot with Multiple Ys Illustration

M N O

Supporting or Linked Tools Supporting tools that might provide input when creating a Boxplot graph include • Data gathering plan to collect the appropriate metrics (See Also

“Data Collection Matrix,” p. 248) • Performance charts and dashboards

A Boxplot graph can provide input to tools such as • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram,”

p. 173) • QFD (See Also “Quality Function Deployment (QFD),” p. 543)

P Q R S T U V W X Y Z

168

Encyclopedia • Statistical analy-

A B C D E F G H I J

sis tools and other root cause analysis techniques (See Also “Statistical Tools,” p. 684) • FMEA with fol-

low-on action planning (See Also “FMEA,” p. 287)

Cause-Effect Diagram Data Gathering (metrics)

QFD Boxplot Graph

Performance Charts and Dashboards

Statistical Analysis Tools

FMEA

Figure B-5: Boxplot Graph Tool Linkage

Figure B-5 illustrates the link between a Boxplot graph and its related tools and techniques.

Brainstorming Technique

K L M N O P Q R

What Question(s) Does the Tool or Technique Answer? How can you elicit new ideas in a short period of time? Brainstorming helps you to • Generate and document a large volume of ideas or concepts in one

meeting. • Create out-of-the-box thinking to develop new ideas.

S T U V W X Y Z

Alternative Names and Variations This tool is also known as • Round-robin brainstorming

Techniques that incorporate brainstorming include • Nominal group technique • Mind-mapping • Affinity diagramming (See Also “Affinity Diagram,” p. 136) • Fishbone diagramming (See Also “Cause-and-Effect Diagram,”

p. 316) • Tree diagram (See Also “Tree Diagram,” p. 712)

Brainstorming Te chnique

169

• Matrix diagram (See Also “Matrix Diagrams,” p. 399) • QFD (See Also “Quality Function Deployment,” p. 543) • FMEA (See Also “FMEA,” p. 287) A

When Best to Use the Tool or Technique When breakthrough ideas are needed, to expand the team’s thinking beyond the traditional concepts, the brainstorming technique can help generate new ideas or perspectives. Use brainstorming when each team member’s participation is desired.

B C D E F

Brief Description Brainstorming features a team of people generating ideas in an intentionally uninhibited fashion. An ideal group size consists of about eight people (from six to ten is acceptable). The best results are produced by a diversity of perspectives, such as cross-functional disciplines, different perspectives, tenure, roles, and so on. Often a mix of written (or silent) and verbal generation techniques help to draw people out and allow for 100% participation. Brainstorming often is accomplished in a single meeting that produces as many ideas as possible. The purpose of brainstorming is to generate several ideas, not necessarily to solve a problem. At the conclusion of the idea generation process, techniques such as multi-voting can help to prioritize or select the top “candidates” for appropriate actions.

G H I J K L M N O P Q R

How to Use the Tool or Technique This technique should be used in a small group and should follow these general guidelines: Step 1.

Step 2.

S T U

Prepare for the meeting.

V

Identify the topic of interest and assemble a small group that represents a cross-section of perspectives. Invite a “neutral” person to serve as the meeting Recorder and facilitator, who will usually not be participating in the process itself. Choose the documentation vehicle (list on paper or individual ideas per sticky note or index card).

W

Initiate the meeting. Review the topic of interest and discuss the brainstorming approach and guidelines to be used in the meeting.

X Y Z

170

Encyclopedia

Step 3.

Individual idea generation. Ask each member to write down as many ideas as possible. Silence should be observed while people are writing down their ideas and until everyone has finished. There are two approaches for documentation—people can either

A

a. List their ideas on pieces of paper, which will be shared later verbally

B C

b. Take a stack of sticky notes or index cards and write one idea per note or card

D E F

Step 4.

Ideas are shared one at a time until all individual ideas are shared with the group. Best results tend to occur when one person shares only one idea at a time and then defers to the next person. This cycle is repeated until all ideas are shared.

G H I J

a. If a list of paper was used in a round-robin fashion, have each person share one idea at a time and then move on to the next person. The meeting Recorder documents the ideas as a list on a flip chart.

K L M

b. If the sticky notes or index cards are used, again in a round-robin fashion, each person shares one idea at a time and then posts the one idea on the wall or flip chart paper and then gives the floor to the next person. (Post index cards with masking tape.)

N O P Q

This approach is preferred if the facilitator’s ideas are important to capture as part of the group.

R S T U V W X Y Z

Share ideas with the group.

Step 5.

Review and build on ideas. Ask the team to reflect on the shared ideas and explore whether an idea spawns an additional idea, expands one, is combined with others, or needs to be modified. Ask the team to reflect on situations outside of the current market segment or change perspectives (such as age, country, and so on) to see if new ideas evolve. Continue to generate additional ideas and record each one as it is shared using the selected documentation approach.

How to Analyze and Apply the Tool’s Output After the brainstorming session is completed, the team can decide how best to use the ideas. The list can be prioritized using multi-voting or a nominal group technique. The ideas can serve as input to an Affinity diagram, Cause-and-Effect diagram (or Fishbone), Mind map, or FMEA.

Brainstorming Te chnique s

171

Hints and Tips •

Select a comfortable environment for the meeting place. Often a casual place away from the workplace (perhaps outdoors) elicits more creativity.

•

All ideas are recorded.

•

Include everyone and avoid inhibiting anyone. Encourage all to participate by utilizing written and verbal idea generation.

•

Allow each person to generate ideas without discussing, evaluating, analyzing, or criticizing any concept at the time of generation/development.

•

Build on one another’s ideas. Sometimes a courteous “freewheeling” verbal approach is used.

•

Expanding on someone’s idea or piggybacking should be encouraged.

•

Initially strive for a large quantity of ideas, regardless of how farfetched they may seem. The process later clarifies, consolidates, and prioritizes the concepts.

A B C D E F G H I J K L M N O

Supporting or Linked Tools Supporting tools that might provide input when conducting a brainstorming session include • Data Gathered (metrics), performance charts, dashboards (See Also

“Data Collection Matrix,” p. 248) • Documentation or written reports

A completed brainstorming session provides input to tools such as • Affinity diagram (See Also “Affinity Diagram—7M Tool,” p. 136) • Cause-and-Effect (or Fishbone) diagram (See Also “Cause-and-

Effect Diagram—7QC Tool,” p. 173) • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) • Matrix diagram (See Also “Matrix Diagrams—7M Tool,” p. 399) • Mind-mapping

P Q R S T U V W X Y Z

172

Encyclopedia • QFD (See Also “Quality Function Deployment (QFD),” p. 543) • Tree diagram (See Also “Tree Diagram—7M Tool,” p. 712)

Figure B-6 illustrates the link between a brainstorming session and its related tools and techniques. A B

Affinity Diagram

C D

Mind-mapping

E F G

Data Gathering (metrics)

H I

Cause-Effect Diagram Brainstorming

Documentation Tree Diagram

J K

QFD and other matrix diagrams

L M N

FMEA

O P Q R S T U V W X Y Z

Figure B-6: Brainstorming Tool Linkage

Cause-and-Effe ct Diagram—7QC Tool

173

C Capability Analysis See Also “Process Capability Analysis,” p. 486.

Cause-and-Effect Diagram—7QC Tool

A B C D

What Question(s) Does the Tool or Technique Answer? What are the potential causes of a problem or problematic outcome? Cause-and-Effect diagramming helps you to

E F G H

• Organize ideas and understand the relationship between potential causes and a problem by formatting, arranging, organizing, and parsing potential causes into themes and sub-themes in preparation for a future cause identification effort. • Stimulate thinking when developing the list of the potential sources of a problem. • Guide concrete action and track the potential causes during an investigation effort to determine if the item significantly contributes to the problem.

I J K L M N O P Q

Alternative Names and Variations This tool is also known as:

R S T

• C and E diagram

U

• Fishbone or fishbone diagram

V

• Ishikawa diagram

W

Variations on the tool include • Cause-and-Effect matrix or Cause-and-Effect Prioritization Matrix (See Also “Cause-and-Effect Prioritization Matrix,” p. 188) • Cause enumeration diagram • Process Fishbone • Time-delay Fishbone • CEDAC (Cause-and-Effect Diagram and Cards)

X Y Z

174

Encyclopedia

• Desired-results Fishbone • Reverse Fishbone diagram • Cause and Prevention diagram A

• Fault Tree analysis (See Also “Fault Tree Analysis,” p. 309)

B C D E F G

When Best to Use the Tool or Technique Before any action is taken, this tool helps to organize and begins to analyze potential sources of problems. The Fishbone is a good tool to use when multiple people (and/or disciplines) should be engaged in the problem-solving and there are many perspectives to capture.

H I J K L M N O P Q R S T U V W X Y Z

Brief Description The Cause-and-Effect diagram is one of the seven basic tools of quality. It goes by several names—a Fishbone diagram because a completed diagram resembles the skeleton of a fish and an Ishikawa diagram after its creator. A professor at the University of Tokyo, Dr. Kaoru Ishikawa, developed the first Cause-and-Effect diagram in 1943. The original intent of the diagram was to sort out and depict the relationship among the several factors impacting Quality Control (QC), wherein the variables that cause dispersion, such as chemical composition, size of parts, or process workers, are called “factors” (causes). The quality characteristics describe the outcome, or effect, such as length, hardness, percentage of defects, and so on. As the tool grew in popularity, it was referred to as an Ishikawa diagram. The Cause-and-Effect diagram is a member of the 7QC Tools (or seven Quality Control tools), attributed to Dr. Ishikawa. The 7QC Tools sometimes are called the seven basic tools because they were the first set of tools identified as the core quality improvement tools. Ishikawa’s original 7QC Toolset includes 1) Cause-and-Effect diagram, 2) Check sheet (or Checklist), 3) Control charts, 4) Histogram, 5) Pareto chart, 6) Scatter diagram, and 7) Stratification. More recently, the 7QC Toolset is modified by substituting the Stratification technique with either a flowchart (or process map) or a run chart (or time series plot). A Fishbone diagram is a focusing tool. It starts with identifying all the ideas for potential causes and then groups and categorizes the potential causes into themes. It takes a snapshot to hypothesize the collective potential causes—what the team thinks is currently happening. It displays potential causes of a problem, with the potential causes depicted as “off-shoots” or bones of a fish stemming from the problematic outcome as the “head” of the fish. Increasing detail about the potential

Cause-and-Effe ct Diagram—7QC Tool

175

cause is displayed as the offshoot, or bone, branches further out from the main bone. The relationships can be depicted as a matrix or Fishbone diagram, wherein the problem (or effect) is placed to the far right. The problem statement is placed in a box or diamond-shape to represent the head of a fish. This represents the “effect” that all the subsequent causes supposedly impact. The root cause themes and sub-themes are placed to the left of the problem. There are two major types of Cause-and-Effect formats depending on the context in which the problem exists; one is a Dispersion, and the other is a Process format. There are some memory techniques that serve as checklists to trigger thinking about potential root cause categories—the 5Ms and P and the 4Ps. The former is typically applied to the Dispersion analysis, and the latter is applied to the Process approach.

Dispersion Analysis Potentially, why does quality dispersion occur? This approach uses individual causes (or dispersion) grouped within a major cause category. The smaller categories on the skeleton of the fish drill down into sub-causes, identifying why this cause potentially happens. Dispersion occurs typically due to differences in

A B C D E F G H I J K L M

• Raw materials’ composition, size, and so on

N

• Tools, machinery, equipment, technology—with respect to their operating performance

O

• Work method, process or procedure—potentially because it could be incomplete, inaccurate, inflexible (accommodating change or a special occurrence), misunderstood

Q

• Measurement—potentially from operator-to-operator error, operator-to-part, or part-to-part.

T

When using the Dispersion analysis approach, there is no right answer or set of categories names or themes. They should be modified to fit the situation or problem. However, a convention that often is used as a starting point is referred to by the acronym, 5Ms and P.

5Ms and P Memory Triggers A standard categorization technique for manufacturing industries is called 5Ms and P [pronounced “five Ms and a P”] and can be used as a checklist to identify common cause themes: • Machines (equipment, technology—hardware and/or software) • Methods (process, procedure, approach, policy, or practice) • Materials (raw materials, components, information, or data)

P R S U V W X Y Z

176

Encyclopedia

• Measurements (input and output metrics for quality, quantity, process performance metrics, calibration, and inspection) • Mother Nature (or a more encompassing term is environment), which includes external as well as workplace factors such as: A B

• Natural environment [weather (and acts of God), temperature, humidity]

C

• Physical surroundings (facilities, buildings, plant, workspace, office)

D

• Management surroundings (organizational, social, political)

E

• Marketplace

F G H I J K L M N O P Q R S T U V W X Y Z

• People [those involved in the process (directly or indirectly)—customers, employees, managers, suppliers, partners, regulators and shareholders] Sometimes this approach is called The 6Ms, where the term “People” (of 5Ms and P) is changed to Manpower.

Process Classification The Process approach uses the major process steps, instead of major causal categories; however, this is rarely used. This diagram may be depicted in the Fishbone shape or as a Process map “spine” with potential causes linked to the appropriate process step. When using the Process approach, again, there is no right answer or set of category names or themes. They should be modified to fit the situation or problem. However, the services industry has a convention that often is used as a starting point—the 4Ps.

4Ps Memory Triggers Common categories that work well for the service(s) industry is called the 4Ps, which represents • Policies (company and HR policies, including roles and responsibilities, performance metrics, reward and recognition, promotion, overtime, comp-time/vacation, and so on, and organizational—cultural, training versus apprenticeship, and so on) • Procedures (methodology, approach) • People (employee profile, partners, sub-contractors, and so on) • Plant/Technology (workspace, hardware, software, support tools) Because the services industry is dependent on its human capital, the policies are critical to shaping its “personality” in the marketplace, and often it is the make-or-break for attracting the right talent. Hence, the Policy category often needs to be parsed further to get to the appropriate detail.

Cause-and-Effe ct Diagram—7QC Tool

177

How to Use the Tool or Technique When developing a Cause-Effect, Fishbone, or Ishikawa diagram: Step 1.

Agree on the topic or problem to be analyzed—for example, a quality characteristic that needs improvement. Use data specifically to describe the problem.

Step 2.

Write the specific problematic outcome to the far right edge of the diagram and draw a box around the text to form the “head” of the fish. Figure C-1 displays the general structure of the Cause-andEffect diagram using the Fishbone analogy.

A B C D E F G H I

Cause

Major Cause

Major Cause

J K

Major Cause

L Effect

Bones

Problem

Spine

M N O P

Major Cause

Major Cause

Major Cause

Q R S T U V

Figure C-1: Basic Cause-and-Effect Diagram (Fishbone) Using 5Ms and P for Dispersion Analysis

Step 3.

Draw a line extending from the left edge of the fish head to the left edge of the diagram to represent the spine.

Step 4.

Referencing data and experience, write the potential causes of the problem and group related topics. Have the participants bring to the meeting completed data collection sheets from their respective areas to inform the brainstorming activities. (See Also “Brainstorming Technique,” p. 168)

W X Y Z

178

A B

Encyclopedia

Step 5.

Name the major themes of each grouping (or high-level process steps).

Step 6.

Draw a line extending from the spine for each theme and write the category name at the outermost end of the line (the end that is not attached to the spine) to represent a “main” Fishbone.

Step 7.

Draw one branch or offshoot to the main Fishbone for each sub-theme (or sub-process step). Label the end of each branch with the sub-theme name to represent smaller bones and to show the linkage to the “higher level” category with which it is affiliated.

Step 8.

Continue Step 7 to drill-down into as much detail as required, defining more bones and linking the relationship.

Step 9.

Review the diagram for completeness and clarity and then modify accordingly.

C D E F G H I J K

a. Use the 5-Why technique to ensure the potential cause is documented (See Also “5-Whys,” p. 305).

L

b. Fill in gaps and streamline wording.

M

c. Eliminate unnecessary redundancy but document duplicity that represents the current situation.

N O

d. Eliminate unrelated items.

P Q R S T

How to Analyze and Apply the Tool’s Output • Review the final diagram and discuss the themes thought to represent the most critical source of the problem.

U

• Look for repeat causes across multiple categories.

V

• Prioritize and select—use the diagram as a checklist to explore further into those critical themes. You could use a QFD tool to prioritize the potential causals. A less desirable selection approach could be a form of voting technique, such as weighted voting.

W X Y Z

• Investigate or study the prioritized causes and/or critical themes to verify that they are in fact causals. Some tools and techniques to begin the analysis include the following: • Force-field analysis • Scatter Plots (See Also “Scatter Diagram—7QC Tool,” p. 640) • Frequency Plots (See Also “Graphical Methods” and “Histogram —7QC Tool,” p. 323 and 330, respectively)

Cause-and-Effe ct Diagram—7QC Tool

179

• Tables of Results (for attribute data) • ANOVA (Analysis of Variance) (See Also “Analysis of Variance (ANOVO)—7M Tool,” p. 142) • Regression Analysis (See Also “Regression Analysis,” pp. 571) • Time Trap Analysis • Design of Experiment (DOE) (See Also “Design of Experiment (DOE),” p. 250)

A B C D

Note This tool documents POTENTIAL causes, not proven causes; so improvement action plans cannot be developed until the causes are verified. Seek the causes thoroughly using the Cause-and-Effect diagram. When the Cause-and-Effect is detected, check and record it on the diagram, by writing the date of occurrence and its measurement next to its corresponding “labeled bone.” If the cause is not detected, then the “labeled bone” will lack any markings. This procedure highlights experienced data and focuses on the detected causes, which can lead to quicker error correction.

E F G H I J K L M N O P

Examples Dispersion Analysis example An example of a Dispersion Analysis, as shown in Figure C-2, uses the “5Ms and P” technique to define the root causes as titles of the major bones off the spine of the Cause-and-Effect diagram. [Recall the 5Ms and P refer to Machine, Materials, Method, Metrics, Mother Nature, and People. See the “Hints and Tips” section later in this entry for more detail on 5Ms and P technique.] Process Classification Analysis example An example of a Process Classification Analysis, shown in Figure C-3, uses the high-level process steps as the titles of the major bones off the spine of the Cause-and-Effect diagram.

Q R S T U V W X Y Z

180

Encyclopedia

Cause

A B C D E

Mother Nature Ingredients Old

Interruptions

Baker Distracted

Measuring spoons Dry goods scale not calibrated

Baker Rushed

Measuring cup dented

Electric Mixer misses bottom of bowl

Wrong Oven Temperature Used

Measurement

Baked cake failed to rise

Failed to scrap bowl edges Ingredients not mixed well Ingredients mixed all at once

No supplement oven thermometer

Kids Jumping During Baking

Person

Oven temperature not calibrated

Cook book unclear

Wooden Spoon used versus Spatula

Urgent phone call Child Injured

Effect

Machines Baking pans burn easily

Wrong

High Altitude

F G

Materials

Too little liquid Used metal scoop rather than glass cup

Method

H I J K L

Figure C-2: Cause-and-Effect Diagram (Fishbone) Using 5Ms and P for Dispersion Analysis

M N O P Q R S T U V W X Y

Read Recipe and Prepare High Altitude Oven temperature not adjusted and calibrated

Gather Ingredients Cookbook not understood

Put cake in cold oven Cooked at wrong temperature

Bake Cake

Necessary ingredients not bought

Too little liquid Ingredients not organized Test for “Doneness” not done

Effect

Baked cake failed to rise

Baker Interrupted Child Injured Urgent phone call Kids Jumping During Baking

Cake iced while still hot

Cool and Ice Cake

Cause Baking powder not added Ingredients mixed all at once Ingredients not mixed well Scrap bowl edges with rubber spatula Electronic Mixer misses bottom of bowl

Oven not warmed Ingredients measured incorrectly Used metal scoop rather than glass cup

Missing Old

Equipment not checked Baking pans burn easily

Mix Ingredients

Other

Z

Figure C-3: Cause-and-Effect Diagram (Fishbone)—Using Process Steps for a Process Classification Analysis

Cause-and-Effe ct Diagram—7QC Tool

181

Process Classification Analysis—a Second Example Another example of a Process Classification Analysis, shown in Figure C-4, diagrams the high-level process steps as the actual spine of the Cause-and-Effect diagram. The bones off the spine align to the root causes directly linked to a particular process step. A

Necessary ingredients not bought

B

Ingredients Missing

Ingredients not organized

Old

Cookbook not understood

Oven temperature not adjusted and calibrated High Altitude Measuring cup dented Baking pans burn easily Equipment not checked

Baking powder not added

Oven not warmed

Read Recipe and Prepare

Gather Ingredients Too little liquid

Used metal scoop rather than glass cup Ingredients measured incorrectly

Cooked at wrong temperature

Baked Cake “Soupy”

Mix Ingredients

C

Test for “Doneness” not done

Put cake in cold oven

D Cool and Ice Cake

Bake Cake

Baked cake failed to rise

Ingredients mixed all at once Electric Mixer misses bottom of bowl

Kids Jumping During Baking

Cake iced while still hot

Scrap bowl edges with rubber spatula

Other

Ingredients not mixed well Urgent phone call Child Injured Baker Interrupted

Figure C-4: Cause-and-Effect Diagram (Process Fishbone)—Using Process Steps for a Process Classification Analysis

E F G H I J K L M N

Poorly Constructed C and E Diagram Example A cause generally contains many complex elements. Therefore when building a Cause-and-Effect diagram, it is important to drill down into a cause to identify the potential root. The drill down can be achieved by using the 5-Whys technique. (See “5-Whys,” p. 305) The following Fishbone diagram is an example of a poor one. Even though the “form” is correct, the diagram simply lists eight items. It lacks detail, which indicates potential lack of knowledge of the process. The Fishbone diagram should be detailed, not too generalized or shallow. The example of a poor Fishbone fails to describe why ingredients are missing and why the batter was not mixed well. The baker could have had the blender working overtime, but if the beaters were not touching all parts of the bowl and the mixing process lacked manually scraping of the sides of the bowl with a rubber spatula, the potential root causes behind why the batter was not mixed well are more obvious. Moreover, the poor Fishbone diagram contains some symptoms—”Baked Cake Soupy,” rather than probing further to understand the driver as to why it’s soupy and document the causes. A well-constructed Fishbone diagram documents all the possible causes in an organized way, depicting the relationship of the major causes themes to subsequent or more detailed causes. Figure C-5 displays a poorly constructed Cause-and-Effect diagram.

O P Q R S T U V W X Y Z

182

Encyclopedia Oven Not Pre-heated

Missing Ingredients

Old Baking Powder

Didn’t Follow Recipe

Baked cake failed to rise

A B C

Batter Not Well Mixed

Baked Cake “Soupy”

Baker Interrupted

Kids Jumping During Baking

Figure C-5: Poor Cause-and-Effect Diagram (Fishbone)

D E F G H I J K L M N O P Q R S T U V W X Y Z

Hints and Tips Building a Cause-and-Effect diagram is best done with a group of people with diverse perspectives, to analyze the potential causes from multiple angles. Capture viewpoints from various roles and disciplines. Ensure everyone agrees on the problem statement first and focus on potential causes, not symptoms. Use the 5-Why technique to drill down and identify the potential source of the cause. (See Also “5-Whys,” p. 305) Capture all ideas (information on “what,” “where,” “when,” “how much…”). Consider all causes, not just those within the group’s control. Use data to specify the problem. Do not criticize or debate ideas during brainstorming or when representing them on the Cause-andEffect diagram. Histograms can reveal the frequency of defects caused by certain items. Simply test for understanding to clarify that an idea is communicated succinctly. Look for duplicates and aggregate them together if possible. Some causes may fit into more than one category (particularly people-related topics). Ideally a cause should be aligned with one major category, but if it is not clear where it best fits, represent it in multiple categories. With further analysis, eventually its best fit may narrow to one area. Take time in developing the diagram, ask as many people as possible, and avoid trying to complete it in one work session. Allow people to study a work-in-process diagram to trigger additional ideas. Prioritizing can be done later using a QFD. The appropriate number of categories depends on the problem and related potential causes. There is no prefect number. For a Process

Cause-and-Effe ct Diagram—7QC Tool

183

Fishbone, the process depicted should be high-level. Therefore, ideally the number of steps is three to five and generally should not exceed 10; otherwise, it becomes too detailed and cumbersome. Category naming can be done two different ways: predetermined categories prior to the brainstorming portion or determined after brainstorming as a “fall-out” from the grouping portion. There are pros and cons to both approaches. The predetermined categories may expedite and better organize the work. If experience determines the categories selection, then the theme titles probably will not inhibit the generation of ideas. However, the addition of an “other” category also will help to prevent the stifling of ideas. If category selection occurs after the brainstorming, it takes time to organize and arrive at a common agreement on the theme. It involves testing for understanding and potentially regrouping and/or further consolidation of the ideas. As a rule of thumb, the drill-down into sub-causes usually stops when a cause is controlled by a different level of management than the main causal bone. Common sense should rule. Rarely are people the root cause of process problems. Often people are compensating for something that is not working, confused by something that is misunderstood, or handling an exception to the rule. Probe further to understand the cause behind a person’s given activity or behavior. Sometimes an “outside” person not involved in the development process can add a new perspective and identify gaps or patterns in the diagram.

A B C D E F G H I J K L M N O P Q R S T U

Supporting or Linked Tools Supporting tools that might provide input when developing a Causeand-Effect diagram include • Histogram (See Also “Histogram—7QC Tool,” p. 330) • Pareto chart (See Also “Pareto Charts—7QC Tool,” p. 445) • Process map to identify potential root causes (See Also “Process Map (or Flowchart—7QC Tool),” p. 522) • VOC and VOB data (See Also “Voice of Customer Gathering Techniques,” p. 737)

V W X Y Z

184

Encyclopedia

• 5Ms and P technique or its variants such as 6Ms and 4Ps (described in detail earlier in this section) • 5-Why Technique (See Also “5-Whys,” p. 305) • Brainstorming technique (See Also “Brainstorming Technique,” p. 168) A

A completed Cause-and-Effect diagram provides input to tools such as

B C

• Brainstorming potential solutions and concept generation methods

D

• Root Cause Analysis Techniques [such as Correlation and Regression, Hypothesis Testing, and Design of Experiment (DOE)] (See Also “Design of Experiment (DOE),” “Hypothesis Testing,” and “Regression Analysis,” p. 250, 335, and 571, respectively)

E F G H I J K L

• QFD (Quality Function Deployment) (See Also “Quality Function Deployment (QFD),” p. 543) • FMEA (Failure Modes and Effect Analysis) (See Also Failure Modes and Effects Analysis (FMEA), p. 287) Figure C-6 illustrates the link between the Cause-and-Effect diagram and its related tools and techniques.

M N

5-Whys Technique

O P

5M and P Technique and its variants

Histogram

QFD

Pareto Chart

Root Cause Analysis Techniques

Q R S T U

Cause and Effect Diagram

Concept Generation Methods

Process Map

V W X Y Z

VOC and VOB Data

Brainstorming Technique

FMEA

Figure C-6: Cause-and-Effect Tool Linkage

Variations Cause-and-Effect Matrix Links prioritize customer requirements with the steps of a process to understand which step(s) is more critical to meeting the customer needs. [See the Cause-and-Effect Matrix entry in this section for more detail.]

Cause-and-Effe ct Diagram—7QC Tool

185

Cause Enumeration Diagram This tool first starts with a brainstormed, exhaustive list of all the possible causes and then groups the items into affinities categories—thus enumerating a large number of likely causes. The very last step is drawing the Fishbone diagram or a matrix that considers the Cause-and-Effect relationship. Benefit: The potential causes listed should be more comprehensive.

A

Weakness: It may be difficult to relate the large number of causes to the problematic outcome, making the diagram difficult to draw.

B

Cause and Prevention Diagram This variation is used as part of risk mitigation planning. Figure C-7 illustrates the general structure of a Cause and Prevention diagram. [See the “Cause and Prevention Diagram” entry in this section for more detail.]

D

C E F G H I J K

Prevention – Risk Response Strategies Risk Response

Risk Response

L

Risk Response

M

Bones Cause Potential Risk

Spine

N O P Q

Risk Response

Risk Response

Risk Response

R S T U V W

Figure C-7: Cause and Prevention Diagram (Fishbone variation)

X Y

CEDAC (Cause-and-Effect Diagram and Cards) Dr. Ryuji Fukuda (from Japan) developed this approach, which integrates problem-solving with the Fishbone diagram analysis, uses color cards, and prominently displays it to collect input from passers-by. The CEDAC approach includes the following steps: ®

Step 1.

Problem is identified and an improvement target and metrics are selected and agreed to.

Z

186

Encyclopedia

Step 2.

Causes are brainstormed.

Step 3.

Causes are sorted into categories.

Step 4.

Using colored cards (or Post-It™ Notes); one color is assigned to a category.

Step 5.

The causes identified within a category are written on a card with the assigned color—one idea per card. This step is repeated until all the ideas are documented on a card with category color assigned to it.

E

Step 6.

The cards are placed on a Fishbone diagram.

F

Step 7.

Improvement ideas for each card are written on a different color card and posted on the Fishbone diagram.

Step 8.

The diagram is then placed in a high-traffic public area (similar to the time-delay Fishbone technique), wherein passers-by can write additional ideas on causes and/or improvements.

K

Step 9.

Each improvement idea is studied for potential value contribution.

L

Benefit: Similar to that of the time-delay Fishbone, plus the color cards contribute an additional communication technique that assists with identifying patterns.

A B C D

G H I J

M N O P Q R S T U V

Weakness: Similar to the time-delay Fishbone, plus this technique adds a level of complexity that may or may not add value to the process of identifying and sorting potential causes. Moreover, the technique blends improvement ideas with the original intent of an Ishikawa diagram. Thus, the natural tendency of people to jump to solutions may be more difficult to overcome using this technique. The potential causes identified in a Causeand-Effect diagram should be further studied (using other techniques such as Hypothesis testing) to identify the actual root causes and which have the greatest impact. If the improvement ideas are simple in nature, it will boost worker morale and require little investment (either budgetary or procedurally), and so by all means the suggestion can be implemented.

W X Y Z

Warning Caution should be taken that implementing simple band-aid improvements might merely treat the symptom temporarily and not address the underlying root cause, which can lull people into thinking they have “fixed” the problem because they have “done something.” Keeping “busy” does not necessarily add value; avoid minutia.

Cause-and-Effe ct Diagram—7QC Tool

Desired-results Fishbone This technique uses the Ishikawa concept but uses a “desired result” as the focus, rather than a problem. Use the same procedure as when building a Cause-and-Effect diagram but write the desired outcome in the head of the Fishbone diagram and brainstorm ways to potentially achieve that goal. Benefit: This technique helps to identify and organize a large number of improvement ideas. Weakness: To avoid the natural tendency to “jump” to solutions before the root cause is understood, it is best to use this approach with new design or creation ideas, rather than solving a problem by rewording the problem in a “positive” way.

Fault Tree Analysis This uses a top-down analytic method and Tree diagram structure with unique symbols to dissect a specific problem or problematic outcome into its related causes. (See Also “Fault Tree Analysis,” p. 309) Process Fishbone This tool is used to analyze the potential causes of a problem within a process, typically a production or assembly, but can also be used with any services process. Start with a high-level process flow as the “main line” or “spine,” with the problem identified in a box at the far right-end of the flow. Draw the separate Fishbone (with its related potential causes) for each process step. Potential causes also can occur in the transition, or hand-off, between steps and should be drawn with a separate Fishbone. Benefit: The tool follows the sequence of the process and thus is relatively easy to develop and easy for a “recipient” to understand. Weakness: If similar potential causes recur throughout the process, they are repeatedly drawn. Thus, the compounding impact of a repeating cause or a combination of several related causes is difficult to illustrate.

Reverse Fishbone Diagram This variation is used when identifying potential consequences to one or more solutions, improvement ideas, or possible actions. It also is called Backward Fishbone, Action and Effect diagram (AED), or Solution Impact diagram. Similar to the desired-result Fishbone approach, the reverse Fishbone focuses on a proposed solution. Follow the basic Ishikawa diagram procedure but write the possible solution at the “head” of the fish. The bones identify possible effects the solution may have: • You can initiate the brainstorming portion of the approach by answering, “What effect could this solution have?”

187

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

188

Encyclopedia

• Both positive and negative impacts of the solution should be explored. • Categorizing techniques of the possible effects could include colorcoding or marking the ideas to distinguish the positive from the negative. A B C D E F G H I J K L M N O P Q R S T U V W X

If several options are being explored, develop one reverse Fishbone per possible solution. Benefit: This approach leverages the knowledge of experienced people, particularly if a diverse cross-section of functional expertise participates. Weakness: The results may fall victim to groupthink or paradigm paralysis, and potentially good solutions may be discounted too quickly, or a sub-optimal solution may slide through the scrutiny process too easily.

Time-delay Fishbone This tool is used to collect input from people over time by posting a large (poster-size or bigger) Fishbone diagram in a high-traffic area (office break room or cafeteria) with pens available nearby. Passers-by are given a standing invitation to add to the diagram. Sometimes copies of the diagram posted in multiple locations are needed to capture a good crosssection of workers. Benefit: More people can participate in building the diagram, and it extends beyond a meeting date and time, allowing people to reflect, research, and contemplate the causes over time while working. In addition, it can create some excitement (a “buzz”) in the office as people contribute. Weakness: Sometimes documenting the final Fishbone requires interpretation without testing to understanding with the author of the idea. Interpretation may be needed because symptoms (rather than causes) are posted, poor handwriting, unclear comments, or apparent redundancy. Moreover, finding an adequate central location for all to see can be difficult if the work environment includes several office travelers or “remote” workers.

Cause-and-Effect Prioritization Matrix

Y Z

What Question(s) Does the Tool or Technique Answer? Which key variables (process steps or process inputs) best meet customer requirements (or the key process output variables)? A Cause-and-Effect Prioritization matrix helps you to • Understand the impact or effect of key process elements cause to key outputs, as defined by customer priorities.

Cause-and-Effe ct Prioritization Matrix

189

• Focus and prioritize activities according to what is important to customers, regardless if the resulting action improves a problem area or creates an innovation or delighter.

Alternative Names and Variations This tool is also known as • Cause-and-Effect matrix (or C and E matrix) • Prioritization matrix Variations on the tool include • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173) • Decision matrix, or Pugh matrix (See Also “Pugh Concept Evaluation,” p. 534) • QFD (Quality Function Deployment) or House of Quality (See Also “Quality Function Deployment (QFD),” p. 543)

A B C D E F G H I J K L M N

When Best to Use the Tool or Technique The Cause-and-Effect Prioritization matrix is an analysis tool to 1) evaluate the relative customer value of what is being provided today, 2) determine the biggest impact (or effect) on meeting customer requirements, and 3) determine where to potentially focus on opportunities for improvement.

O

This is best used when there are differing opinions about which process steps have the biggest impact on fulfilling requirements or when the number of input variables need to be narrowed.

T

P Q R S U V

Brief Description The Cause-and-Effect Prioritization matrix identifies those potential root causes with the biggest impact on an effect, using customer requirements as its evaluation criteria. This analytical tool is similar to a Cause-and-Effect diagram in that it draws a relationship between the potential causes and an effect, but the C and E Matrix has an expanded purpose, which is to 1. Understand the relationship between output variables and customer requirements, based on what is important to the customer.

W X Y Z

190

Encyclopedia

2. Understand the relationship between process steps (or input variables) and the key process output variables, based on what is important to the customer.

A B C D E F G H I J K L

The C and E matrix examines the relationship of various process steps (or inputs) with prioritized customer requirements (or output variables). The purpose of this tool is to determine which process steps (or inputs) have the strongest correlation to the requirements, to determine areas of focus. This tool does not examine current performance. This tool requires customers’ input on not only their requirements, but also the relative priority ranking of each. Collecting this data is important to ensure the validity of the tool. If the data does not exist, the investment of time and money to collect this customer data is an opportunity to better understand customer needs and is a critical input that cannot be overlooked. The customer may be either external or internal. If a known defect or problem is limited to a set of process steps (or input variables), the application of the C and E Prioritization matrix may be restricted only to those that are suspected to be causing variability (or nonconformance) to one or more of the key output variables. Hence, the C and E matrix prioritizes which of those process steps (or input variables) is likely to be causing the biggest impact.

M N O P Q

How to Use the Tool or Technique When developing a Cause-and-Effect Prioritization matrix, you must Step 1.

R S

Identify key customer requirements and their relative importance (or ranking). Potential Voice of the Customer (VOC) information sources include • Key outputs from the Process map

T U

• VOC studies (focus groups, interviews, surveys, and so on)

V

• CTQ matrix or tree

W

Step 2.

List the key customer requirements across the top of a grid as column headings (one topic per column).

Step 3.

Place the customer’s relative importance of each requirement in the row just beneath the column headings. These are collectively called the Priority Score. The scale usually ranges from 1 to 10 (with 10 being the most important or having the largest impact) and no duplicate ratings.

Step 4.

Create the following extra columns in the grid:

X Y Z

Cause-and-Effe ct Prioritization Matrix

191

• Total Score—as the last column in the grid, placed to the far right of the last key customer requirement column. • Process steps (or inputs)—as the first column in the grid, placed to the far left of the first key customer requirement column. Step 5.

Step 6. Step 7.

Identify all the process steps and key inputs from the Process map and list each one as an individual row heading, below the row containing the Priority Scores.

A B C

Divide each open cell in half, with a diagonal line from the upper right corner to the lower left.

D

Determine the Correlation Rating. Examine each step (or input) individually and evaluate how well the step (or input) influences or fulfills each customer requirement using the following rating scale:

F

• Blank or 0 = No correlation

E G H I J

• 1 = Remote correlation

K

• 3 = Moderate correlation

L

• 9 = Strong correlation

M N

Step 8.

Place the Correlation Rating score for each process step in the upper left of each cell.

O

Step 9.

Within each cell, multiply the Correlation Rating score times the Priority Score and write that product in the lower right corner of the cell. (Correlation Rating x Priority Score) For example

Q

• Step 1 and Customer Requirement 1 = (3 correlation x 10 priority) = 30 • Step 2 and Customer Requirement 2 = (9 correlation x 8 priority) = 72 • Step 3 and Customer Requirement 3 = (1 correlation x 5 priority) = 5; until… • Step n : Customer Requirement 4 = (3 x 3) = 9 Step 10.

Determine the Total Score. Add the multiplication products (in the lower right corner) across each row to determine the Total Score for each process step (or input). Record that number in the far right column. Continuing with the preceding example • Step 1 Total Score = 30 + 0 + 5 + 0 = 35

P R S T U V W X Y Z

192

Encyclopedia

• Step 2 Total Score = 0 + 72 + 0 + 9 = 81 • Step 3 Total Score = 90 + 0 + 5 + 27 = 112; being the highest total • Step n Total Score = 10 + 0 + 15 + 9 = 34 A

Step 11.

B C D

Evaluate the Total Scores for each process step (or input) for reasonableness. The Total Scores should mirror the strength of the relative correlation that each process step (or input) has with the customer requirements. If something seems amiss, double-check to see if the following error conditions exist:

E

• Incorrect correlation rating score

F G

• Missing process step or input variable

H

• Unrelated customer requirement (or output variables) with the process

I

• Unrecognizable linkage between the process and the customer requirements/deliverables

J K L M N

Table C-1 provides a Cause-and-Effect matrix template to display its generic structure using the example priority scores and calculations described in the preceding steps: Table C-1: Generic Cause-and-Effect Matrix Template

O P Q R

Customer

Customer

Customer

Customer

Requirement 1

Requirement 2

Requirement 3

Requirement 4

Process Steps 10

8 Priority Scores

S T

Step 1

0

Y Z

9 0

Step 3

9

• • • Step n

3

0

0

1x5=5

1

3x3=9

9 1x5=5

(30 + 5) 81

3 0

0

35

0

0 9 x 8 = 72

9 x 10 = 90

5

1 0

3 x 10= 30

Step 2

W X

0

3

U V

Total Score

(72 + 9)

112 9 x 3 = 27

(90+5+27)

3x3=9

(10+15+9)

SUMMATION 1

0 1 x 10 = 10

3 0

3 3 x 5 = 15

34

Correlation Scale: 0=None; 1=Remote; 3=Moderate; 9=Strong

Cause-and-Effe ct Prioritization Matrix

193

How to Analyze and Apply the Tool’s Output • The highest ranked process steps (or inputs) indicate the strongest influence on meeting customer requirements. • You might wish to indicate the top Total Scores (by circling, bolding, or changing the color of the number). Alternatively, display the findings in a Pareto chart visually to depict the relative contributions of the process steps (or inputs) to the customer requirements (or outputs). • Determine the current performance to determine if there is room for improvement to identify the areas with the largest correlation and the poorest performance. • Confirm the results and determine the magnitude of impact. Tools to support this next step include:

A B C D E F G H

• Stratified Dot Plots

I

• Scatter Plots

J

• Testing “quick fixes”

K L

• Multi-vari charts

M

• Correlation analysis

N

• Hypothesis testing (that is, Regression, Logistic Regression, ANOVA, or Chi-Square Test) • Design of Experiment • The results can serve as input to developing a Control Plan or FMEA.

O P Q R S T U

Examples Cause-and-Effect Matrix Example Table C-2 presents a sample Cause-and-Effect matrix that explores the relationship between the selling process steps for computer consulting services and customer requirements.

V W X Y Z

194

Encyclopedia Table C-2: Cause-and-Effect Matrix Example for Computer Services Selling Quality Product

Prompt Service

Professional & Knowledgeable People

Competitive Price

10

8

9

5

Process Steps

A B C

Identify Prospect Understand Customer Needs

9

F

Present Value Proposition

9

G

Negotiate Deal

D E

3

3 132

90

27 3

90

15 9

27

45 9 45

H I J K

Total Score

Deliver Offering

3

Support Offering

9

9

3

30

72 9

9 27

9

90

72

45 9

81

45

162 45 174

288

L M

W

Poorly Constructed C and E Matrix Example The following example examines the relationship between customer requirements for purchasing a high-end stereo versus the selling process steps at a retail store. Notice that many of the retail selling process steps have very low impact on the customer requirements. If the retail store owner wanted to better understand why the store was experiencing low sales, the owner might probe further into customer requirements that are more applicable to the selling process or examine other controllable processes at the store (such as brands inventoried versus those of discount stores). Moreover, when the requirements were gathered, the customer did not discriminate among all the needs, but simply gave them equal weight, which provides little insight. Table C-3 shows an ineffective use of a Cause-and-Effect matrix in that the answer is obvious and does not require a matrix to sort out the answer.

X

Table C-3: Ineffective Use of a Cause-and-Effect Matrix—Selling a High-end Stereo Example

N O P Q R S T U V

Y Z

Big Sound

Silver Exterior

Remote Control

Illuminated Dial

Lowest Purchase Price

10

10

10

10

10

Process Steps

Total Score

Identify Prospect Understand Customer Needs

1 10 10

Cause-and-Effe ct Prioritization Matrix Big Sound

Silver Exterior

Remote Control

Illuminated Dial

Lowest Purchase Price

Process Steps Present Value Proposition

1

3 10

Negotiate Deal

30

Total Score

40 0

3

30 30

A B

Install Stereo

Service Stereo

195

1

10 10

C D E F

Hints and Tips

G

The number of key customer requirements should be limited to about

H

five. Select the appropriate customer requirements driving the set of

I

process steps (or input variables) being evaluated. The relationship

J

between the process steps (or inputs) and the Customer Require-

K

ments may produce a grid with about a half to two-thirds of the grid

L

with empty cells (a blank influence rating) showing no correlation. If

M

the grid contains a majority of cells filled in with Correlation Rating

N

scores, then evaluate whether a relationship presumption is being

O

forced. If the grid contains too few cells filled, examine whether the

P

customer requirements are appropriate for the given process steps

Q

(or input variables).

R

The Correlation Rating simply determines the magnitude of impact

S

or correlation between a process step (or input) and customer

T

requirements (or outputs) and is indifferent to a positive or negative

U

relationship. Hence the scale should always be positive number—no

V

negative numbers. The Correlation Rating scale conventionally used

W

is “Blank, 1, 3, 9” to create a spread that better distinguishes high

X

correlation relationships from the remainder of the considerations.

Y

This tool evaluates the magnitude of impact between a process step

Z

(or input) and customer requirements (or outputs); it does NOT evaluate current performance.

196

Encyclopedia

Supporting or Linked Tools Supporting tools that might provide input when developing a Causeand-Effect Prioritization matrix include

A B C

• Key outputs from the process map (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522) • VOC studies (focus groups, interviews, surveys, and so on) (See Also “ Voice of Customer Gathering Techniques,” p. 737)

D

• CTQ matrix or tree (See Also “Critical to Quality (CTQ),” p. 242)

E

• Fishbone diagram (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173)

F G H I

A completed Cause-and-Effect Prioritization matrix provides input to tools such as

J

• Stratified Dot Plots (See Also “Dotplot,” p. 280)

K

• Scatter Plots (See Also “Scatter Diagram—7QC Tool,” p. 640)

L

• Testing “quick fixes”

M N

• Root Cause Analysis Techniques:

O

• Multi-vari charts (See Also “Multi-vari Chart,” p. 439)

P

• Correlation and Regression analysis (See Also “Graphical Methods” and “Regression Analysis” p. 323 and 571 respectively.)

Q R S T U V W X Y Z

• Hypothesis Testing (that is, Logistic Regression, ANOVA, or ChiSquare Test) (See Also Analysis of Variance (ANOVA)—7M Tool and “Hypothesis Testing” p. 142 and 335 respectively.) • Design of Experiment (See Also “Design of Experiment (DOE),” p. 250) • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) • Control Plan (See Also “Matrix Diagrams—7M Tool,” p. 399) Figure C-8 illustrates the link between the Cause-and-Effect Prioritization matrix and its related tools and techniques.

Cause-and-Effe ct Prioritization Matrix

197

Stratified Dot Plots

Scatter Plots Process Map

VOC and VOB Data

A Testing “quick fixes” Cause and Effect Matrix

CTQ Matrix or Tree

Root Cause Analysis Techniques

B C D E F

FMEA Fishbone Diagram

G H I

Control Plan

Figure C-8: Cause-and-Effect Tool Linkage

J K L

Variations Cause-and-Effect Diagram Depicts the relationship between a given problem and its potential causes. (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173 for more detail.) Decision Matrix Uses similar principles as the Cause-and-Effect matrix but has a slightly different purpose. The Decision matrix (sometimes called the Pugh matrix) is a selection tool used to determine, from a list of options against a weighted set of criteria, which option to pick. (See Also “Matrix Diagrams—7M Tool,” p. 399 for details on different types of matrices.)

QFD (Quality Function Deployment) QFD is a more in-depth tool than the Cause-and-Effect matrix but uses similar principles. The QFD tool (sometimes called the House of Quality) translates customer requirements or needs (using CTQs—critical to quality) into actions and designs that build and deliver a quality offering (product and/or service). (See Also “Quality Function Deployment (QFD),” p. 543.)

M N O P Q R S T U V W X Y Z

198

Encyclopedia

Cause and Prevention Diagram What Question(s) Does the Tool or Technique Answer? What are the potential risk response strategies to a potential risk or failure? A

A Cause and Prevention diagram helps you to

B C D E F G H I

• Organize risk response strategies to understand the relationship between various response strategies to a potential failure (or risk) by formatting, arranging, organizing, and parsing the strategies into themes as part of risk mitigation planning. • Stimulate thinking when developing the list of the responses to a potential problem. • Guide concrete action. If the risk occurs, the diagram serves as a checklist or tracking tool for those responding to the given situation.

J K L M N

Alternative Names and Variations Variations on the tool include: • Cause-and-Effect diagram

O P Q R

When Best to Use the Tool or Technique Before any action is taken, this tool helps to define and organize risk response strategies to a potential problem (or risk).

S T U V W X Y Z

Brief Description A preventative tool used for risk mitigation planning. It focuses people on the various response action plans, depending on the given situation, and is a variation on the Cause-and-Effect diagram (or Fishbone) tool. (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173) Though the two tools have a similar structure, the Cause and Prevention diagram warrants its own entry to highlight its purpose as a proactive risk tool, versus the C and E matrix’s post-mortem approach of examining potential root causes. The Cause and Prevention diagram starts with the identification of a potential failure or problem and places it at the “head” of the Fishbone structure. Each major bone represents a major risk response category, stemming off the spine. The brainstorming technique is used to identify various risk response action plans. These response plans are mapped and organized with the corresponding risk response categories and depicted as the related “bones” stemming from the spine.

Cause and Prevention Diagram

199

It can be depicted as a matrix or Fishbone diagram wherein the problem (or risk) is placed to the far right. The problem is placed in a box or diamond-shape to represent the head of a fish. The prevention strategy themes are placed to the left of the problem. There are three major types of risk response classification: 1. Acceptance—Accept the consequences passively or actively; retain the risk. 2. Avoidance—Eliminate a specific threat, usually by eliminating the cause. 3. Mitigation—Reduce the expected monetary value of a risk event by reducing the probability of occurrence. a. Reduction—minimize its occurrence and effect b. Transfer—all or a portion of it to another party by using or implementing the following: i. Insurance for direct property damage

A B C D E F G H I J K

ii. Indirect consequential loss (often performed by a contractor, for example, debris removal or equipment replacement)

M

iii. Legal liability (design errors, public bodily injury, performance failure)

O

iv. Personnel (employee bodily injury/Worker’s Compensation)

Q

v. Employee replacement costs vi. Resulting business losses

L N P R S T U V

How to Use the Tool or Technique When developing a Cause and Prevention diagram, consider the following:

W

Step 1.

Y

Step 2.

Start with a prioritized list of potential risks or failures, which could come from an FMEA matrix. Agree on the risk or potential failure to be analyzed. (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) Write the potential risk to the far right edge of the diagram and draw a box around the text to form the “head” of the fish, as shown in Figure C-9.

X Z

200

Encyclopedia

Prevention – Risk Response Stategies Risk Response

A B

Risk Response

Risk Response

Bones

C

Cause

D

Potential Risk

Spine

E F G

Risk Response

H

Risk Response

Risk Response

I J K L M

Figure C-9: Generic Cause and Prevention Diagram

N O

Step 3.

Draw a line extending from the left edge of the fish head to the left edge of the diagram to represent the spine.

Step 4.

Agree on the major response categories: Acceptance, Avoidance, Reduction, Transfer, Other, and any additional major response approach theme.

Step 5.

Reference policies, standard operating procedures and experience, brainstorm and write down the potential response action plans, and group them within the related major response categories.

P Q R S T U V W X

Hints and Tips

Y

Have the participants attending the meeting bring policies, proce-

Z

dures, instructions, and guidelines from their respective areas to inform the brainstorming activities. (See Also “Brainstorming Technique,” p. 168)

Cause and Prevention Diagram

Step 6.

Draw a line extending from the spine for each theme and write a major response category name at the outermost end of line (the end that is not attached to the spine) to represent a “main” Fishbone.

Step 7.

Draw one “branch” or offshoot to the main Fishbone and label the end of each branch with the specific response action plan idea to represent smaller bones and to show the linkage to the “higher level” category with which it is affiliated.

201

A B C

Step 8.

Continue Step 7 to drill-down into as much detail as required defining more bones and linking the relationship.

D

Step 9.

Review the diagram for completeness and clarity and then modify accordingly.

F

E G

a. Fill in gaps and streamline wording.

H

b. Eliminate unnecessary redundancy.

I

c. Eliminate unrelated ideas.

J K

How to Analyze and Apply the Tool’s Output

L M

• Review the final diagram for completeness and clarity.

N

• Incorporate this Cause and Prevention diagram in the appropriate control documentation:

O

• Risk Mitigation Plan • Transition Plan • Implementation Plan

P Q R S T

• Communication Plan

U

• Control Plan

V

• Standard Operating Procedure manual.

W X Y Z

202

Encyclopedia

Hints and Tips Prioritize the most likely to occur negative events and develop one Fishbone per potential major “negative” outcome. This is a good tool A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

to develop an action plan for the high-priorities following an FMEA. (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) It might help to categorize the different types of risk according to the cause or source of risks: Business Risks (chances for a profit/loss associated with any business endeavor) and Pure (Insurable) Risk (only a chance for loss). Rank the risk according to the ability to manage effective responses: Opportunity (probability of occurrence certainty) and Threat (magnitude if occurred and the amount of information you have about the risk). Then look at possible cumulative effect of several risk events occurring in conjunction with each other and/or dependencies. Identify the Tools and Techniques/Processes to organize, document, and respond to risk if it occurs. Key Questions to ask during the response plan generation step are as follows: 1. How could this risk be avoided? 2. Can this risk be reduced? 3. Can the risk be shared or transferred? 4. Should we face the risk, and if so, should scheduling and financial allowances be made? 5. Can we contain the risk? Benefits: Can be a thorough, in-depth analysis of possible response strategies. Weakness: Need multiple Fishbone diagrams for the multiple possibilities of risk occurrences. This requires time to develop the analysis, and the planning may never be needed. It may be difficult to prioritize and identify the possible negative outcomes.

Cause and Prevention Diagram

203

Supporting or Linked Tools Supporting tools that might provide input when developing a Cause and Prevention matrix include • FMEA (Failure Modes and Effect Analysis) (See Also Failure Modes and Effects Analysis (FMEA), p. 287) A completed Cause and Prevention matrix provides input to tools such as

A B C

• Checklist defining an action plan or standard response procedure (See Also “Checklists—7QC Tool,” p. 204)

D

• Control Plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

E

• Risk Mitigation Plan (See Also “Risk Mitigation Plan,” p. 601)

F G

• Transition Plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

H

• Implementation Plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

I

• Communication Plan (See Also “Matrix Diagrams—7M Tool” p. 399)

J K

• Standard operating procedure (SOP). Figure C-10 shows the link between the Cause and Prevention diagram and its supporting tools and techniques. Control Plan

L M N O P Q

Risk Mitigation Plan

R S

Transition Plan FMEA

Cause and Prevention Diagram

T U V

Implementation Plan

W X

Communication Plan

Standard Operating Procedure

Figure C-10: Cause and Prevention Tool Linkage

Y Z

204

Encyclopedia

Checklists—7QC Tool

A

What Question(s) Does the Tool or Technique Answer? What activities or deliverables are required to meet requirements? Alternatively, what was observed?

B

Checklists help you to

C D

• Outline (or list) the set of tasks or deliverables to be completed.

E

• Remind someone of the various components to fulfill requirements.

F

• List potential considerations to determine if appropriate or required for a given situation.

G H I J K

• Recap the standard operating procedure (SOP) activities or deliverables. • Collect data about “how many” or “what type” of something occurred.

L M N

Alternative Names and Variations This tool is also known as a check sheet.

O P Q R S

When Best to Use the Tool or Technique This tool precedes work and summarizes a set of pre-determined activities, action items, deliverables, or questions to be referenced before and/or during work.

T U V W X Y Z

Brief Description The checklist is a generic tool that may provide recommendations or considerations of “best practices” or “options” to be evaluated for a given situation. Alternatively, the tool may outline the standard operating procedure (the set of required tasks) and/or outputs that must be completed to satisfy a given set of requirements. This generic tool can be as simple or as complex as the situation requires. It is a powerful tool in that it is flexible and customizable to reflect a range of complexity, depth, and breadth needed to complete a work procedure. The checklist is a member of the original 7QC Tools, (or seven Quality Control tools), attributed to Dr. Kaoru Ishikawa. The 7QC tools sometimes are called the “seven basic tools” because they were the first set of tools identified as the core quality improvement tools. Ishikawa’s original 7QC Toolset includes: 1) Cause-and-Effect diagram, 2) Check sheet (or

Che cklists—7QC Tool

205

checklist), 3) Control charts, 4) Histogram, 5) Pareto chart, 6) Scatter diagram, and 7) Stratification. Structure the checklist to include the critical parameters of a task, such as the following: • Required procedure—The sequence of activities or tasks to fulfill a certain standard or set of requirements. It can include who is to do what task or who is involved, as illustrated in Table C-4, and sometimes may include who the customers, decision-makers, and subject matter experts are.

Process Steps

Sales Specialist

B C D E

Table C-4: Generic Procedure Checklist Example for Selling Sales Rep

A

Field Engineer

Approver: District Manager

F G H

1. Identify Prospect

X

2. Understand Customer Needs

X

X

3. Develop Value Proposition, Price and negotiation boundaries

X

X

X

K

X

X

L

4. Present Value Proposition 5. Negotiate, if necessary

X

I J

M

X

6. Sign Contract

N

X

7. Deliver Offering

X

O

8. Support Offering

X

P Q

• List of deliverables or outputs—This can include critical features or functionality aspects. Can include who is accountable for producing the deliverable, as seen in Table C-5. Sometimes may include who are the (internal or external) customers and subject matter experts. Table C-5: Generic Deliverables Checklist Example for Selling

Deliverables

Sales Rep

Sales Specialist

Field Engineer

R S T U V

Approver: District Manager

W X

1. Prospect identified and entered into sales tracking database.

X

Y

2. Customer needs understood and documented

X

Z

3. Offering value proposition and price documented in presentation format.

X

consulted

approved

informed

informed

X

4. Negotiation boundaries defined.

continues

206

Encyclopedia Table C-5: Continued

Deliverables

Sales Rep

Sales Specialist

X

support

5. Value Proposition presented.

A B C D E F

6. Negotiations conducted, if necessary, final agreement reached. Customers signature secured on contract 7. Contract signed by manager, copy, then sent to customer.

Field Engineer

Approver: District Manager

X X

approved

8. Offering delivered and installed

X

9. Customer trained and offering supported.

X

G H I J K

• List of planning considerations (best practices or conditional what-ifs to evaluate prior to or during an activity)—If the added complexity is useful, the potential conditions should include the corresponding next steps or alternatives, as seen in Table C-6. Table C-6: Generic Planning Checklist Example for Selling

L M N O P

Considerations to select appropriate selling support literature: 1. Identify industry segment 2. Identify role of key contact 3. Who has budgetary responsibility? What is the relationship with the key contact?

Q

4. What is the total department budget?

R

5. Are the funds allocated for this initiative? If so, how much?

S

6. What is the timeframe for the initiative?

T

7. What are the key issues (pain) trying to be solved?

U V W X Y Z

8. What are the departmental near term and long term goals? How are they performing against those goals? 9. Do other departments have similar issues? 10. Are there any critical (internal or external) suppliers? 11. Who are the (internal and / or external) customers? What are their requirements? How does the key contact know? How recent is the information? 12 How predisposed is the key contact to our solutions (exploring options, not sure if will buy; interested but undecided which is the best approach; ready to buy)

• A collection tool—To capture frequency of events or conditions needing to be tallied (attribute data), such as number of something (for example, defects, phone calls, and so on) or type of something (categories, names, colors), as illustrated in Table C-7.

Conjoint Analysis

207

Table C-7: Generic Tracking Checklist Example for Selling Quarterly Selling Activity:

Q1

Q2

Q3

Q4

1. # incoming leads from trade shows and advertising 2. # phone calls made 3. # customer emails sent

A

4. # exploratory appointments / meetings scheduled

B

5. # Value proposition meetings conducted

C

6. # deal negotiations in process 7. # deals closed 8. # customer contracts signed 9. # offerings delivered and installed 10. # of customer “relationship” building meetings conducted

D E F G H

The power of the checklist comes about with active use. Its strongest attributes—simplicity, flexibility, and scalability—are also its weakness. It must be used and maintained to reflect the current thinking or best practice to reap its powerful benefits; otherwise, it could become obsolete or irrelevant. As with other soft tools, the checklist should be part of an evergreen process to maintain its applicability. The refresh process keeps people actively engaged using it because it will reflect the latest lessons learned and best practices.

Conjoint Analysis

I J K L M N O P Q R

What Question(s) Does the Tool or Technique Answer? What product and/or services features and functionality do customers prefer and would be willing to buy? A conjoint analysis helps you to • Determine consumer preference and behavior by making trade-offs between choices. • Understand price sensitivities to various product and/or services attributes and levels to determine the most appealing combinations. • Characterize and verify customer requirements (both implicit and explicit).

S T U V W X Y Z

208

A

Encyclopedia

When Best to Use the Tool or Technique A conjoint analysis collects the Voice of the Customer (VOC) and derives preference profiles from a portfolio of options to predict buying behavior. The technique is best used during the design phase of a new product and/or services offering or when an existing offering needs to be modified or refreshed to help drive sales and market share. It can also be used during the define phase of a project to determine what is important to the customer.

B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Brief Description The conjoint analysis technique was developed in the early 1970s as a market research tool to assess the potential profitability, sale revenue, and market share for a new or modified offering. It employs a structured methodology to determine consumer preference and buying behavior by studying the joint effects of multiple attributes. The term “conjoint” means two or more items joined together or combined. The research subjects are asked to select their preference given a set of alternatives. The trade-off decisions indicate price sensitivity and determine the most appealing combination of attributes for an offering. The conjoint analysis reveals the underlying preferences that guide purchasing decisions. It determines which features to offer that will appeal to maximum number of potential purchasers. It also identifies segments that will prefer a particular set of features and measure how desirable that combination may be to target. The conjoint model decomposes a potential product and/or services offering into a few major attribute categories, wherein each attribute is further decomposed into different levels. The levels can be non-metric and metric levels. The potential offering scenarios are created by assembling different attribute combinations (or bundles) using a fractional factorial design (from the DOE technique) to create a subset of the possible options. The research respondents evaluate each “test” alternative combination to determine their preferences. This selection asks the respondents to trade-off various attributes for others to decipher relative degrees of importance and/or price sensitivities. The conjoint analysis defines a set of part-worths (also referred to as utilities) for each attribute to understand the relative importance of each feature and the value of its different levels when making a purchasing decision. There are three different conjoint analysis methods: The adaptive conjoint, which is probably the most popular approach, employs a rough ordering of importance and then uses pair-wise tradeoffs to compute the part-worth. The hybrid model weighs each attribute from zero to ten, and then 100 points are allocated across all attributes and respective levels. The bridging design is used for a large number of attributes, and it evaluates bundles on a subset of attributes common across several respondents.

Conjoint Analysis

209

The conjoint analysis examines the relative value of attributes considered jointly versus in isolation. This technique’s power is that it more closely aligns to real-world trade-off scenarios, using a statistical approach to economically evaluate a fraction of the possible combinations, ultimately to predict consumer choice among multi-attribute offerings. A

How to Use the Tool or Technique The conjoint analysis procedure involves three steps, which build on the DOE fundamentals. As with a traditional DOE, the important and often the most time-consuming activity in the process is the planning step. The general conjoint analysis procedure is as follows: Step 1.

Design conjoint study. a. Select attributes relevant to product (or service) category. This refers to appearance, size, or price of an offering, for example. A good source of input is from the various data collection methods such as focus groups of targeted customers. Also consider utilizing secondary sources of data, such as desktop research, Web blogs, and periodicals. Consider whether the target audience is segmented. If so, unique market segments may require additional or unique attributes. b. Select levels for each attribute.

B C D E F G H I J K L M N O P

A level represents the values or options for each attribute, such as the different appearance options, different size options, or the different price options. The more options or levels introduced into the study, the greater the complexity for the respondents to select an option, and the greater the complexity of the test design and analysis.

Q

Try to include the same number of levels for each attribute to simplify evaluation for respondents and to avoid misleading results on the importance of attributes. Consider picking similar ranges found in existing products or services to compare with their “delighters” and “dissatisfiers” (from a Kano Analysis). Figure C-11 illustrates potential attributes and respective levels for a new computer concept. Notice that although the design has a relatively large number of attributes, it features an equal number of levels per attribute, thereby simplifying the analysis.

V

R S T U W X Y Z

210

Encyclopedia Computer

Target Audience

A

College Student

Attributes

B C

Sales Channel

Plastic Color

Camera

Screen Size

Price

D E

Attribute Levels

F G H I

G Black

N Orange

Internet

University

Embedded

None

12

20

$999

$2599

Figure C-11: Computer Example for Conjoint Attributes and Levels

J K L M N O P Q R S T U V W X Y Z

c. Develop relevant option bundles (or combinations) to be evaluated. Determine appropriate bundles of attribute profiles to define a potential offering to test. A bundle contains one option (or level) of each attribute selected for analysis. The combination of attribute options within a bundle defines the product or services offering. The complete set of combinations defines the possible design options to be tested, using a fractional factorial (orthogonal array) design to reduce the number of evaluation options instead of full factorial design with all possible combinations. (See Also “Design of Experiment (DOE),” p. 250) If every attribute at every level were to be tested in the computer example shown in Figure C-11 (a full factorial design), the test would involve 64 different possible combinations (2x2x2x2x2x2). If the design were run at halffractional factorial DOE, the number of unique runs reduces to 32. If the Taguchi Orthogonal Array were used, only 16 runs would be required. The experimental design is based on the amount of acceptable confounding balanced by the available time, resources, and funding. In DOE terminology, the “resolution” must reflect the business conditions.

Conjoint Analysis

211

To double-check the number of runs required for a design, using MINITAB, follow the procedure outlined as follows: • Full Factorial Design (for 6 attributes, 2 levels each)—The drop-down menu to highlight: Stat>DOE>Factorial>Create Factorial Design… to open the main screen. Select 2-level factorial (default generators) and enter 6 into the Number of Factors dialog box. Click OK. The resulting session window display 64 runs. • Half-Factorial Design (for 6 attributes, 2 levels each)—The drop-down menu to highlight: Stat>DOE>Factorial>Create Factorial Design… to open the main screen. Select 2-level factorial (default generators) and enter 6 into the Number of Factors dialog box. Click the Designs… button and select the 1/2 fraction 32 VI 2**(6-1) design option and click OK to close the Design window. In the main screen, click OK again. Click OK. The resulting session window displays 32 runs. (Note this is a Resolution IV design.) • Taguchi Orthogonal Array Design (for 6 attributes, 2 levels each)—The drop-down menu to highlight: Stat>DOE>Taguchi>Create Taguchi Design… to open the main screen. Select 2-Level Design, and enter 6 into the Number of Factors dialog box. Click the Designs… button and select the L16 2** 6 design option and click OK to close the Design window. In the main screen, click OK again. The resulting session window displays 16 runs. The MINITAB Worksheet can be used as the design structure and the data collection sheet for the respondents’ “preference score.” The design structure defines how to combine the various attributes and respective levels to ensure randomness. For those fractional factorial experiments, the design structure also identifies which runs to select while maintaining balance. In addition, these procedures are illustrative and lack “replicates.” Replicates should be added to the design to make the experiment more robust and improve the experiment’s ability to predict.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

212

Encyclopedia

If the design involved three levels for each attribute, the complexity would be compounded, resulting in 729 runs for a full factorial design (3x3x3x3x3x3); however, an Orthogonal Array design would result in only 27 runs. Step 2. A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Obtain data from a sample of respondents. a. Choose the format to present the bundles. Select the form in which the respondents will receive the bundled options. Typically, the bundles can be represented as a verbal presentation, written description, pictorial presentation, or samples (that is, proto-types). Often a “card” is used to display the configuration of bundled attributes and appropriate levels. Each run requires a unique “card,” whereby the set of cards represent the full portfolio of multiple configurations options. The respondent, then, sorts the cards in rank-order from highest preference to lowest. b. Design a data collection procedure. The respondents assign part-worth for each level of each attribute. Rank-order approach: The product bundles could be rankordered (for example, one to sixteen), with one representing the highest or most preferred rank and sixteen the lowest. Rating approach: If a ratings scale is used, respondents evaluate each product on scale of 0 to 100 points with larger sums indicating preference. The advantage of this approach is that the Least Squares Regression can be applied with dummy variables to compute part-worth functions. [A constant sum also can be allocated (100 points, for example) across 16 bundles.] c. Select computation method for obtaining part-worth function. The data is entered in the data collection worksheet of a statistical software application for analysis (MINITAB). Obtain the part-worth functions by computing the preference score for an attribute and computing the average scores of the bundles that contain that attribute. For example, if four bundles contain one attribute, add up the respondent’s preference scores for those four bundles and divide by four to calculate its average. This computed number represents the part-worth.

Conjoint Analysis

213

The part-worth model is one of the simpler models to define the various attributes. There are also linear models (using vectors) and quadratic models (using ideal-points). Step 3.

Evaluate product design options. a. Segment customers based on part-worth functions. Determine how to arrange and aggregate responses. Cluster the part-worth data to see if any patterns emerge. Cluster according to one of the following options: • the individual responses • the pooled responses from all the participants into a one part-worth function • similar preferences Let the part-worth data reveal the number of segments (or latent class), if any. Heterogeneous grouping assumes all customers belong to same segment but differ as specified in part-worth distribution. b. Design market simulations to assess likely success of a proposed new product concept under various simulated market conditions. c. Select choice rule to transform part-worth into choices customers are likely to make to determine the best product and/or services offering. The choice rules include • Maximum Utility—Compute the market share by counting the number of customers for whom that offering provides maximum utility and divide by the number of customers in the study. • Share of Utility—Select the higher the utility of a product corresponds with the higher likelihood that a customer will choose it. • Logit Choice—Determine the proportion of times that product has maximum utility; the brand with the maximum utility varies randomly. • Alpha—Weight the maximum utility and the share of preference wherein the chosen weight ensures that the simulated market share reflects the actual market share of existing products.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

214

A B C D E F G H I J K L M

Encyclopedia

Examples Laptop Computer Functions Example This scenario involves a computer manufacturing’s marketing department trying to increase market share and brand loyalty by wooing a new target audience— high school students preparing to go to college. The target market wants mobility, “hip” design and functionality, and plenty of storage capacity for its large music files—all in all, the best value package. The marketing department wants to leverage current technology but create a new image that convinces the consumers that the product meets their needs. To keep the example simple, presume the high school student population is homogeneous and that a sample size of three is sufficient. The study’s design involves five attributes with two levels and two replicates. (Given the two replicates, each respondent will complete the study twice; however, the card deck must be placed in a different random order between runs.) Design Elements—The computer attributes of interest are • Plastic color (2 levels)—Graphite Black; Neon Orange • Sales Channel (2 levels)—Internet; University

N

• Camera (2 levels)—Embedded; None

O

• Screen Size (2 levels)—12 inches; 20 inches

P

• Price—$999; $2599

Q R S T U V W X Y Z

Figure C-11 illustrates the attributes and their respective numbers of levels, which is unequal across the options. The total possible number of combinations for these five attributes at two levels is 32 (2x2x2x2x2), and with two replicates, the runs increase to 64. Given economic and time pressures, the study will use a halffractional factorial, Resolution V design, requiring only 32 runs. Using MINITAB to build the study’s worksheet, access the main screen by using the following drop-down menus: Stat > DOE >Factorial >Create Factorial Design…. Enter the study’s attributes and level information in the main screen and its three “options” screens, as illustrated in Figure C-12. (See also “Design of Experiment (DOE)” p. 250 for more details.) The respondents will rate the multiple configuration options using “cards.” Each card will contain a unique configuration option. The MINITAB Worksheet displays the bundle alternatives. The final Session Window and Worksheet for this example are shown in Figure C-13. Note that the Worksheet displays the attributes levels as coded integers (1, 2, 3, 4).

Conjoint Analysis

215

1. 3.

2.

A B 5. 4.

C D E F G H I

Figure C-12: MINITAB Set-up for Computer Example

Label the Worksheet column C10 as the “Response” column. Create the 32 response cards, following the design in the MINITAB Worksheet. (One card would feature Neon Orange, Internet sales, Embedded camera, 20 inch screen for $2599 to mirror the first line in the MINITAB Worksheet.) Run the study with multiple respondents and record the results on the MINITAB Worksheet.

J K L M N O P Q R S T U V W X Y Z

Figure C-13: MINITAB Worksheet and Session Window for Computer Example

216

Encyclopedia

For purposes of this example, the conjoint will cover only four respondents with each respondent sorting the cards twice.

A B C D E F G

To analyze the results, from MINITAB’s drop-down menu, select Stat > DOE > Factorial > Analyze Factorial Design…. Enter the column containing the study’s responses as the response variable in the dialog box. Click OK or select any of the analysis options of interest. The Graphs option screen contains the option to select a Pareto diagram. The results of this study show that color has the most significant impact on preference, with all but the screen size contributing to preference, as shown in Figure C-14, which captures the session window and Pareto chart. The respondents indicated that they prefer a neon orange computer to graphite black. The students reported price sensitivity as the second largest impact on preference, favoring the lower of the two prices, at $999.

H I J K L M N O P Q

Figure C-14: MINITAB ANOVA Session and Pareto for Computer Example

R S T U V W

The Session Window Output found in Figure C-14 contains additional terms not defined within this entry. For a discussion on the meaning of R-Sq and R-Sq(adj), reference the Regression Analysis entry. (See Also “Regression Analysis,” p. 571) For a discussion on the meaning of the Analysis of Variance terms, reference the ANOVA entry. (See Also “Analysis of Variance (ANOVA)—7M Tool,” p. 142)

X Y Z

Supporting or Linked Tools Supporting tools that might provide input when developing a conjoint analysis include • Fishbone (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173) • VOC/VOB data (See Also “Voice of the Customer Gathering Techniques,” p. 737) • CTQs (See Also “Critical to Quality (CTQ),” p. 242)

Control Charts—7QC Tool

217

• QFD (See Also “Quality Function Deployment (QFD),” p. 543) • Solution selection matrix (See Also “Solution Selection Matrix,” p. 672) A completed conjoint analysis provides input to tools such as A

• QFD (See Also “Quality Function Deployment (QFD),” p. 543)

B

• FMEA (See Also “Failure Modes and Effects (FMEA),” p. 287)

C

• Pugh matrix (See Also “Pugh Concept Evaluation,” p. 534)

D

Figure C-15 illustrates the link between the conjoint analysis and its related tools and techniques.

E F G H

Fishbone

I J

VOC and VOB Data

QFD

K L

CTQ Matrix or Tree

Conjoint Analysis

Pugh Matrix

M N O

QFD

FMEA

P Q

Solution Selection Matrix

Figure C-15: Conjoint Analysis Tool Linkage

R S T U

Variations DOE (See Also “Design of Experiment (DOE),” p. 250)

Control Charts—7QC Tool

V W X Y Z

What Question(s) Does the Tool or Technique Answer? How is the process behaving; is it in statistical control? Control charts help you to • Determine whether a process is stable (in statistical control) over time.

218

Encyclopedia

• Measure, monitor, and control processes. • Determine what is a special cause event (unexpected variation) that needs to be identified and eliminated. A B C D

• Decipher common cause variation (predictable and expected) or the inherent variation in a process versus unexpected variation. • Predict changes and indicate the need to improve process performance or determine if an improvement actually reduces process variation. (Key triggers include quality, cost, and capacity.)

E F G

Alternative Names and Variations This tool is also known as

H

• Statistical process control (SPC) charts

I

• Process behavior charts

J K L M

Variations on the tool that depend primarily on the type of data being plotted include • Charts for variable data include

N

• X-bar and R chart (or averages and range)

O

• X-bar and S chart (or averages and standard deviation)

P Q

• I and MR chart (or “individuals and moving range”, or I-MR, Xchart, X-R chart, IX-MR chart, XmR, moving range)

R

• Moving average-moving range chart (or MA-MR)

S T U V

• Target charts (or difference charts, deviation charts, or nominal charts) • CUSUM charts (or “cumulative sum”)

W

• EWMA (or “exponentially weighted moving average”)

X

• Multivariate chart (or Hotelling T2)

Y Z

• Charts for attribute data include • p-chart (or “proportion” or “percentage” chart) • np-chart (or number within a proportion (or affected units)) • c-chart (or count chart)

Control Charts—7QC Tool

219

• u-chart (or unit chart/counts-per-unit) • P Prime SPC chart • Charts for either type of data include • Short run charts (or stabilized charts/Z-charts) • Group charts (or multiple characteristic charts)

A B C

When Best to Use the Tool or Technique Control charts should be used on a regular basis as part of the monitoring and management of a process to control variation. They ensure process stability over time—a prerequisite for process capability analyses. Control charts also aid in distinguishing between special cause variation and common cause variation as a guide for management.

D E F G H I

Brief Description A control chart often is referred to as a super-charged run chart or time series plot used to monitor a process over time. It is a frequency distribution plotted over time in the sequence that the data occurred or was produced and adds three reference lines for interpreting patterns in the data. These reference lines comprise a centerline (the average or mean) and control limits—an upper and lower control limit. The control limits are defined as three-sigma on either side of the collected data’s mean. Why three standard deviations? It balances the likelihood of a false signal and maintains sensitivity to detect real signals. A Statistical Process Control (SPC) is the application of statistical methods to identify and control special cause variation in a process. Control charts are a graphical tool used to monitor changes that occur within a process by distinguishing variation that is inherent in the process (common cause) from variation that indicates a change in the process (special or assignable cause). This change may be a single point or a series of points in time—each is a signal that something is different from what was previously observed and measured. Unusual variation is signaled by any point outside or specific patterns within the control limits. Control limits are established at a distance of three standard deviations on either side of the centerline, or mean, of data plotted on a control chart. Do not confuse control limits with specification limits. Control limits help to identify special cause variation and confirm stability over time—the Voice of the Process (VOP) metric. Specification limits describe conformance to customer expectations—a Voice of the Customer (VOC) metric, which are utilized in process capability charts.

J K L M N O P Q R S T U V W X Y Z

220

A

Encyclopedia

Recall that a “bad” part is determined by 1) dispersion in the process, where it is wider than the specified spread, or 2) off-target, where the process variation could be “skinnier” than the control limits but have drifted (that is, are off-center). (See Also “Process Capability Analysis,” p. 486) Therefore, it is important to note that whether or not a process is meeting its control limits depends on the desired outcome itself, whether

B

• larger is better,

C

• smaller is better, or

D E F G H I J K L M N O P Q R S T U V W X Y Z

• nominal (target) is best. As just stated, the type of control chart primarily depends upon the type of data being plotted. Variable data is quantitative results or data where measurements are used for analysis. Sometimes variable data is called continuous data because it is measured on a continuous scale such as temperature, distance, cycle time, profit, or mass and pressure, rather than in discrete units or yes/no options. Ranks are also a type of variable data, such as the customer satisfaction scores on a scale of one to ten. Variable data control charts come in pairs, with the top chart plotting the average or centering of the process data, and the bottom plot displaying the range of the data distribution. Continuous data is used to create an IMR chart (or X-MR), X-bar and R chart, and X-bar and S chart. Attribute data, on the other hand, is qualitative, rather than quantitative in nature. It can be counted for recording and analysis and is measured strictly by either conforming or not. Therefore, a specification is imbedded in the response as to whether or not a criteria was met. Attribute data (also called “discrete” data) is a characteristic that may take on only one value (0 or 1, for example) and be counted in discrete units such as items or and yes/no options. Discrete examples include binary responses, such as “yes/no” or “pass/fail,” and counted data, such as number of defects. Attribute data is used to create p-charts (percent chart), np-charts (number of affected units chart), c-charts (count chart), u-charts (counts-perunit chart). Figure C-16 summarizes the breakdown of data types into the specific control chart produced. In comparison, variable data is more informative than attribute data, as attribute data is more limited in describing a process. Plus, the sample size required for attribute data must be larger than an equivalent variable data measurement to be statistically significant. In the context of data type, control charts measure variation between samples for attribute data and variation both between and within sample groups over time for variable data.

Control Charts—7QC Tool

221

Data type determines Control Chart to be used

Variable Data

Attribute Data

Measured and plotted on a continuous scale (e.g. time, temperature, cost figures)

Count or Classification data plotted as discrete events (Classified as % waste, errors, etc.)

A B

Sample Size of 1

Fixed Sample Size (small 3-5)

X Chart and R (Average and Range)

Variable Sample Size (large usually > 10)

X Chart and S (Average and Standard Deviation)

Count (Defects, Incidents or Nonconformances)

Failure to meet 1 of the acceptance criteria. (Note: a defective unit might have multiple defects.)

Classification (Defectives or Nonconforming Units)

An entire unit fails to meet the acceptance criteria, regardless of the # of defects on the unit.

C D E F G

Fixed Opportunity (Constant Sample Size)

Variable Opportunity (Variable Sample Size)

Fixed Subgroup Size (Constant Sample Size)

Variable Subgroup Size (Variable Sample Size)

C Chart (Number of Incidents)

U Chart (Incidents per Unit)

NP Chart (Number of Defectives)

P Chart (Percent Defective)

Figure C-16: Summary Tree of Different Control Charts

H I J K L M

Guidelines or tests exist to identify unexpected patterns in data, wherein it is unlikely that these patterns occurred by chance alone. These special cause guidelines presume the observations are independent. Independence means the value of the given data point is not influenced by the value of another data point. If data are not independent, the data values will not be random. This means the rules for determining special cause variations cannot be applied (because they are based on rules of statistical probability). Different statistical authorities vary slightly on the exact guideline details; however, in general there are eight tests for special cause events. MINITAB’s guidelines cover the following scenarios:

N O P Q R S T U V

1. 1 point beyond control limit (3σ)—Detects a shift in the mean, an increase in the standard deviation (σ), or a single aberration in the process. Check your R-chart to rule out increases in variation.

W

2. 9 points in a row on one side of the mean—Detects a shift in the process mean.

Y

3. 6 points in a row steadily increasing or decreasing—Detects a trend or drift in the process mean. Small trends will be signaled by this test before the first test. 4. 14 points in a row alternating up and down—Detects systemic effects, such as two alternately used machines, vendors, or operators.

X Z

222

Encyclopedia

5. 2 out of 3 points in a row @ 2 σ or beyond—Detects a shift in the process average or increase in the standard deviation (σ). Any two out of three points provide a positive test. 6. 4 out of 5 points @ 1 σ or beyond—Detects a shift in the process mean. Any four out of five points provide a positive test. A B C D E F G H I J K

7. 15 points in a row on both sides of the centerline within the 1 σ “zone”—Detects stratification of subgroups—appears when observations in a subgroup come from sources with different means. 8. 8 points in a row on both sides of the centerline all beyond 1σ zone—Detects stratification of subgroups when the observations in one subgroup come from a single source but subgroups come from different sources with different means. MINITAB indicates the special cause pattern by plotting the unexpected event in red and places a footnote alongside the red data point to reference the specific rule. MINITAB provides the specific rule information about a given special cause event, which can be found on its session window. Figures C-17 and C-18 illustrate these different special cause guidelines.

L M N O P Q R S T U V W X Y Z

Figure C-17: Interpreting Control Chart Patterns for Variable and Attribute Data

Control Charts—7QC Tool

223

A B C D E F G H I Figure C-18: Interpreting Control Chart Patterns for Only Variable Data

J K L

How to Use the Tool or Technique When creating control charts, they can be drawn by hand on graph paper using special control limit factors unique to what is being plotted—or created using statistical software. For illustration purposes, MINITAB will be used in this section to plot each of the different types of control charts.

M

Individuals Control Chart The Individuals control chart is one of the most commonly used tools, and variable data is used to construct it. It can go by several names other than Individuals chart—I-MR, Individuals-Moving Range, X-MR, or chart of individuals. I-MR charts are used in the following scenarios:

Q

N O P R S T U

• Low volume manufacturing process (such as commercial airplanes)

V

• Batch-type processes (such as producing a vaccine)

W

• Infrequently reported transactional processes (such as monthly financials) Scenarios wherein rational sub-grouping exists, I-MR charts do not apply. I-MR charts assume that the observations are independent from one another. If the data are not normal, they need to be transformed using a constant such as lambda. MINITAB applies the Box-Cox as the default transformation function to the non-normal by simply selecting the utility. It is a good chart to use to gain a quick understanding of the process to help support improvement and control efforts.

X Y Z

224

Encyclopedia

The example scenario, illustrating the Individuals control chart, involves a company monitoring its invoice payments. In this case, a random sample of one paid invoice each day was selected and identified the number of days from when the invoice was sent to the customer. A B C D

Note In this example, smaller is better; the fewer days between the invoice paid, the better the invoicing company’s cash flow.

E F G H I J

Given that variable data has been entered into the MINITAB Worksheet, the command to access the Individuals charts from its drop-down menu is Stat > Control Charts > Variables Charts for Individuals > I-MR…. Figure C-19 displays sample MINITAB screens where the appropriate variable data set is selected (Area 1) and x-axis labels are identified (Area 2) to produce the final Individuals control chart in Figure C-20.

K L M N

Area 2 Area 1

O P Q R S T U V

Figure C-19: Example MINITAB I-MR Main Screen

W X Y Z

Notice that the top graph plots the individual data points, in this case the days it took for an invoice to be paid over time. The mean (denoted as x-bar) is 31.8 days, and the UCL (upper control limit) is 64.15 days, and the LCL (lower control limit) is -0.55 days. There is one data point out of control, as indicated by the point above the UCL with “1” footnote beside it. The special cause guideline that a MINITAB reference is “1 Point Beyond Control Limit (3σ): Detects a shift in the mean, an increase in the standard deviation (σ), or a single aberration in the process. Check your R-chart to rule out increases in variation.” Figure C-21 provides this example’s MINITAB session window.

Control Charts—7QC Tool

225

A B C D E Figure C-20: Example MINITAB I-MR Control Chart

F G

It is best to determine whether a process is in or out of control by checking the moving range chart because it is the average moving range (MR-bar) that defines the width of the control limits for both graphs. The bottom graph displays the “moving range” (MR), which is the difference between two consecutive individual data points, hence the MR graph has one less data point than the Individuals chart. The average moving range (denoted as MR-bar) is 12.16 days, and the UCL is at 39.74 days, and the LCL is at 0 days. There are two data points out of control (calculated by the absolute difference between the one out of control individual data point and those points collected on either side of it). They are indicated by the two points above the UCL with “1” footnote beside them—the same rule referenced in the top chart. Figure C-21 provides this example’s MINITAB session window.

H I J K L M N O P Q R S T U V W X

Figure C-21: Example MINITAB I-MR Session Window

Recalculated Control Limits Control limits may be recalculated if and only if the process was intentionally modified; otherwise, the control limits should hold constant. If that is the case, the data collection sheet used to monitor the process should indicate when the special cause event occurred—the data of the changed process.

Y Z

226

A B C D E

Encyclopedia

If this example scenario were unacceptable and an improvement to enhance cash flow were implemented, the control chart should indicate a “special cause” event because the process was changed. In this example, let’s say that the process improvement was a prominent “reminder” notice of the 45-day policy on the subsequent invoices after the March timeframe. Follow the same procedure just detailed, using the following MINITAB drop-down menu sequence: Stat > Control Charts > Variables Charts for Individuals > I-MR… . Next, select the I-MR Options button and click the Stages tab, as shown in Area 3 of Figure C-22. The resulting Individuals control chart with the recalculated control limits is illustrated in Figure C-23.

F G

Area 3

H I

Area 1

J K L M N O P Q

Figure C-22: Example MINITAB I-MR Main Screen and Stages Tab

R S T U V W X Y Z Figure C-23: Example MINITAB I-MR Control Chart with Recalculated Control Limits

Did the process improve, knowing that smaller is better? Notice that both the top and bottom graphs have a vertical dashed line to indicate where the process change occurred. Both of their means and the control

Control Charts—7QC Tool

limits became smaller. In the top graph, the mean fell to 23.31 days (from 31.8), the UCL decreased to 54.19 days (from 64.15), and the LCL at -7.57 days (from -0.55). In the new period to the right of the dashed line, no points are identified as out of control. Similarly, in the bottom graph, the average moving range (MR-bar) fell to 11.61 days (from 12.16), the UCL at 37.93 days (from 39.74), and the LCL constant at 0 days—again with no points out of control. In conclusion, the improvements successfully reduced cash flow on average by 8.49 days.

Normal Versus Non-normal Data—Normal Probability Plots Individuals control charts assume that the data are distributed normally; otherwise, it must be transformed. There are several ways to check if the data distribution is normal: 1. Visually, by examining a frequency plot such as a histogram, Dotplot, or Boxplot. Figure C-24 illustrates a normally and non-normally distributed histogram. A definitive conclusion using only visual graphs is difficult. (See Also “Graphical Methods,” p. 323) Visual check of Normal versus Non-normal Histogram looks fairly normal

Histogram clearly not normal

227

A B C D E F G H I J K L M N O P Q R S T

Figure C-24: Histograms of Normal and Non-normal Data

2. Statistically, by running a normal probability plot. If data are normal then a normal probability plot will show a linear relationship. Using MINITAB to create a normal probability plot, simply select the following commands from its drop-down menu: Graph > Probability Plot > Single… Select the data to be analyzed and place it in the dialog box. Click OK. The resulting graph may look similar to either half of Figure C-25—showing normal data on the left and nonnormal data on the right of the figure.

U V W X Y Z

228

Encyclopedia Normal data: Plot is linear

Non-normal data: Plot is not linear

P-value is > 0.05 and > 95% of the data lies within the confidence bounds suggesting the data is normal.

P-value is < 0.05 and < 95% of the data lies within the confidence bounds suggesting the data is not normal.

A B C D E F G

Figure C-25: Histograms of Normal and Non-normal Data

H I J K L M N O P Q R S T U V W X Y Z

Transformation of Non-normal Data (Plus Normal Probability Plots) The control limits on an Individuals charts are based on a normal distribution. If the data are not normal, then the control limits are not appropriate for assessing special cause variation. A transformation of the data usually will correct non-normality (symmetry) issues. Control limits can then be used to assess special cause variation patterns in the transformed data. Hence, non-normal data must be transformed before constructing an Individuals control chart. A generic transformation method, called the Box-Cox (power) transformation, changes the shape of the exponential data to look normal. The form of the transformation is YT = YΛ, wherein the constant of lambda (Λ) is used. MINITAB applies the Box-Cox, as the default transformation function, to find the optimal value of lambda (Λ) that minimizes the variability in the data. The MINITAB applies Box-Cox method from several areas within the application software. It can be accessed directly through the main menu drop-down: Stat > Control Charts > Box-Cox Transformation… Another way to access the Box-Cox method is within the control chart area of the software with the following the drop-down procedure: Stat > Control Charts > Variables Charts for Individuals > I-MR…. Next, select the IMR Options… button and click the Box-Cox tab, (the tab just to the right of the Stages tab shown in Area 3 of Figure C-22). Two sample Individuals control charts, pre- and post-Box-Cox transformation, can be found in Figure C-26. Notice how the non-normal, non-transformed chart on the right shows the data “squished” toward the top of the graph—the data is skewed left. After the transformation, the data looks more randomly and evenly distributed (without a pattern) about the mean.

Control Charts—7QC Tool Non-normal data before Box-Cox transformation

229

Non-normal data after Box-Cox transformation

A B C D Figure C-26: Comparison of Non-normal Data not Transformed and Transformed Using Box-Cox

E F

X-bar and R Chart Variable data create the X-bar and R charts to comprised of a paired set of averages and range control charts. Typically these charts are used for highvolume processes, wherein sampling occurs on a regular time-based interval. The averages data is derived from a subgroup sampling scheme, such as a sample of two to five items taken at regular intervals, and the average of that subgroup represents a control chart data point. The criteria for sampling frequency focuses on ensuring the sample consists of homogeneous items (low variability) within the sample and allows for greater variability from sample to sample.

G H I J K L M N O

Note

P

The Central Limit Theorem purports that “averages” vary less than indi-

Q

vidual data points and tend to be normally distributed as the sample

R

size increases. A distribution of the averages of sampling subgroups

S

exhibits a tighter spread around the mean (standard deviation) than a

T

distribution plot of the individual data points. Hence, testing for nor-

U

mally distributed data is unnecessary for this type of control chart.

V W

The X-bar and R chart works for both normal and non-normally distributed data, wherein the process contains little to no changes and produces the data frequently. Manufacturing is a common process for X-bar and R charts since a sample of a few items can represent several hundred pieces. Given that variable data has been entered into the MINITAB Worksheet, the drop-down menu sequence of commands is: Stat > Control Charts > Variables Charts for Subgroups > Xbar-R… Figure C-27 displays a sample MINITAB screen where the appropriate variable data set is selected (Area 1) to produce the final X-bar and R control chart in Figure C-28.

X Y Z

230

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

The conclusions from studying Figure C-28 might entail the fact that the process appears stable and random—no apparent special causes. Hence, a capability analysis could be conducted next to compare the process with the customer specifications.

Area 1

X-bar and S Chart The X-bar and S chart is similar to the X-bar and R charts except Figure C-27: Example MINITAB X-bar and R Main that the sample data’s stanScreen dard deviation is plotted, instead of its range, to assess the sample to sample variability. These charts frequently are used when subgroup size is larger than the X-bar and R scenarios (that is, the sample size (n) is greater than five ( > 5)). When the sample size is ten or more, the range is no longer an efficient statistic for measuring variability; hence, the less Figure C-28: Example MINITAB X-bar and R Control Chart sensitive standard deviation calculation is used. The X-bar and S chart work for both normally and non-normally distributed variable data. It is commonly used to quickly detect very small process changes. Healthcare scenarios often call for utilization of X-bar and S charts.

Area 1

Given that variable data has been entered into the MINITAB Worksheet, the Figure C-29: Example MINITAB X-bar and S Chart Main Screen drop-down menu sequence of commands is: Stat > Control Charts > Variables Charts for Subgroups > Xbar-S… Figure C-29 displaysa sample MINITAB screen where the appropriate variable data set is selected (Area 1) to produce the final X-bar and S control chart in Figure C-30.

Control Charts—7QC Tool

231

A B C D E Figure C-30: Example MINITAB X-bar and S Control Chart

Figure C-30 displays a general, fairly random pattern. The sample sets two and eight in the top charts may require some investigation to verify the lack of any special cause event.

Attribute Control Charts Attribute data contains less information and less granularity than variable data. Attribute data defines if a product is working or not working, but it fails to describe the magnitude of the failure. Attribute data may describe that the battery life failed, but it lacks the discriminating information as to how far from target. Moreover, attribute data represents “after-the-fact” information, causing the process players to be reactive. If an item fails to meet specifications, it needs to be scrapped or reworked. When possible, move upstream in the process to try to identify leading indicators (variable data) that enable proactive work to prevent failed products before they occur. Only four of the eight guidelines to determine special cause variation are applicable for attribute control charts. Figure C-17 displays the appropriate four tests for attribute data. Attribute control charts only feature one graph and do not require normal data. Hence, MINITAB’s attribute control chart screens lack the final four special cause tests and the BoxCox transformation option. There are two types of control charts developed from attribute data. Data depicting the number or percentage of completely “bad” (or good) units—known as defective or non-conforming units—make up the first category to construct either the p or np charts. A defective is a unit that fails to satisfy requirements due to one or more defects. Hence, the entire product is classified as “good” or “bad,” and “rejected” or “un-shippable” describes a bad unit. These type of charts are based on a binomial distribution—a two-state scenario with constant probability of either state occurring.

F G H I J K L M N O P Q R S T U V W X Y Z

232

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

The second type of attribute control chart is defined by collected data about flaws or defects when the unit (as a whole) is still “acceptable.” The c-chart and u-chart make up this category of defect attribute control charts. A defect is a single non-conforming quality characteristic. Hence, the data may represent a folded page of a book coming off the printing production line, but the entire book is classified as “shippable.” These type of charts are based on a Poisson distribution—describing the area of opportunity (time or space, for example). The Poisson distribution represents fairly large total populations (measured as discrete data) that are difficult or impossible to count and models random occurrences over time, distance, area, and volume. Figure C-31 summarizes the scenario that requires which attribute chart.

Fixed Sample Size

Varying Sample Size

Supported Distribution

c u Poisson p-Chart Defects The p-chart uses binary np p Binomial attribute data to monitor Defective defective units. It is the most sensitive attribute con- Figure C-31: Summary of Attribute Control Charts trol chart. The data may contain pass/fail of a test or errors detected or not detected. The p-chart plots the proportion (p) of data for either aspect (passed or failed) of criteria; hence the chart is based on the binomial distribution. A p-chart can accommodate unequal group sizes of non-conforming or defective items. Often the data is collected as a percentage or fraction of the total lot and may contain multiple types of nonconformity. The data collection sheet needs to identify the different types of non-conformity to better determine the cause(s) of variation.

Given that variable data has been entered into the MINITAB Worksheet, the drop-down menu sequence of commands is: Stat > Control Charts > Attribute Charts > P… Figure C-32 displays sample MINITAB screen where the appropriate attribute data set is selected (Area 1) to produce the final p-chart in Figure C33.

Area 1

Figure C-32: Example MINITAB p-Chart Main Screen

Control Charts—7QC Tool

233

A B C D E Figure C-33: Example MINITAB p-Chart

Figure C-33 illustrates the classic “city skyline” of the unequal sample size proportional display of a p-chart. Notice the vertical axis is labeled “proportion” because the chart uses percentage data to accommodate the unequal sample sizes. The changing limits is also a function of the changing sample size. Statistically there is more confidence with more data, thus tighter control limits. For smaller sample sizes, the confidence decreases; therefore, the control limits loosen (or get larger). Figure C-33 contains four flagged data points that require further investigation.

P Prime The P Prime SPC chart is a variant that basically combines the IndividualMoving Range (I-MR) chart and a p-chart. This chart is used when the pchart produces wrong control limits (that is, compresses) because it is not adequately showing the within-sample variation, typically resulting from a very large sample size with very few defects. This chart corrects the pchart by using the moving range to adjust the control limits with an estimate of the within-sample variation. Dr. David Laney (of Stamford University) created this tool and has written much on the topic.

F G H I J K L M N O P Q R S T U

np-Chart The np-chart uses binary attribute data to monitor defective units. The np-chart uses the same data structure as a p-chart (hence the binomial distribution applies); however, it plots the number of event occurrences rather than proportions. The equation can be written as

W

# occurrences = (subgroup size) x (proportion) or np = n*p

Z

Different from the p-chart, the np-chart’s subgroup size must remain constant. Given that variable data has been entered into the MINITAB Worksheet, the drop-down menu sequence of commands is: Stat > Control Charts > Attribute Charts > NP… Figure C-34 displays sample MINITAB screen where the appropriate attribute data set is selected (Area 1) to produce the final np control chart in Figure C-35.

V X Y

234

Encyclopedia

Area 1

A B C D E

Figure C-34: Example MINITAB np-Chart Main Screen

F G H I J K L M N O P Q R S T U V W X Y Z

Figure C-35: Example MINITAB np-Chart

Figure C-35 illustrates the sample np-chart. Notice that the vertical scale on the graph is count versus the p-chart’s proportion. The attribute data charted in Figure C-35 is Invoice Errors; therefore, smaller is better, and the graph indicates an improvement over time. A question arises as to whether this improvement is by random chance alone. It should be investigated, and if changes have been made, they should be standardized in order to maintain this downward trend.

c-Chart The c-chart monitors the count (c) of defects in a process when an individual item may have multiple defects. The Poisson distribution is the basis for the construction of the chart, assuming an equal number of opportunities for defects must be reasonable; hence, the subgroup size must remain constant. Given that variable data has been entered into the MINITAB Worksheet, the drop-down menu sequence of commands is: Stat > Control Charts > Attribute Charts > C… Figure C-36 displays a sample MINITAB screen where the appropriate attribute data set is selected (Area 1) to produce the final c control chart in Figure C-37.

Control Charts—7QC Tool

Figure C-37 illustrates the sample c-chart. Notice that the vertical scale on the graph is counted similar to the np-chart. There is one point out of control Area 1 in this example, where bad solder joints is plotted. In this case, smaller is better, so the upward trend also may cause concern worth investigating. In addition, the upward trend includes Figure C-36 Example MINITAB c-Chart Main seven points above the mean Screen count, which is approaching the rule of “9 data points above (or below) the centerline” and may indicate a shift change.

u Chart The u-chart uses the same data structure as a c-chart except that the number of defects per unit (u) is plotted instead of the counts of defects. Again, the Poisson distribution is the basis for the construction of this Figure C-37 Example MINITAB c-Chart chart, so it assumes an equal number of opportunities for defects must be reasonable. The u-chart, considering it monitors the Area 1 number of defects per unit, can accommodate unequal subgroup sizes. Given that variable data has been entered into the MINITAB Worksheet, the drop-down menu sequence of Figure C-38: Example MINITAB u-Chart Main commands is: Stat > Control Screen Charts > Attribute Charts > U… Figure C-38 displays sample MINITAB screen where the appropriate attribute data set is selected (Area 1) to produce the final u control chart in Figure C-39.

235

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

236

A B C D E F G H I

Encyclopedia

Figure C-39 shows the sample u-chart whose vertical scale is “count per unit,” as a percent (or proportion) to the given sample size of 50 (as indicated in Figure C-38). As with the c-chart, this example indicates one point out of control, where bad solder joints is plotted. The scenario calls for a “smaller is better,” so the upward trend also may cause concern worth Figure C-39: Example MINITAB u-Chart investigating. In addition, the upward trend includes seven points above the mean count, which is approaching the rule of “9 data points above (or below) the centerline” and may indicate a shift change.

J K

Hints and Tips

L

The process control limits are the only limits that belong on a control

M

chart; not customer specification limits. Control charts should be con-

N

structed real time as data is produced in the process, not from inspec-

O

tion records. Control means the process is consistent, not necessarily

P

that the results are meeting requirements. Statistical control means

Q

there are no special causes affecting a process—as indicated by ran-

R

domly dispersed points within the control limits around the average

S

line. Points within control limits that indicate a trend, shift, or instabil-

T

ity are special causes requiring investigation. Points outside the con-

U

trol limits should be removed from control limit calculations (but still

V

plotted) once cause is identified. Change control limits only when the

W

process is changed for the data collected after the process change.

X Y Z

• Sampling method and plan • To establish inherent variation and allow process to run without sampling. • Determine sample size • Attribute at least 50. (c and u-charts, samples to average 5+ defects) • Variable at least three to five

Control Charts—7QC Tool

237

• Ensure random samples • Consistent conditions (same machine, operator, lot, and so on) when sampling • Collect 20 to 25 different sample groups before calculating control limits. • Patterns determine sampling frequency. If process is in “control,” frequency can be reduced.

A B C D E

Supporting or Linked Tools Supporting tools that might provide input when developing a control chart include • Data collection sheets (See Also “Data Collection Matrix,” p. 248)

F G H I J

• Graphical tools—particularly frequency plots such as a histogram, Dotplot, or Boxplot (See Also “Graphical Methods,” p. 323)

K

• Normal probability plot (Figure C-25)

L

A completed control chart provides input to tools such as • Fishbone diagram (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173)

M N O P

• Root cause analysis techniques (See Also “Hypothesis Testing” and “Regression Analysis” p. 335 and 571, respectively.)

Q

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287)

S

R T

• Brainstorming technique (See Also “Brainstorming Technique,” p. 168)

U

• Concept generation methods

V

• Control plans (See Also “Matrix Diagrams—7M Tool,” p. 399)

W X

• Risk mitigation plan (See Also “Risk Mitigation Plan,” p. 601)

Y

• Transition plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

Z

• Implementation plan (See Also “Matrix Diagrams—7M Tool,” p. 399) • Communication plan (See Also “Matrix Diagrams—7M Tool,” p. 399) • Standard operating procedure (SOP) Figure C-40 illustrates the link between a control chart and its related tools and techniques.

238

Encyclopedia

Fishbone Diagram

Root Cause Analysis Techniques

FMEA

Control Plan

Risk Mitigation Plan

A

Data collection sheet

B C D

Transition Plan Graphics (Frequency Plots)

Control Chart Implementation Plan

E F G H

Normal probability plot Brainstorming Technique

Concept Generation Methods

I

Standard Operating Procedure

J K L

Communication Plan

Figure C-40: Control Chart Tool Linkage

M N

Cost/Benefit Analysis

O P Q R S T U V W

What Question(s) Does the Tool or Technique Answer? What is the payback time period for an investment? A cost/benefit analysis helps you to • Understand if and when net cash benefits are equal to or begin to out-weigh the net costs or outflows. • Evaluate if the project is worthwhile and whether or not to proceed with an investment.

X Y Z

Alternative Names and Variations This tool is also known as • Benefit cost analysis Variations on the tool include • Internal Rate of Return (IRR) • Return on Investment (ROI)

Cost/Benefit Analysis

239

• Return on Assets (ROA) • Net Present Value (NPV)

When Best to Use the Tool or Technique Conducting a cost/benefit analysis is an important management tool to determine whether to proceed with a project and/or an improvement investment. This technique should be utilized at the onset of a project, during its planning stage, and re-evaluated throughout the project lifecycle (including the project close) to monitor the project’s financial performance and healthiness and validate assumptions.

A B C D E F

Brief Description The cost/benefit analysis serves two key purposes: 1) determining financial healthiness of a decision and 2) helping to monitor risk of success and sustainability. The technique compares the cost or expenses associated with a process or project and contrasts it with its respective benefit to the organization. The cost/benefit analysis provides input to go/no go decisions and risk mitigation planning. If the benefits outweigh the costs, it generally leads to a “go forward” decision. Conducting a cost/benefit analysis requires input from those familiar with the process, often supported by a finance expert if not an active member of the project team. The data may include hard and/or soft costs, which should be categorized based on accounting principles employed by the organization. Hard costs and savings easily are associated with money and are evident on the bottom-line of a profit and loss (P&L) statement. Soft costs or savings imply financial benefit and can be quantifiable in monetary terms but have a more nebulous impact on the bottom-line P&L. Soft benefits include time freed-up to do other activities and improved satisfaction and morale. However, depending on the root cause of a problem, some improvements such as improved organization or tidying-up may be a soft or hard benefit, given that those improvements are part of the Lean 5S strategy and may have direct positive safety or waste reduction impacts that are traceable to the bottom-line. (See Also, “5S,” in Part I “Lean and Lean Six Sigma,” p. 29) Thus, categorizing the hard and soft money should be done in conjunction with the organization’s financial expert(s).

How to Use the Tool or Technique Conducting a cost/benefit analysis generally consists of the following steps:

G H I J K L M N O P Q R S T U V W X Y Z

240

Encyclopedia

Step 1.

a. Document assumptions about the project/investment. They may include current problem, its metrics, the period of time involved, the root causes, any boundary conditions, the proposed improvements/investments, and expected impact (both positive and negative, financial and non-financial).

A B C D

b. Gather required cost factors: materials, labor, other resources and the cost of capital (interest rate or financing charge) if an investment is under consideration and identify when they will be incurred, as a one-time event or ongoing.

E F G H

c. Identify the benefits as monetary, time, or other and identify when they will be realized, as a one-time event or ongoing.

I J K L

Step 2.

Using the cost and savings factors, develop a calendar that schedules the net project gain (or loss) and calculate appropriate financial ratio(s).

Step 3.

Decide whether to proceed with or kill the project/investment or make appropriate modifications to the assumptions and recalculate the financial ratios.

M N O P Q R S T U V W X Y Z

Gather and categorize the financial and non-financial impact of the project or potential investment as to the potential costs and savings.

The various financial ratios used in a cost/benefit analysis include the following calculations. Consult with the organization’s financial expert to assist in the selection of the appropriate ratio.

Return on Assets (ROA) ROA = Net Income divided by Total Assets, where net income equals the expected revenue or earnings. Return on Investment (ROI) ROI = Net Income divided by Total Investment, where net income equals the expected revenue or earnings. Net Present Value (NPV) NPV = Sum of the ratios of cash flow in a time period divided by one plus the cost of capital for that same time period. The NPV formula is NPV


Σ ( (1  r)t t ) t
0 n

CF

Cost/Benefit Analysis

241

Where “n” is the number of time periods, “t” is the specific time period, “r” is the cost of capital (interest rate) for the given time period, and “CF” represents the cash flow in that time period.

Internal Rate of Return (IRR) The IRR is used to evaluate a portfolio of projects. The IRR of one project (or opportunity) is compared with other projects. The go/no go decision for a project is based on the relative IRR ratios among a portfolio of opportune projects; wherein the larger IRR is preferred. The IRR calculation defines the cost of capital (“r”) at the point when the NPV equals zero. The exact calculation is an iterative process to determine the exact IRR. The IRR formula is IRR
NPV @ 0


Σ ( (1  r)t ) n

CFt

t
0

A B C D E F G

Where “n” is the number of time periods, “t” is the specific time period, “r” is the cost of capital (interest rate) for the given time period, and “CF” represents the cash flow in that time period.

H

Payback Period This is a simple, thus commonly used cost/benefit calculation. It defines the fulcrum point at which the cash flow becomes positive, when the net cash benefits (inflows or revenues) outweigh the net costs (outflows).

K

The calculation for the breakeven point usually ignores the time value of money and simply determines the cumulative net cash flow over time: Payback = Investment divided by Cash Inflow Where the investment is both any initial one-time payment and any incremental ongoing outlays of cash. The cash inflow is any resulting savings or revenues on a one-time and ongoing basis.

I J L M N O P Q R S T

Supporting or Linked Tools Supporting tools that might provide input when developing a cost/benefit analysis include • Data collection sheet (See Also “Data Collection Matrix,” p. 248) A completed cost/benefit analysis provides input to tools such as • Project charter (See Also “SMART Problem and Goal Statements for a Project Charter” for more on project charter, p. 665) • Phase-gate reviews • FMEA and risk mitigation planning (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) • Control plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

U V W X Y Z

242

Encyclopedia

Figure C-41 illustrates the link between a cost/benefit analysis and its related tools and techniques.

Project Charter

A B C D

Phase-gate Reviews Data collection sheet

Cost / Benefit analysis FMEA

E F G

Control Plan

H I

Figure C-41: Cost/Benefit Analysis Tool Linkage

J K L

Critical Path Method (CPM)

M

See Also “Activity Network Diagram (AND)—7M Tool,” p. 127

N O

Critical-to-Quality (CTQ)

P Q S

What Question(s) Does the Tool or Technique Answer? How does my work relate to the customer requirements, and how do I know when I have fulfilled them?

T

CTQ helps you to

R

U V

• Understand customer general requirements in more specific terms.

W

• Translate customer requirements into specific, actionable, measurable language for those who work in the process (process worker, for example).

X Y Z

When Best to Use the Tool or Technique The Critical-to-Quality (CTQ) elements of a customer requirement should be defined as early in the project as possible, immediately following the Voice of the Customer (VOC) gathering activities.

Critical-to-Quality (CTQ )

Once customer input has been gathered and categorized, the requirements should be translated into specific, measurable terminology, referred to as a CTQ. CTQ can be used to evaluate how well the current process works or how well the current deliverables meet customer requirements. CTQ also can be used to select improvement options, such as process improvements or new features/functionality.

243

A B

Brief Description CTQ is a simple, yet powerful tool that translates customer needs into a meaningful, measurable, and actionable metric for the person or persons doing the work needed to deliver the requirement. Customers typically describe their requirements using vague words or generalities. In addition, customers’ descriptions may or may not mirror one another’s value system or criteria given a strong likelihood of different perspectives. In the example of prepared food, the requirement of food to “taste good” may not be perceived the same from person to person. If the company making prepared food wants to meet customer requirements, it needs to understand the characteristic of “tastes good” and its qualities. To do so, the company needs to drill down and dissect its meaning and interpret the quality of “tastes good” until it arrives at meaningful and actionable terms for the process of cooking, packaging, and delivering the food to the marketplace. To eliminate ambiguity, CTQs translate customer needs into internally meaningful, specific, and measurable terms . Understanding and interpreting a customer’s requirements is important because they define or help identify the important elements to satisfy their own needs—that which is critical to providing or delivering the quality they expect. CTQs serve as a bridge between the internal process and its deliverables and customer satisfaction. Therefore, an accurate translation is critical to understanding the customer’s perspective. A fully developed CTQ has four elements: 1) output characteristics, 2) output metric (or “Y”), 3) target value, and 4) specification/tolerance limit. The inclusion of these four elements eliminates ambiguity around the interpretation of the customer needs to help the business achieve customer satisfaction. In a Six Sigma project, essentially there may be two types of CTQs. The first and most prevalent type addresses the process and its outputs that the project aims to improve. The second type of CTQ describes what must be done to meet the requirements of the project itself. This is a onetime set of CTQs needed to accomplish the specific project deliverables and may or may not exist. Ultimately, a CTQ translates a VOC requirement into a metric that is actionable by either a process player or project team member.

C D E F G H I J K L M N O P Q R S T U V W X Y Z

244

Encyclopedia

How to Use the Tool or Technique Developing a Critical-to-Quality metric is fairly simple. The key is to arrive at a measure that is meaningful and actionable. Follow the subsequent procedural guidelines: Step 1. A

Gather the customer needs as verbatim input (using their language). If a large number of requirements exist, often it helps to organize the VOC input into an Affinity diagram. (See Also “Affinity Diagram—7M Tool,” p. 136)

B C D E

Step 2.

F H I J

• Continue to break down and translate the VOC need until a meaningful and actionable measure exists for each of the process players involved in delivering that requirement. Stop when the last level of detail is measurable.

K L M N

• Repeat this step until a CTQ is aligned with each VOC requirement or category.

O P

• A given CTQ can be aligned with multiple VOC requirements.

Q R

• A given VOC need can elicit multiple CTQs to address multiple dimensions or multiple process player roles.

S T

Step 3.

Organize and document the drill-down work into a matrix or a Tree diagram, linking the CTQ to the original VOC requirement.

Step 4.

Validate the CTQs with the customer to ensure proper understanding and translation of the original requirement.

V W X Y

Translate the customer need into a CTQ. • Addressing one customer need or category at a time, begin to parse it down into more specific meaningful descriptions, translating what is critical to the customer ultimately into an actionable metric for either a process player or project team member.

G

U

Identify the customers and their requirements.

Z

Examples CTQ Tree Figure C-42 illustrates a CTQ tree for a restaurant patrons’ food requirements. Notice that one customer requirement, “Good Food,” generated two CTQs: “order taken correctly” and “cooked correctly.” Each of those CTQs link to a different process player. The order taking CTQ links to the wait staff, whereas the cooking properly CTQ links to the chef (or kitchen staff).

Critical-to-Quality (CTQ ) BASIC NEED

1st LEVEL

Accuracy

Good Food

Timeliness

2nd LEVEL

MEASURE

Order Taken Correctly

Yes / No

Cooked Per Request

Yes / No

A B

Order to Arrival Time (seconds) Freshness

245

Scale: 1 (stale) – 10 (fresh)

C D

General Need

Behavioral (Requirement)

E F

Figure C-42: Sample CTQ Tree

G

CTQ Matrix Figure C-43 illustrates another example of a restaurant patron’s food requirements displayed as a CTQ matrix.

H I J K

VOC

Issue Translation

Key Item

CTQ

Measure

L

Cooking

Temperature

Temperature when fully cooked

Fahrenheit

M

Delivery

Quick Service

Transfer Time from kitchen to table

seconds

I like my food hot.

N O P Q

Figure C-43: Sample CTQ Matrix

R S

Hints and Tips

T

Start with customer needs. Use nouns without adjectives; adjectives

U

can introduce ambiguity, allowing for multiple interpretations or per-

V

spectives (that is, how hot is hot; what is good; how blue is blue).

W

Using a CTQ tree approach, the branches to the right show increasing

X

detail of that requirement, not a new requirement. The number of lev-

Y

els (in a matrix) or branches (in a tree) needed to define a CTQ

Z

depends on how specific the requirement is to the required work. For example, if the requirement is an internal customer requirement from an area manager, more than likely, the resulting CTQ would require

continues

246

Encyclopedia

less granular parsing to arrive at something meaningful, measurable, and actionable, than compared with a requirement from the external customer. A given CTQ can be aligned with multiple VOC requireA B C D E F G H I J K L M N O P Q R S T U V W X

ments. A given VOC requirement may elicit multiple CTQs to achieve and address multiple dimensions or multiple process player roles, as illustrated in Figure C-42. CTQs centered around the process or product and/or services offering ultimately should be aligned to a process player’s role and reflected in commensurate performance documents such as personnel job appraisal and standard operating procedure (SOP). Process-specific CTQs: A CTQ developed for a process or product becomes part of standard work wherein the generic project requirements translate into something meaningful, measurable, and actionable for the process player and can become part of his/her job performance metrics. Process CTQs are relevant until customer requirements or the process is changed. An improvement project may aim to define, clarify, or refine the process CTQs as part of its improvement deliverables and incorporate them as part of the transition and control plan. Project-specific CTQs: A CTQ can be developed for a project (for example, an improvement project), wherein it is achieved by completion of the project, thereby satisfying a customer requirement. Hence, the generic project requirements translate into something meaningful, measurable, and actionable for the project team member. Upon project completion, that CTQ is no longer relevant. While on a project, if both process and project-specific CTQs exist, keep them separated to avoid confusion.

Y Z

Supporting or Linked Tools Supporting tools that might provide input when developing a CTQ include • VOC and VOB data (See Also “Voice of the Customer Gathering Techniques,” p. 737) • Current Process map (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

Critical-to-Quality (CTQ )

247

A completed CTQ matrix or tree provides input to tools such as • QFD (See Also “Quality Function Deployment (QFD),” p. 543) • Root cause analysis techniques • Concept generation methods

A

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287)

B

• Standard operating procedure (SOP)

D

Figure C-44 illustrates the link between a CTQ and its related tools and techniques.

C E F G

QFD

H I

Root Cause Analysis Techniques

VOC and VOB Data

J K L

CTQs

Concept Generation Methods

Process Map FMEA

M N O P Q

SOP

R S

Figure C-44: CTQ Tool Linkage

T U V W X Y Z

248

Encyclopedia

D Data Collection Matrix A B D

What Question(s) Does the Tool or Technique Answer? What information is available about the current process, product, or services offering?

E

A data collection matrix helps you to

C

F G

• Plan and organize current data sources and collection plan.

H

• Identify what data is needed, what is available, where to get it, and who is responsible to get it.

I J K L M N

Alternative Names and Variations Variations on the tool include • Check sheets or checklist (See Also “Checklists—7QC Tool,” p. 204)

O P Q R S T U V W X Y Z

When Best to Use the Tool or Technique Develop a data collection matrix during both the project planning phase to decipher what data is needed and what is available, as well as the data collection phase to monitor the activity.

Brief Description A data collection matrix is often overlooked because it is a simple tool, and its value goes unnoticed. As is true with all matrix tools, the data collection tool power unveils itself when used in concert with other tools. Use the matrix to plan and organize the data collection process for the project. Near completion of the project, the project team may need to transition the responsibilities to the process players for ongoing operations. If so, then the data collection matrix describes the critical variable to monitor. Hence, the document serves as part of the project’s control plan deliverable. A data collection matrix should be customized to the unique needs of the organization and the type of data needed to understand its critical parameters. However, a fundamental structure provides a good starting point from which to modify. The tool provides a roadmap as to what data is required and organizes that which is collected to ensure nothing is missing or there’s no duplicity. At a minimum, the matrix should identify the following items:

Data Colle ction Matrix

249

• The metric of interest (the CTQ) • The official source of the data to ensure integrity and avoid duplications or voids in the data sources • The supplier of the data, who measures it, and how A

• Specifications or targets and/or operational definition The data collection matrix is the “master plan” for a process’ data collection (and/or sampling plan). It identifies the various sources, including supporting tools that contain the actual data of interest. Another related but separate tool is the actual data collection sheet. This document is used to gather and record the data collected and often is referred to as a “checklist.” Similar to the data collection matrix, the checklist should be tailored to the work scenario and becomes the basic elements of a “frequency” checklist. (See Also “Checklists—7QC Tool,” p. 204)

B C D E F G H I J

Examples Data Collection Matrix Figure D-1 shows a sample data collection matrix template.

K L M

Data Collection Matrix

N

Process Name: Process Owner:

Prepared by:

Page:

Approved by:

Document No:

Location:

Approved by:

Revision Date:

Area:

Approved by:

Supersedes:

Int/Ext

Contact Info:

Process Step

CTQ Metric KPIV

KPOV

Data Type

Operational Definition

Specification/ Requirement USL

LSL

Measurement Method (how measured and collected)

Sample Size

Frequency

Who Measures (supplier)

Data Source Reporting Method and Location (System, location, and Customer owner)

of

O P Q R

SOP Reference

S T U V W X Y Z

Figure D-1: Sample Data Collection Matrix

250

Encyclopedia

Supporting or Linked Tools Supporting tools that might provide input when developing a data collection matrix include • Standard operating procedure (SOP) A B

• Process maps (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

C

• CTQs (See Also “Critical to Quality (CTQ),” p. 242)

D E

A completed data collection matrix provides input to tools such as

F

• Data collection checklists (See Also “Checklists—7QC Tool,” p. 204)

G

• Graphical tools (See Also “Graphical Methods,” p. 323)

H I

• Process maps (See Also “Process Maps,” p. 522)

J

• Control plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

K

Figure D-2 illustrates the link between a data collection matrix and its related tools and techniques.

L M N O

Data Collection Checklists

Standard Operating Procedure

P Q

Graphical Tools Process Maps

Data Collection Matrix

R S T

Process Maps CTQs

U V W

Control Plan

Figure D-2: Data Collection Matrix Tool Linkage

X Y Z

Design of Experiment (DOE) What Question(s) Does the Tool or Technique Answer? Which variables or combination of variables proves the best in producing the desired results, based on experimental data?

De sign of Experiment (DOE)

251

DOE helps you to: • Identify which factors are most important in producing the desired outcome. • Determine the optimal “setting” for the critical factors.

A B

Alternative Names and Variations This tool is also known as: • D of E

C D E F

• Designed experiments

G

Variations on the tool include

H

• Conjoint Analysis (See Also “Conjoint Analysis,” p. 207)

I J K

When Best to Use the Tool or Technique The scientific investigative nature of a DOE provides actual information about a causal relationship in a more efficient and effective manner than one factor at a time, also known as OFAT [pronounced “O-fat”]. A DOE is useful when selecting the key factors affecting the desired outcome when trying to choose between alternatives and when trying to model a system to identify its optimal settings. Within a Six Sigma project, the DOE technique helps to answer several questions and thus has several applications. The DOE approach helps to identify the most important variables needed to produce a desired outcome. Typically, a DOE can be used during the design and optimize phases of a DFSS project, and the analyze and improve phases of a DMAIC project, depending on the key business question being asked.

L M N O P Q R S T U V W

Brief Description Design of Experiments (DOE) is a structured and efficient approach to investigate Cause-and-Effect relationships between input variables (referred to as “X”s) and outputs (referred to as “Y”s) within a process. George Box claimed that “to determine what happens with a process when you interfere with it, you have to interfere with it, not passively observe it.” In other words, to understand how best to produce the desired results, “adjust” the critical variables in a system to understand the consequences or impact of the action.

X Y Z

252

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

Through careful planning, a well-constructed DOE tests several factors (Xs) simultaneously. The resulting analysis explains the impact on the desired outcome of not only each individual factor (called a main effect), but also the interaction of the multiple factors (the combined or confounded effect). A test variable that proves not to be highly significant does not mean that it is insignificant, so examining the interaction between variables sometimes reveals important insights. DOE as an investigative technique is more efficient and effective as compared with the “trial-and-error” and the one factor at a time (OFAT) approaches. The DOE technique’s unique features include the following: • It is a balanced experiment allowing multiple factors to be studied simultaneously. • Interactions are comprehended (detected and measured). • Outputs are controlled, minimizing or eliminating the effect of noise factors through blocking techniques and quantifying experimental error. • Fewer experiments are required, using sound scientific and mathematical methodology. The DOE has two primary advantages: • Real-time data from the DOE technique actually demonstrates a Cause-and-Effect relationship—adjust a critical variable (X) to observe the outcome (Y). In the regression model, the technique relies on predicted data (that is, educated guesses) from either historical or happenstance scenarios where the relationship is theorized. • The structured and active experimentation technique from a DOE provides more information about the causal relationship versus a model-building technique using less-structured, happenstance data such as OFAT or trial-and-error.

Alternative Experimental Techniques In a controlled lab environment of a middle and high school science laboratory classroom, with a known (or predetermined) outcome, the teacher wants to demonstrate the response. This is the scenario most people relate to when thinking about experimentation. However, recall the conditions. The discovery process is controlled for the students; the critical variables and the outcome are known (by the teacher and the textbook author); and the time allotted for the procedure is reduced to fit the class period. The common techniques used individually or in combination by classroom

De sign of Experiment (DOE)

teachers include: One Factor At A Time (OFAT), Stick With A Winner, Implement Many Solutions (process), or test at extremes (design). However, in most real-world situations in business, the critical factors may not be known (or isolated), and the Cause-and-Effect relationship might be only a hypothesis (not proven). To fully understand the power of the DOE technique, let us contrast it with these three different alternatives. One Factor at a Time The OFAT technique depends on pure luck to find the optimal combination and settings of multiple factors. It is a classic experimental approach (often taught in secondary school and college science labs), wherein only one factor is the focus, and all else should be held constant. However, in a complicated process, it is extremely difficult, if not impossible, to hold constant. This technique’s shortcomings also include that it requires a significantly large number of experiments to study the effects of all the input factors. Plus, the technique may never reveal the optimum combination of the multiple variables. Moreover, the OFAT cannot distinguish any interaction between factors, uncovering if the behavior and optimal setting of one factor may depend on another. Thus often OFAT conclusions are wrong, misleading, or statistically inconclusive. It produces little usable data, after relatively large commitments of time and effort, and the observed effects often are unexplainable. For illustration purposes, the OFAT experimental scenario involves identifying if two components aid in accelerating the dissolving time of an instant drink mix—temperature and type of liquid. The experimental design examines the temperature at two levels (room temperature and boiling), and two types of liquid (tap water and carbonated water). Figure D-3 displays this scenario for the instant drink mix design and its two adjustable factors; wherein the “-” symbol indicates the base configuration (room temperature and tap water), and the “+” symbol indicates the adjusted factor (boiling temperature and carbonated water). The experiment design adjusts one factor at a time, holding all other factors at the base level. The experiment starts by adjusting the liquid type only and holding the temperature steady at room temperature. Thus, the first test is run with faucet water as the base configuration, and the second test is run with carbonated water. The results showed that the instant drink mix dissolved in 45 seconds in the tap water (Run 1) and 41 seconds in the carbonated water (Run 2). Because the carbonated water dissolution time is faster, the third run maintains the carbonated water and adjusts the second variable using the boiling temperature. The results of Run 3 for carbonated water at a boiling temperature is 47 seconds, slower than Run 2 for carbonated water at room temperature. It is believed that Run 2 produced the fastest dissolving time.

253

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

254

A B C D E F

Encyclopedia

Given time and resource pressures, the experiment stops at this point and concludes that the combination of carbonated room temperature water (Run 2) is the optimal variables setting for the fastest dissolving time. The final combination of tap water at boiling temperature is never tested. If Run 4 were run, the dissolution time would have been 37 seconds, as shown in Figure D-3. The wrong conclusion is drawn. The conclusion depended on which test combination ran first and at what point the experiment ceased. If a scenario involved more than two factors that needed testing, the likelihood of coming to the incorrect conclusion using this technique increases. Water Type

Run 2

Run 3

+1

41

47

Run 1

Not Tested

45

37

Carbonated

G H I

-1 Tap

J K L M

-1

Room

+1

Boiling

Temperature

Figure D-3: Results of the OFAT Instant Drink Mix Example

N O P Q R S T U V W X Y Z

Stick with a Winner For illustration purposes, the Stick with a Winner experimental scenario involves building a paper helicopter to determine which design descends the slowest when dropped from a six-foot high stairwell landing. The design contains multiple components that can be adjusted to obtain the slowest speed. For this scenario, six factors will be tested: the type of paper it is made of, the weight of an added paper clip, the wing length, the wing width, the body length, and the body width. To start the Stick with a Winner technique in this experiment, the paper helicopter’s base construction is noted, and the time it takes to descend is recorded. Next, each component is changed one at a time, and descent time is recorded. If the change produces the desired result, the factor is left at that level; otherwise revert back to the previous setting for the remainder of the experiment. In this example, if the resulting time is slower, then hold (or stick) with the adjustable factor at that level. Figure D-4 displays this scenario for the base helicopter design and its six adjustable factors wherein the “+” symbol indicates the one factor adjustment, and the “-” symbol indicates the base construction.

De sign of Experiment (DOE)

255

Paper Type

Paper Clip

Wing Length

Wing Width

Body Length

Body Width

Base

-

-

-

-

-

-

Paper Change

+

-

-

-

-

-

Add Paper Clip

-

+

-

-

-

-

Wing Length Change

-

-

+

-

-

-

A

Wing Width Change

-

-

-

+

-

-

B

Body Length Change

-

-

-

-

+

-

C

Body Width Change

-

-

-

-

-

+

D

Design

Figure D-4: Paper Helicopter Stick with a Winner Design

E

After running the OFAT experiment, the times were recorded, as shown in Figure D-5. The slowest descending time was 2.7 seconds for the combination of the new paper type and longer wing length. However, notice not all of the different combinations were tested. Based on the previous instant drink OFAT experiment, what is the likelihood that this 2.7 second combination is the optimal one? More than likely, a better combination exists, but we cannot be certain.

F G H I J K L M

Paper Type

Paper Clip

Wing Length

Wing Width

Body Length

Body Width

Flight Time

Base

-

-

-

-

-

-

2.2

Paper Change

+

-

-

-

-

-

2.4

O

Add Paper Clip

+

+

-

-

-

-

1.9

P

Wing Length Change

+

-

+

-

-

-

2.7

Q

Wing Width Change

+

-

+

+

-

-

2.5

R

Body Length Change

+

-

+

-

+

-

2.4

S

Body Width Change

+

-

+

-

-

+

2.5

T

Design

Figure D-5: Results of the Paper Helicopter Stick with a Winner Example

N

U V

Implement Many Solutions or Test at Extremes Often project improvement teams generate multiple solutions to enhance the process and implement these simultaneously for quick results. Given time and resource pressures, these circumstances frequently occur and describe the typical conditions for an Implement Many Solutions (process) or test at extremes (design) situation. Continuing with the paper helicopter scenario from the previous example, the approach would examine the extreme conditions only to select the combination that produces the slowest descent time. Per the design approach, not all the potential combinations would be tested. Regardless of the resulting times, the likelihood of discovering the best configuration is slim with this all-

W X Y Z

256

Encyclopedia

or-nothing approach. Figure D-6 illustrates the test at extremes approach for the paper helicopter example.

Design

Paper Type

Paper Clip

Wing Length

Wing Width

Body Length

Body Width

A

Standard

-

-

-

-

-

-

B

Change All Parameters

+

+

+

+

+

+

C D E F G H I J K

Figure D-6: Paper Helicopter Test at the Extremes Design

DOE Terminology To understand the DOE technique, one first must understand some of its unique concepts and terminology. The DOE terms listed as follows represent the fundamentals, but know that there are whole college-level, semester-long, courses in the DOE techniques, and this entry is just skimming the surface to explain how to conduct some simple DOE tests. The key DOE terminology includes

L

• Alias—A synonym for confounded.

M

• Blocking—Method of negating the effect of uncontrollable factors (such as noise) or factors not of interest. It reduces unwanted variability in the response due to a nuisance variable (a factor impacting the response but not of interest). For example, depending on the business questions some blocking variables may include: shifts, day, location, batch, machine, operator, tenure, or age.

N O P Q R S T U V W X Y Z

• Design Structure • Balanced design—When the test design involves an equal number of runs for each combination of each setting (or level) for each factor. • Orthogonal design (or array)—When the test design reflects a balanced matrix that avoids any confounding of the main effects or any interactions. A full factorial is described as balanced and orthogonal because the design represents an equal number of runs for each level and combinations of all the factors. • Confounded—When the effects of two or more factors are indistinguishable from one another (also called alias).

De sign of Experiment (DOE)

257

• Factor (x)—There are two types of factors (See Also “Y=f(x),” p. 758): • Control Factor—An input of a process having an effect on a response and whose value can be easily selected. • Noise Factor—An input to a process having an effect on a response but is not easily managed. • Factorial Design Structure

A B C

• Full Factorial—A design wherein all the possible combinations of all levels of all factors are tested; none are omitted. A full fractional design for a two-level (indicated by either + or -), three-factor (A, B, and C) is shown here:

D E F G

Run

A

B

C

H

1

-

-

-

I

2

-

-

+

J

3

-

+

-

L

4

+

-

-

M

5

-

+

+

6

+

+

-

P

7

+

-

+

Q

8

+

+

+

This design can be denoted as 23 indicating two-levels for three-factors (2x2x2, for example), represented by a total of eight possible combinations to represent a full factorial. (The notation nk denotes n-levels for “k” factors.) • Fractional Factorial—A design describing fewer combinations than a full design are tested. For example, if a halffractional design for a two-level (indicated by either + or -), three-factor (A, B, and C, for example) would have only four runs (half of the full factorial design 23 = 8 ) of the possible eight total combinations. (The notation nk denotes nlevels for “k” factors.) For example, a half-factorial 23 design might select the odd runs of the full factorial design example:

K

N O

R S T U V W X Y Z

258

Encyclopedia Run

A

B

C

1

-

-

-

2

-

-

+

A

3

-

+

-

B

4

+

-

-

5

-

+

+

E

6

+

+

-

F

7

+

-

+

H

8

+

+

+

I

• Interaction—The effect of a single factor (X) on the output (Y) is dependent on another factor. There are two types of interaction: “synergistic” (mild to moderate and useful in its effect) and “antisynergistic” (strong and disruptive in its effect).

C D

G

J K L

The interactions represented in a full fractional design for a twolevel (indicated by either + or -), three-factor (A, B, and C, for example) is shown here with two-way interactions AB, AC, and BC, and a three-way interaction ABC:

M N O P Q

Note

R

Notice the “+” and “-” signs of the main effects (A, B, or C) are multi-

S

plied together to denote the appropriate interactions (AB, AC, BC,

T

ABC). Recall the rules of multiplying signs: (-) x (-) = (+); (-) x (+) = (-);

U

and a (+) x (+)=(+).

V W X

Run

A

B

C

AB

AC

BC

ABC

Y

1

-

-

-

+

+

+

-

2

-

-

+

+

-

-

+

3

-

+

-

-

+

-

+

4

+

-

-

-

-

+

+

5

-

+

+

-

-

+

-

6

+

+

-

+

-

-

-

Z

De sign of Experiment (DOE) Run

A

B

C

AB

AC

BC

ABC

7

+

-

+

-

+

-

-

8

+

+

+

+

+

+

+

259

• Levels—Settings or value for a given factor. A two-level setting often is indicated by a “+” to represent the high setting and “-” for the low setting.

B

• Main Effect—The effect of a single factor (X) that is independent of any other factors.

D

• Orthogonal Array (OA)—A design that is balanced wherein each factor level appears the same number of times. Originally defined by Taguchi. The simplest is an L4 OA design (four trial runs) for two levels:

F

A C E G H

Column 1

Column 2

Column 3

Trial 1:

+

+

+

J

Trial 2:

+

-

-

K

Trial 3:

-

+

-

M

Trial 4:

-

-

+

N

Notice that each factor (+ and -) appears twice within each column of data. • Randomization—Method of assigning the order in which the treatment combinations are measured. Minimizes the effect (bias) of a “lurking’ (unaccounted for) variable.

I

L

O P Q R S T

• Replication—Repeating all treatment combinations more than once (not repeating a measurement of a single treatment combination). Provides a better estimate of random error.

U

• Response (Y)—A measurable output of a process.

W

• Resolution Level—The amount of confounding found in a design, either main effects with interactions or interactions with interactions, based on the number of factors and levels in a design, wherein the higher the resolution, the less confounding exists. Selecting the appropriate resolution level depends upon the tolerable level of uncertainty about the experimentation balanced against both time and economic constraints on the number of experiments (runs) that can be conducted.

X

Table D-1 summarizes some of the various resolution levels wherein “Full” represents a full factorial design with no confounding. Resolution II (or RII) contains some of the main effects confounding with one another (or 1+1=II), which is undesirable. Resolution III (or RIII) is a

V

Y Z

260

Encyclopedia

fractional factorial design with no main effects confounding each other, but the main effects are confounded with two-factor interactions (or 1 main effect + 2 factor interactions = 1+2 = III). Resolution IV (or RIV) is a fractional factorial design with no main effects and two-factor interactions are confounded each other, but two-factor interactions may be confounded with each other (or 2+2=IV). A Resolution V has no confounding with main effects and two-factor interactions, but twofactor interactions may be confounded with three-factor and higher interactions or confounding of a main effect and a four-factor interaction (or 2+3=V=1+4). This matrix can also be found within MINITAB or in the appendix of many DOE textbooks.

A B C D E F G

The confounding found in the following example of a half-factorial design of a two-level, three-factor, 4-run (out of a possible eight combinations) structure contains three sets of confounding:

H

Number of Factors

I

Runs

J

4

K

8

L

16

M

32

N

64

O

128

3

2 Full

4

5

6

7

IV

III

III

III

Full

V

IV

IV

IV

Full

VI

IV

IV

Full

VII

V

Full

Full

8

III Full

P Q

Run

A

B

C

AB

AC

BC

ABC

R

2

-

-

+

+

-

-

+

T

3

-

+

-

-

+

-

+

U

4

+

-

-

-

-

+

+

8

+

+

+

+

+

+

+

S

V W X Y Z

The confounding scenarios (wherein the runs are exactly the same) represented in the above example are • Column “A” is confounded with the “BC” column. • Column “B” is confounded with the “AC” column. • Column “C” is confounded with the “AB” column. Each of the main effects (A, B, and C, for example) is confounded with a 2-way interaction factor (perhaps BC, AC, and AB), hence, this is a Resolution III design, which is a low (and often undesirable) resolution.

De sign of Experiment (DOE)

261

• Treatment Combination—Combination of control factors and levels, sometimes simply called runs or combinations. • Treatment Order • Run Order—When the test is conducted in a random order to eliminate bias. • Standard Order—When the test is conducted such that its treatment combinations are in order. A DOE design could represent a single type of experiment or a suite of different types of experimentations wherein the carefully designed combination of tests builds a deeper understanding of the process. Not surprisingly, there are various classes of designed experiments, each having strengths for use in certain applications. The main categories are

A B C D E F G

• Screening Design—To sort through several variables using a two-level design (minimum and maximum) to identify the critical ones. They allow for the estimation of main effects only and are useful when faced with many potential factors to assess. Screening DOEs typically use a Resolution III design and nearly always are followed by a higher resolution design to select the optimal level of selected significant variables.

H

• Full-Factorial Design—To test all the possible combinations and levels of the factors. It allows for the estimation of all effects so there is no confounding of effects. Due to size of experiment, it is useful with only a small number of factors (typically less than 6).

M

• Fractional Factorial Design—To test a subset of a full-factorial design, to save on time, resources, and money, and is one of the most common types of DOE. Often follows after a screening design to consider potential interactions. It allows for the estimation of main effects and some interactions. It typically uses a Resolution IV and V design but can be folded over to increase resolution and simplify confounding structure. • Optimizing DOE—To determine optimum levels of the variables, using more complex designs such as Response Surface methodology or Evolutionary Operation (EVOP). • Response Surface Design—To discover the shape of the response using geometric concepts to depict the relationships being discovered. Often used when a response needs to be optimized. It is able to estimate nonlinear (usually quadratic) effects. Because of size of experiment, only a small number of factors can be considered (generally less than eight). Often preceded by a screening design or fractional factorial design in order to focus on a few main factors of interest. • Robustness (Taguchi) Design—To find a region of low variability in the response and then moving the response to minimum, target, or maximum requirements, ultimately to optimize signal-to-noise ratios.

I J K L N O P Q R S T U V W X Y Z

262

Encyclopedia

• Confirming DOE—Method that tests the factors at the optimum level to verify that the prediction matches reality.

A B C D E F G H

• Mixture Design—To test variables expressed as proportions of a blend, which when summed would equal 100%, as the whole blend. Used when there is a dependency structure within the factors. This design type frequently is found in chemical applications or other recipe-type applications. • Conjoint Analysis—To investigate the attribute data associated with customer preference and purchasing behavior. (See Also “Conjoint Analysis,” p. 207)

How to Use the Tool or Technique The methodology, in its most basic sense, involves four key steps: Step 1.

Plan the experiment

J

Step 2.

Conduct the experiment

K

Step 3.

Analyze the data

L

Step 4.

Draw conclusions

I

M N

Step 1 Plan the Experiment

O

A majority of the effort in conducting a DOE should be budgeted for the planning phase.

P

The planning tasks include

Q R S T U V W X Y Z

A. Identifying the process for experimentation and the experiment’s objective(s) or hypotheses. B. Defining the response variables (Ys). C. Gathering the knowledge about the process (via Process maps, Cause-and-Effect diagrams, and so on). D. Developing a list of process variables (Xs) and classifying them as independent, dependent, uncontrollable, or noise factors. E. Assigning the appropriate level to each variable. To simplify the experiment, a two-level (high/low) might a good initial setting. F. Assessing the measurement system. If the measurement process is new, run several trials to become comfortable with the measurement system and validate its accuracy prior to running the experiment. (See Also “Conjoint Analysis,” p. 207) G. Selecting the appropriate design. H. Conducting a “sanity” check on the treatment combinations to ensure a reasonable experimental region.

De sign of Experiment (DOE)

263

I. Determining the appropriate sample size. (See Also “Sampling,” p. 618 for more information in determining the correct sample size.) The DOE Planning deliverables include • Specified hypothesis of interest

A

• Experimental protocol documented

B

• All treatment combinations specified

C

• Run order of treatment combinations defined

D

Generic 23 Design Example Using MINITAB Set up a generic MINITAB DOE design by following this procedure:

E F G

1. Open the MINITAB application software; from its toolbar, select the following sequence: Stat > DOE > Factorial > Create Factorial Design…. If the scenario of a 2-level, 3-factor design were continued (a 23 design), the MINITAB dialog boxes are illustrated in Figure D-7.

H

2. Select the 2-level factorial (default generators) (2 to 15 factors) because this example involves a 2-level factorial design, indicated by Area 1 in Figure D-7.

K

3. In the Number of Factors area, select 3 from the drop-down list for the three factors in the example (that is, A, B, and C), indicated by Area 2 in Figure D-7. 4. Select the Designs… button on the main dialog box, and a second dialog box opens called Create Factorial Design—Designs, indicated by Area 3 in Figure D-7. In this case, because there are only three factors and two levels, the only options are “Resolution III” or “Full.”

I J L M N O P Q R S T U V

Note

W

As needed, select the number of center points, replicates (typically

X

three or more are selected, but to keep the example simple, one is

Y

selected), and the number of blocks (one is selected to represent dif-

Z

ferent operators who will be involved in the experiment).

a. Select Full factorial 8 Full 2**3 because the example will test all possible combinations of the three factors at two levels, versus a fractional-factorial design. Leave the other options at their default levels.

264

Encyclopedia

b. Select OK in the Create Factorial Design—Designs dialog box. c. Select OK in the main Create Factorial Design dialog box.

A

A MINITAB Worksheet opens with the specified experimental design as its structure. This window is to be used as the data collection template in the experiment.

B C D E F G H I J K L

Figure D-7: Creating the 23 Design in MINITAB

M N O

After OK is selected in both dialog boxes, a MINITAB Worksheet will be created. Figure D-8 displays a sample one for this scenario.

P Q R S T U V W X Y

Figure D-8: Sample MINITAB Worksheet Contents in (Random) Run Order

Z

Note The data in your “StdOrder” (C1) column more than likely is different from that in Figure D-8 because the order is generated randomly to eliminate testing bias.

De sign of Experiment (DOE)

265

Recall, that this MINITAB Worksheet serves as the DOE “data collection form” referred to in the subsequent DOE Step 2a instructions.

Step 2 Conduct the Experiment To conduct the DOE, the tasks include A. Utilize a data collection form that mirrors the design’s resolution and run order, including replications. (See Figure D-06.)

A B

B. Actively collect the data.

C

C. Note any unusual outcomes.

D E

The DOE conducting deliverables include

F

• Completed and documented experiment results.

G

Catapult Example Using MINITAB For purposes of example, a catapult experiment will serve as the DOE for the remainder of this section. Most Six Sigma practitioners experience a catapult at some point in their careers because it is such a good teaching tool. Figure D-9 provides a simple illustration of a catapult device.

H I J K L

Ball Type

M N

3 Cup Position

2

O

1 3 Rubber Band Attach Point

Pullback Angle

P

4 2

3 2

Stop Position

1 6

5

4

3

Rubber Band Hinge Point

1 2 1

Q R S T U V W X

Figure D-9: Illustration of a Simple Catapult

Figure D-9 illustrates the multiple variables that can be adjusted to determine the optimal distance that the ball can be projected. For purposes of this illustration, the three factors will be assigned as follows:

Y Z

266

Encyclopedia

• A = “Team” conducting the experiment, of which there are two in this example, with its two levels indicated as team “A” or “B.” • B = “Projectile” type, specifying what type of ball is used in the experiment. In this case, the two levels (or types) of balls are “plastic” or “golf.” A B C D E F

• C = “PBA” (Pull Back Angle) on the catapult arm, with its two levels identified as “40” and “50” degrees as indicated on the catapult device itself. This “factor” information can be entered into the MINITAB by selecting the Factors button on the Create Factorial Design dialog box and entering the preceding data, as shown in “Area 4” of Figure D-10.

G H

Note

I

If the MINITAB Worksheet window is in view, a shortcut to return to

J

that dialog box from the MINITAB Worksheet is by simultaneously hit-

K

ting the Ctrl+E keys on your computer keypad. The most recent active

L

dialog box should appear.

M N O P Q R S T U V W X Y Z

1. Select Ctrl+E simultaneously on the computer keypad. Notice the prior Steps 1, 2, and 3 data still are selected. If not, please re-enter the example information stated earlier. 2. Select the Factors button on the MINITAB dialog box (shown as “Area 4” in Figure D-10), and enter in the experiment design by providing the specific input variable and respective levels: a. Type Team, select Text in the respective drop-down window, given that the factor is attribute data (category name, for example), and type the two experiment levels A and B for the names of the two teams. b. Type Projectile to represent the ball-type, select Text in the respective drop-down window, given that the factor is attribute data (category name), type the two experiment levels Plastic and Golf for the two types of balls to be used. c. Type PBA to represent the “pullback angle” in the experiment, select Numeric in the respective drop-down window, given that the factor is continuous data (inches, for example), and type the two experiment levels 40 and 50 for the pullback angle distance on the catapult.

De sign of Experiment (DOE)

267

d. Select OK in the Create Factorial Design—Factors dialog box. e. Select OK in the main Create Factorial Design dialog box.

A B C D E F G H I Figure D-10: Defining the Catapult Factors of a 23 Design in MINITAB

For purposes of illustration, two teams conducted the catapult experiment, following the RunOrder instructions from the MINITAB Worksheet for the prescribed factors at the prescribed levels. Recall that the RunOrder is randomized by MINITAB to eliminate bias, such that the teams A and B alternate randomly, and their respective experimental setup is randomly alternated. The teams recorded their results of the catapult experiment in the MINITAB Worksheet (serving as the data collection form). Figure D-11 shows a MINITAB Worksheet containing the recorded data in column C8, “Distance,” where the distance of the two different projectiles were recorded in inches for the eight runs. Entered data from experiment

J K L M N O P Q R S T U V W X Y Z

Figure D-11: MINITAB Worksheet Showing Sample Catapult 23 Results

268

Encyclopedia

Step 3 Analyze the Experiment To analyze the DOE results, the tasks include A. Plot the data. A

B. Separate sources of variability in the response via ANOVA. (See Also “Analysis of Variance (ANOVA)—7M Tool,” p. 142)

B

C. Distinguish special cause variation from common cause variation.

C

D. Create plots of factor effects.

D E F G H

E. Develop a simplified (parsimonious) model. F. Validate model assumptions. The DOE analyzing deliverables include

I

• Graphed of the results

J

• Modeled assumptions

K L M N O P Q R S T U V W X Y Z

Catapult Example Using MINITAB (continued) For purposes of the catapult example using MINITAB, we first generate the graphical plots that show the impact of each factor (or input variable— the two teams doing the experiment, pullback angle, and type of ball) on the response variable (distance the ball traveled) to see which one had the greatest impact. We will consider both of the input variables and the interaction of the three. Using MINITAB, the sequence from the toolbar would be Stat > DOE > Factorial > Factorial Plots… First generate the plots that show the impact of each input factor on the response by selecting the following items in the MINITAB DOE screens: 1. Using the completed MINITAB Worksheet with the recorded experimental data; select the following toolbar sequence: Stat > DOE > Factorial > Factorial Plots … to open the Factorial Plots main screen. 2. Select the Main Effects Plot on the main dialog box, indicated by Area 5 in Figure D-12. a. Notice that as you select the checkbox, its respective Setup… button becomes highlighted. Select the Setup… button, and a Factorial Plots—Main Effects dialog box opens, as shown by Area 6 in Figure D-12.

De sign of Experiment (DOE)

269

b. Select the response variable in the far left of the screen and double-click it to insert it into the Response dialog area. c. Find the screen’s Available list of main effects variables located in the center of the screen and select all the factors by clicking the > button to select one at a time or by clicking the double-stacked > button to select all at once.

A

d. Click OK.

B

3. Select the Interaction Plot on the main dialog box, also indicated by Area 5 in Figure D-12. a. Notice that as you select the checkbox its respective Setup… button becomes highlighted. Select the Setup… button, and a Factorial Plots—Interactions dialog box opens, as shown by Area 6 in Figure D-12. b. Same as the preceding step, select the response variable in the far left of the screen and double-click it to insert it into the Response dialog area.

C D E F G H I J K

c. Find the screen’s Available list of main effects variables located in the center of the screen and select all the factors by clicking the > button to select one at a time or by clicking the double-stacked > button to select all at once.

M

d. Click OK.

O

L N P Q R S T U V W X Y Z

Figure D-12: Analyzing the Catapult Factors of a 23 Design in MINITAB

270

Encyclopedia

The resulting graphical analysis from MINITAB produces two graphs, one for the main effects and one for the interactions among them, as shown in Figure D-13. Main Effects plot

Interactions plot

A B C D E F G

Figure D-13: Graphical Analysis of the Catapult Factors of a 23 Design in MINITAB

H I J K L M N O P Q R S T U V W X Y Z

What do these graphs indicate? The greater the slope in the line (that is, the less horizontal and more vertical the line is oriented), the greater the impact. In the Main Effects plot, the line between team A and B is horizontal, indicating no difference between the two teams. The Projectile (ball-type) and PBA (pullback angle) graphs both indicate a difference in effect between their two respective levels. In the Interactions plot, both the Team versus Projectile-type (in the upper-left corner) and the Team versus PBA (in the upper-right corner) feature two lines almost on top of one another indicating little ability to discriminate. However, in the graphical segment in the lower-right corner for the Projectile versus PBA, it features two separate lines with slightly different slopes. Is the interaction between the Projectile and PBA large enough to include in the experimental model? This question is answered in the next step.

Step 4 Draw Conclusions To draw conclusions from the DOE analysis, the tasks include A. Interpreting results using the process language that makes sense to the process players (not statistical terms alone). B. Verifying the conclusions with additional runs, if necessary. C. Planning future experiments, based on new the new knowledge, if applicable. D. Preparing and distributing a report with the findings and conclusions, including what new knowledge was learned about the process and any lessons learned from using the DOE technique.

De sign of Experiment (DOE)

271

The DOE Conducting deliverables include • Documented report with key findings, conclusions, and lessons learned.

Catapult Example Using MINITAB (continued) MINITAB will answer the question of statistical significance for the factors that were graphed as part of Step 3. To analyze the data and plots, follow the procedure in the MINITAB DOE screens: 1. Select the following MINITAB toolbar sequence: Stat > DOE > Factorial > Factorial Design …to open the main Analyze Factorial Design screen. 2. Select the response variable from the list of options on the left of the screen by double-clicking the correct factor or clicking the Select button. Notice it will populate in the Response dialog box. In this example, “Distance” is selected. 3. Click the Graphs button to open the Analyze Factorial Design— Graphs screen, indicated by Area 7 in Figure D-14.

A B C D E F G H I J K L M N O P Q R S T U

Figure D-14: Analyzing MINITAB Graphs of the Catapult Example (23 Design)

a. Select two graphical tools of the Effects Plots checkboxes, Normal and Pareto, to use to analyze the data. Keep all other defaults selected, as shown by Area 7 in Figure D-14. b. Click OK. 4. MINITAB produces two outputs: the Session Window Output to understand the amount of error produced in the model, shown in Figure D-15, and the Normal plots and Pareto graph, shown in Figure D-16.

V W X Y Z

272

A B C

Encyclopedia

Current model for all terms (factors) selected.

There are no degrees of freedom remaining to estimate error after fitting all model terms.

D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure D-15: MINITAB Session Window of the Catapult Example (23 Design)

The ANOVA is used to analyze the model. (See Also “Analysis of Variance (ANOVA)—7M Tool,” p. 142.) The MINITAB Session Window Output provides the current model for the catapult experiment. Recall that this example selected all of the factors as part of the model, as per Step 3) “Analyze the Experiment”, sub-step 3,” (illustrated in Figure D-12). Figure D-15 shows that model results with zero “Residual Error”, which means that the model has no degrees of freedom remaining to estimate error after fitting all the terms in the model. Hence, the model has be simplified to release some degrees of freedom. Only the input factors (X) that impact the response variable (Y) the most should be included in the model. Which are the most significant input variables? Look to the MINITAB graphical output for the answer. Next, examine the MINITAB graphical output, shown in Figure D-16. As shown in Figure D-16, both the Normal plot and Pareto chart depict that the Projectile type (Golf or Plastic ball, indicated by “B”) Is significant in impacting the distance of the projectile when tossed from the catapult. Hence, it should be included in the model. The Pareto chart also indicates that the Pullback Angle (Factor “C”) is very close to having a statistically significant influence on the distance output; therefore, it could be included in the model as well.

De sign of Experiment (DOE)

273

A B C D E 3

Figure D-16: MINITAB Normal and Pareto Graphs for the Catapult Example (2 Design)

F G

The interaction of any main effect showing significance must also be included in the model. In this case, that would be the 2-way interaction of the Projectile/Pullback Angle. The 3-way interaction of Projectile/Pullback Angle/Team appears equally significant when comparing the lengths of the Pareto chart bars, so it should be included in the model.

H

As a result, the model to analyze next should involve the following factors: Projectile (B), Pullback Angle (C), and Projectile/Pullback Angle interaction (BC) and the Projectile/Pullback Angle/Team interaction (ABC).

M

Next re-run the simplified model with only the four factors: B, C, BC, and ABC, as illustrated in Figure D-17.

Q

5. Select the following MINITAB toolbar sequence: Stat > DOE > Factorial > Factorial Design … to open the main Analyze Factorial Design main screen.

S

6. Select the Terms button to open the Analyze Factorial Design— Terms screen. a. Select the four terms (B, C, BC, and ABC) and place in the Selected Terms dialog box. b. Click OK to return to the main Analyze Factorial Design main screen. 7. Select the Graphs button to open the Analyze Factorial Design— Terms screen. a. Select the Four in One Residual Plots graphs button to view all the graphical tool analyses in one screen.

I J K L N O P R T U V W X Y Z

274

Encyclopedia

b. Click OK to return to the main Analyze Factorial Design main screen. c. Click OK again. A B C D E F G H I J K

Figure D-17: Re-run MINITAB Analysis on Simplified Model for the Catapult Example (23 Design)

L M N

The resulting MINITAB Session Window Output is shown in Figure D-18, and its graphical analysis is found in Figure D-19.

O P Q R S T U V W X Y Z

Figure D-18: MINITAB Session Window of the Re-run Catapult Example (23 Design)

De sign of Experiment (DOE)

275

This example’s simplified, 4-factor model can be written as Y = 71 - 9.5*Projectile + 6.25*PBA This simplified model has sufficient degrees of freedom (that is, five) to estimate error. The p-value (or probability of being significant at the 5% alpha default level) is a test for significance. Examining the p-value in the far right column of the Session Window Output indicates that for this simplified model, the Projectile is at 0.000, and the PBA is at 0.003, both of which are the critical value of < 0.05, hence they are significant. (See Also “Hypothesis Testing,” p. 335 for additional information on p-values.) Notice that the Session Window Output also provides the coefficient of determination (“R-Sq,” or R-squared or R2), indicating how well the overall “goodness of fit” for the model. The R-square is defined as the ratio of explained variation to unexplained variation in the model. In the case of this example, the simplified model has an R-square at 95.26%, which is good. The coefficient of determination is bounded between zero and one, or 0% and 100%. It measures the percent of variability in the response explained by the model with the predictor. The “rule of thumb” to determine the R-square’s acceptability, or strength of the relationship, includes • Mechanistic and Physical models: • Excellent = 90 to 100%

A B C D E F G H I J K L M N O P

• Good = 75 to 89%

Q

• Weak = 60 to 74%

R

• Behavioral and Human models:

S T

• Excellent = 70 to 100%

U

• Good = 50 to 69%

V

• Weak = 30 to 49%

W

Hence, the projectile and the pullback angle are useful in predicting the distance. By referencing the “Effects” column in the Session Window Output, the effect on distance is modeled. The golf ball travels 19 inches farther than the plastic golf ball. Pulling the catapult’s arm 10 degrees (from 40 to 50) increases the distance by 12.5 inches, or pulling it back by simply one degree increases the distance by 1.25 inches.

X Y Z

276

A B

Encyclopedia

The simplified model’s four-in-one graphical output indicates that the residual analysis passes model assumption tests, as shown Figure D-19. The data falls close to the normal probability plot. However, there is a slight smile-shape to the Residuals Versus the Fitted Values chart, suggesting a quadratic term. The graph should look random, without any pattern at all. If the BC interaction is placed back into the model, and the MINITAB analysis is re-run, the slight smile pattern becomes more random. See Also “Regression Analysis,” p. 571.

C D E F G H I J K L M N O P Q R S

Figure D-19: MINITAB Residual Plots for the Re-run Catapult Example (23 Design)

Typically a two-level model (as in the case of this catapult example) is insufficient to model a non-linear event. The best approach is to start with a two-level model first to eliminate insignificant factors. Then conduct a threelevel experiment on only those remaining critical variables. Notice that the number of test runs in a three-level full factorial model increases. For a threefactor, two-level (23) design, the number of runs is eight (2x2x2). For a threefactor, three-level (33) design, the number of full factorial runs is 27 (3x3x3).

T U V W X

Hints and Tips Planning a DOE is the most critical step in the process, and an appropriate amount of time should be taken to understand the variables of

Y

interest.

Z

Planning a DOE Brainstorm a list of key independent and dependent variables with people knowledgeable of the process. Use Process Knowledge to select the factor settings. Try to select extreme settings (high and low) to begin to distinguish boundary conditions. Understand the process conditions and standard operating procedures. Determine any constraints—cost,

De sign of Experiment (DOE)

277

time, materials, resources (including people). Consider conditions, and if these elements represent a non-linear effect, a large number of runs may be needed to define the curvature of the model, which often requires a larger budget.

A

Rule of Thumb Sequence of the Different Types of DOEs Type of DOE:

Screening

Fractional Factorial

Full Factorial

Response Surface

Usual # of Factors

>4

3–15

1–5

Dotplot > One Y > Simple…. Figure D-21 displays sample MINITAB screens where the appropriate graph type is selected (Area 1), and the appropriate variable data of interest (that is, Yield) are selected (Area 2) to produce the final sample Dotplot found in Figure D-22.

W X Y Z

282

Encyclopedia

A B C D E F

Figure D-21: Example MINITAB Dotplot Main Screen

G H I J K L M N O P Q R S T U V W X Y Z

Figure D-22: Example MINITAB Dotplot

Figure D-22 illustrates how each data point is displayed in a Dotplot. If the data contains stratified information, the figure lacks the detail to distinguish one category from another.

Plotting Stratified Data of One Variable To examine the stratification within a data set, select With Groups in MINITAB’s Dotplot main screen to delineate that stratified data exits. For example, Figure D-24 illustrates the scenario wherein the data set also contains shift and location information about the produced yield. Select the following MINITAB commands from its drop-down menu: Graph > Dotplot > One Y > With Groups…, as shown in Figure D-23, identify the data to be plotted (shown in Area 2), and identify the data to be stratified (shown in Area 3) to produce the final sample Dotplot found in Figure D-24.

Dotplot

283

A B C D E F G H I Figure D-23: Example MINITAB Dotplot With Groups Main Screen

J K

Figure D-24 illustrates how there appears to be similar variability across Shifts 1, 2, and 3 and between Plants A and B. However, Plant A looks to be producing about a 10% higher yield than Plant B.

L M N O P Q R S T U V W X Y Z

Figure D-24: Example MINITAB Dotplot with Groups

284

Encyclopedia

Plotting Stratified Data of Two or More Variables Dotplots handle not only stratification within a data set well, but also display multiple variables of interest well, provided the total number of data points remains in the 15 to 20 range. MINITAB refers to multiple variables as Multiple Y’s. A B C D E F G H I J K

For example, Figure D-26 illustrates the example scenario wherein the data set also contains both the stratification information of shift and location and the output of two different products (Yield2 and Yield3). Select the following MINITAB commands from its drop-down menu: Graph > Dotplot > Multiple Y’s > With Groups…, as shown in Figure D-25, identify the data to be plotted and the scale (shown in Area 2), and identify the data to be stratified (shown in Area 3) to produce the final sample Dotplot found in Figure D-26. Figure D-26 illustrates that again there is similar variability across shifts and between plants and products (or yields). There are many mean differences: Plant A has lower yields with Product 3 (Yield3); Plant B has differences among the three shifts with Shift 1 being the best and Shift 2 being the worst. (See Also “Stratification—7QC Tool,” p. 697, for more detail on stratifying data.)

L M N O P Q R S T U V W X Y Z Figure D-25: Example MINITAB Dotplot Multiple Y’s and with Groups Main Screen

Dotplot

285

A B C D E F G H Figure D-26: Example MINITAB Dotplot Multiple Y’s and with Groups

I J K

Supporting or Linked Tools Supporting tools that might provide input when creating a Dotplot include

L M N

• Data gathering plan to collect the appropriate metrics (See Also “Data Collection Matrix,” p. 248)

O

• Performance charts and dashboards

Q

A Dotplot can provide input to tools such as

P R S T

• Cause-and-Effect diagram. (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173)

U

• QFD (See Also “Quality Function Deployment (QFD),” p. 543)

V

• Statistical analysis tools and other root cause analysis techniques (See Also “Hypothesis Testing” and “Regression Analysis,” p. 335 and 571, respectively.) • FMEA with follow-on action planning (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) Figure D-27 illustrates the link between a Dotplot and its related tools and techniques.

W X Y Z

286

Encyclopedia Cause-Effect Diagram Data Gathering (metrics) Dotplot

A B C

Performance Charts and Dashboards

D

G H I J K L M N O P Q R S T U V W X Y Z

Statistical Analysis Tools

FMEA

E F

QFD

Figure D-27: Dotplot Tool Linkage

Failure Mode s and Effe cts Analysis (FMEA)

287

F Failure Modes and Effects Analysis (FMEA) A

What Question(s) Does the Tool or Technique Answer? What can go wrong? What can be done to prevent or minimize it? What are the potential failures that could occur, and what is the best response (action) plan to minimize its impact if it does happen?

D

An FMEA helps you to

E

• Identify what could go wrong. • Analyze the magnitude of potential risk associated with a possible

problem or failure event. • Prioritize the potential problems based on a three-fold approach

integrating the ability to detect their occurrence, their likelihood of occurrence, and potential severity of impact if they do happen. • Document an action plan if the failure event occurs, particularly

those posing the biggest risk.

B C

F G H I J K L M N

Alternative Names and Variations This tool is also known as • FMEA

Variations on the tool include • Design FMEA

O P Q R S T

• Healthcare FMEA

U

• Product FMEA

V

• Process FMEA

W

• Systems FMEA

X Y

When Best to Use the Tool or Technique FMEA is a proactive tool. To remain competitive and prevent possible failures, this tool is used in risk planning and management, often prior to implementing something new—a change or improvement. The FMEA serves as an integral tool used in Six Sigma projects to evaluate and reduce the risk in a new process. It is also used in ongoing operations to evaluate risks and develop contingencies. The FMEA should be part of an evergreen process and a critical component of a control management plan.

Z

288

A B C D E F G H I J K L M N O P Q R S T U V W

Encyclopedia

Brief Description The Failure Mode and Effects Analysis (FMEA) is a popular and welldocumented tool that uses a systematic method to define a problem-prevention strategy when implementing a new and/or managing a process, project, system, or offering of product and/or services. Failure mode means the various ways something could go wrong. Effect Analysis describes the potential impact of a failure if it occurs and the potential counter-measures to put in place to minimize or prevent the impact. It features different evaluation criteria scales customized for different situations. If constructed properly, using data as input, this analysis can represent a time-consuming process, thus the contingency planning portion of the process focuses on the highest priority risks. The FMEA tool is designed to identify potential failure modes early to minimize their effect (or their impact) if they happen. It aids in developing a robust design and respective control plans. FMEA identifies potential critical and significant characteristics in a design. The tool establishes a priority for actions and documents the rationale used. The approach used to develop an FMEA is frequently referred to as a bottoms-up systems analysis, trying to identify possible failure modes before they occur. This is a complementary analysis to the Fault Tree Analysis (FTA), which employs a tops-down method. The United States military developed the tool in the 1940s, and NASA used it in the space program in the 1960s. Initially, the tool had two primary purposes: 1) to analyze possible failures of product components (hardware, software) and 2) to analyze the functionality (what something does) relative to its designed performance and output. Now multiple industries have embraced it such as manufacturing, high-tech, and healthcare. The FMEA popularity has expanded its application to existing or newly developed product (hardware, software, or system), service, process, or project. The purpose of an FMEA is to prevent problems with ongoing operations and implementing improvements or newly designed items or events. A well-constructed FMEA produces the following deliverables:

X

• List of potential failure modes

Y

• List of potential critical and potential significant characteristics

Z

• List of effects • List of causes • Documentation of current controls • Requirements for new control plans

Failure Mode s and Effe cts Analysis (FMEA)

289

• Documentation and prioritization of improvement activities • Documentation of the history of improvements

Figure F-1 provides a blank FMEA template that contains the deliverables listed here, labeled with the primary steps involved in building its content. The next section, “How to use the Tool or Technique,” covers this entire procedure.

Sales Kick-off Event

Step 1 b

Date Completed:

Severity Criteria:

Occurrence/Frequency Scale:

Detection Scale:

10 = Hazardous without warning

10 = Absolute Uncertainty

10 = > 1 in 2

or Very High

9 = Hazardous with warning

9 = 1 in 3

or Very High

9 = Very Remote

8 = Very High

8 = 1 in 8

or High

8 = Remote

7 Date Refreshed:

= High

Step 3

7 = 1 in 20

or High

7 = Very low

6 = Moderate

6 = 1 in 80

or Moderate

6

5

5 = 1 in 400

or Moderate

5 = Moderate

4 = Very low

4 = 1 in 2,000

or Moderate

4 = Moderately high

3 = Minor

3 = 1 in 15,000

or Low

3

2 = Very Minor

2 = 1 in 150,000

or Low

2 = Very High

1 = None

1 = < in 1,500,000

or Remote

1 = Almost certain

= Low

E

= Low

F

= High

Controls

Triggers

RPN = Sev X Occ X Det

Potential Causes

Current

Detectability

Effects (Impact)

Step 1 a

G H

Action Planning Occurrence

Mode (What)

Severity

Risk Category

D

Developed by:

Risk Assessment Potential Failure

Action Type (Accept Avoid, Mitigate– Reduce or Translate)

Contingency Plan

Outcome Measure

Person Accountable

Step 4 b

Step 4 c

Step 4 d Step 4 e

I J

Step 2 Step 4 a

B C

FMEA Template Topic:

A

Step 5

K L M

Figure F-1: Blank FMEA Template

N O P

How to Use the Tool or Technique After the initial planning and organization activities, there are basically four major deliverables to develop an FMEA document: 1. Failure modes identified

Q R S T U

2. Risk categorizations defined for Severity, Occurrence, and Detection

V

3. Effects and risk impact determined

W

4. Contingency action plan developed

X

The general 5-step procedure is as follows: Step 1.

Plan and Organize for an FMEA. a. Assemble diverse team to develop the FMEA, ideally comprised of multiple disciplines, backgrounds, experience, roles, rank, and years of experience. Record their names and function.

Y Z

290

Encyclopedia

b. Gain agreement on target topic, and associated outputs, metrics, and/or critical parameters of the target topic. c. Gather “background” data from a variety of sources including • Historical documents on current topic when examining an existing product, process, or service

A B

• Historical documents on similar topics if analyzing a new design, process, service, or project

C D E

• Insights from subject matter experts

F

• Data sources, which include VOC data (Customer satisfaction and VOC surveys), CTQs, QFD, SWOT, benchmarking, industry and technical analyst critiques, detailed Process map, Fishbone diagrams, Cause-and-Effect matrix, historic FMEAs, customer support, service and regulatory logs, control charts, financial data and models, dashboards, and so on

G H I J

Step 2.

Identify and group potential failure modes. Answer the hypothetical question, “What could go wrong?” The team references the background information and its diverse perspectives and employs any or all of the following approaches: 1) Brainstorming, 2) Affinity diagramming, and 3) Tree diagramming.

M N O P Q

a. Record the failure modes and corresponding category (or group) on an FMEA template. Figure F-2 shows the Risk Assessment portion of an FMEA template.

R S T U W

Risk Category Mode (What)

Effects (Impact)

Potential Causes Severity

V

Potential Failure

Current

Controls

X Y Z

Figure F-2: Sample Risk Assessment Portion of an FMEA Template

Triggers

RPN = Sev X Occ X Det

Risk Assessment Detectability

L

Occurrence

K

Failure Mode s and Effe cts Analysis (FMEA)

Step 3.

291

Define the three evaluation criteria to be used and record them on the FMEA as a reference key or footnote. (Customize the evaluation criteria definition and scales to what is the most appropriate for the situation and the industry.) a. Severity Rating (SEV)—1 to 10, for example, where 10

indicates most severe.

A

b. Occurrence or Frequency (OCC)—1 to 10, for example,

where 10 represents the highest likelihood of occurrence. The occurrence criteria can indicate: i. Number of times the failure may occur

B C D E F

ii. Time period (measured in appropriate increments: minutes, days, months, and so on)

G

iii. Percent probability or terms describing probability

H

c. Detection Rate (DET)—1 to 10, where 10 represents the

lowest detectability. Figure F-3 provides an example scale for Evaluation Criteria portion of an FMEA:

I J K L M

Severity Criteria:

Occurrence/Frequency Scale:

Detection Scale:

N

10 = Hazardous without warning

10 = > 1 in 2

or Very High

10 = Absolute Uncertainty

O

9 = Hazardous with warning

9 = 1 in 3

or Very High

9 = Very Remote

P

8 = Very High

8 = 1 in 8

or High

8 = Remote

Q

7 = High

7 = 1 in 20

or High

7 = Very low

6 = Moderate

6 = 1 in 80

6 = Low

R

or Moderate

5 = Low

5 = 1 in 400

or Moderate

5 = Moderate

4 = Very low

4 = 1 in 2,000

or Moderate

4 = Moderately high

3 = Minor

3 = 1 in 15,000

or Low

3 = High

U

2 = Very Minor

2 = 1 in 150,000

or Low

2 = Very High

V

1 = None

1 = < in 1,500,000

or Remote

1 = Almost certain

W

Figure F-3: Sample FMEA Rating Criteria

If unique evaluation criteria are used, gain agreement on the scale and criteria and then document it on the FMEA template.

S T

X Y Z

292

Encyclopedia

Step 4.

Determine effects and risk impact. a. Evaluate the potential impact (or effect/consequences) of each failure mode, using financial models and sensitivity studies and document and rate the Severity according to the agreed-to rating. Record both items on the FMEA template.

A B

b. Identify the potential causes for each failure mode and rate the likelihood of occurrence (or frequency). Record both items on the FMEA template.

C D

c. Identify the controls and/or indication triggers (if any) and rate the ability to detect each failure mode. Record both items on the FMEA template.

E F G

d. Calculate the Risk Priority Number (RPN) for each failure mode effect and record it on the FMEA template.

H I

i. RPN Calculation is the product of multiplying the

J

three risk scores together: Severity x Occurrence x Detection Rate = RPN

K L

ii. The RPN number is a guide and requires judgment

M

and scrutiny when determining the relative importance of any given failure mode.

N O

e. Identify the high-priority failure modes by selecting the largest RPNs or those with the highest possible Severity or Occurrence rating, particularly if it impacts the customer.

P Q R

i. Sort the FMEA by RPN, highest to lowest.

S T U V

Step 5.

Develop the contingency action plan for each of the failure modes with the highest priority or risk as a contingency and record it on the FMEA.

W X

Note

Y

Those items with medium to low priorities should remain listed on

Z

the FMEA but do not require an Action Plan be developed for them unless the have particularly high probability of occurrence or evoke “special interest.”

Failure Mode s and Effe cts Analysis (FMEA)

293

a. Provide immediate attention to any failure identified with

the highest priority. b. Decide which action plan is appropriate for each failure

mode cause. There are three major Action Types or Risk Response classifications: i. Acceptance—Accept the consequences passively or

actively; retain the risk. ii. Avoidance—Eliminate a specific threat, usually by

eliminating the cause. iii. Mitigation—Reduce the expected monetary value

of a risk event by reducing the probability of occurrence. • Reduction—Minimize its occurrence and effect. • Transfer—All or a portion of it to another party

via • Insurance for direct property damage • Indirect consequential loss (often performed by a contractor—debris removal or equipment replacement, for example) • Legal liability (design errors, public bodily injury, performance failure) • Personnel (employee bodily injury/Worker’s Compensation) • Employee replacement costs • Resulting business losses c. Develop the specific contingency plan or action that

addresses the root cause. d. Identify the outcome measurements to ensure the recom-

mended contingency was achieved per plan; used to analyze or test success per the plan. e. Assign one and only one person to be accountable for

carrying out the recommended contingency plan if needed. f. Figure F-4 provides a sample Action Planning section of

an FMEA template.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

294

Encyclopedia Action Planning Action Type (Accept, Avoid, Mitigate -Reduce or Transfer)

A

Contingency Plan

Outcome Measure

Person Accountable

B C D E F G

Figure F-4: Sample Action Planning Portion of an FMEA Template

H I J K

Step 6.

Refresh and re-evaluate the FMEA periodically (updating it with new threats) or re-score it as failures occur and action is taken.

L M N O P Q

How to Analyze and Apply the Tool’s Output Evaluate internal and external risk position and document contingency plans if the risk were to happen. And then take action; prevent possible risk or respond to the occurrence of one. This is when the value of the FMEA analysis comes to light. (See Hints and Tips that follow for additional detail.)

R

• The Occurrence number (OCC) is the primary item to reduce.

S T

• After an action has been implemented, update the Risk Priority Number (RPN).

U

• Typically, the severity number stays the same.

V W X Y Z

• Detection actions alone are taken only as a last resort. Periodically refresh and update the document as part of an evergreen process. • Communicate and distribute the FMEA to appropriate parties. • Incorporate the FMEA into the control plan.

Failure Mode s and Effe cts Analysis (FMEA)

295

Examples FMEA Example—Acme Inc. Sales Kick-off Event Acme Inc. FMEA Template Topic:

Sales Kick-off Event

Date Completed:

Severity Criteria:

Occurrence/Frequency Scale:

Detection Scale:

10 = Hazardous without warning

10 = Absolute Uncertainty

10/6/2006

Date Refreshed:

10 = > 1 in 2

or Very High

9 = Hazardous with warning

9 = 1 in 3

or Very High

9 = Very Remote

8 = Very High

8 = 1 in 8

or High

8 = Remote

7 = High

7 = 1 in 20

or High

7 = Very low

6 = Moderate

6 = 1 in 80

or Moderate

6

5

= Low

Developed by:

= Low

5 = 1 in 400

or Moderate

5 = Moderate

4 = Very low

4 = 1 in 2,000

or Moderate

4 = Moderately high

3 = Minor

3 = 1 in 15,000

or Low

3 = High

2 = Very Minor

2 = 1 in 150,000

or Low

2 = Very High

1 = None

1 = < in 1,500,000

or Remote

1 = Almost certain

Speakers miss event

Participants miss event

Method

Kick-off Notebooks printed incorrectly

Participant satisfaction

Mgmt. disagree on date

Participants unavailable

Budget not agreed to

Bad Weather

Meeting Facility too small

2

64

2

4Q Outlook Meeting

2

32

1-week lead time policy

Microsoft Outlook task reminder

2

72

4Q Outlook Meeting

2

80

8

6

Presentation materials not finalized in time to meet printer’s lead-time.

6

8

Conflict with own team meeting

5

Cost overruns

Unclear whose budget will cover event (HQ versus Region)

4-week lead time policy 3

2

144

2

72

6

240

9

576

None

VP Sales to communicate policy

More people attend than invited (2nd level management) 3 All participants did not RSVP 8

Microsoft Outlook task reminder

4Q Outlook Meeting and Invitation Memo restate policy

6

126

7

168

None 3

E Contingency Plan

Outcome Measure

Person Accountable

F G H I J

4Q Outlook Meeting

8

7

D

Action Type (Accept Avoid, Mitigate– Reduce or Translate)

Avoid

None

Winter time frame

Participants dissatisfied

Triggers

8

8

Lack of participant‘s attention

VP Sales inspection VP Sales inspection

Nobody assigned the task.

Next event reports same mistakes

Controls

3

Invitations sent during year-end vacations.

Cost overruns

Unable to travel to/ from meeting

4Q Outlook Meeting

VP Sales inspection

Lack of management agreement

5

Mother Nature

VP Sales inspection

Lack of management agreement

4 Participants not surveyed at end

4

8

8 RSVPs not received

Detectability

Kick-off Agenda not available

Current

RPN = Sev X Occ X Det

Materials

Effects (Impact)

Potential Causes

B

Action Planning Occurrence

Mode (What)

Severity

Potential Failure

A C

Risk Assessment

Risk Category

George S. Sam T. Lois M. Bill M. Laura H. Julie H.

Sr Field Mgmt. to establish a policy

Policy communicated

VP of Finance

K L

Avoid

Assign a person to the task

100% participants surveyed

VP of HR

Reduce

Accommodate early travel time when needed

Voice mail; when needed

VP of HR

Avoid

Turn away “uninvited” guests

Person turned away at meeting door

VP of HR

Avoid

Call those who did not RSVP

100% invites accounted for

Natasha G., of HR

Figure F-5: Example: Acme Inc. FMEA for Sales Kick-off Event

M N O P Q R S T

Hints and Tips

U

Fully develop the list of failure modes first before jumping to fixes or

V

causes. Otherwise a significant potential risk may be overlooked

W

because many potential failures build on one another.

X

Include subject-matter experts and those with no knowledge of the

Y

topic under review to help develop a list of potential failures. Those

Z

with little to no knowledge provide a fresh perspective on the topic, and at times their seemingly naïve questions give insight into a root cause. Consider assigning team roles to those assembled to develop the FMEA and avoid having one person assigned multiple roles. The recorder of ideas should not participate in the idea creation to avoid

296

Encyclopedia

issues such as power of the pen and lack of time to think before contributing. (See Also “Brainstorming Technique,” p. 168) There can be multiple failures to any one item or process step being A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

examined, and they should be listed and analyzed separately. There can be multiple effects for any one failure and they should be listed and analyzed separately. Definitions of and scale for Severity, Occurrence, and Detection should be modified to fit the appropriate situation or industry. The highest score(s) indicate the need for significant contingency planning. Typical default scale is 1 to 10, where 10 represents the most severe, highest likelihood of occurrence, and lowest detectability. FMEA templates should reflect the purpose of the risk analysis (product, process, healthcare) and the needs of the organization. Some FMEA templates also add an Action Results section to the far right of the template that often includes items such as: • Action Taken and Date (what) • Management Approval (who) and Date • Recalculation Columns for Post-Action: Severity, Occurrence, Detection, and Risk Priority Number (RPN) • Implement whichever approach best reinforces the importance of using the FMEA as an evergreen document. Some industries associated with products or processes that involve safety issues or compliance with government regulations may need special controls to monitor potential risks. In that case, some add a “Critical Characteristic” column to the FMEA template to capture the specific measurements or indicators required. When adding the Critical Characteristics column, a new risk calculation is introduced, called Criticality, which is determined by multiplying the Severity times the Occurrence. Both of these columns are often placed just to the right of the RPN column. Similar to the RPN number, the Criticality is a guide and requires judgment and scrutiny when determining the relative importance of any given failure mode. A well-constructed FMEA is time-intensive to build and requires thorough research, exploration, and idea generation; hence, many think this tool is best suited for non-redundant systems or projects.

Failure Mode s and Effe cts Analysis (FMEA)

297

Note Risk Priority Number (RPN) There is no RPN threshold value needed to classifying a potential failure mode as high risk. There is no set RPN value that triggers a mandatory “Recommended Action” needed. Likewise, there is no minimum RPN value below which the team is automatically excused from action planning.

A B C D E F

Caution

G

Don’t focus on the Risk Priority Number (RPN) alone because other

H

important items may be overlooked, such as those with a high Sever-

I

ity or of particular importance to the customer or a regulatory agency.

J K

Taking Action 1. To reduce RPN, focus on reducing probability of occurrence, not on increasing the ability to detect an error. Lowering the likelihood of an error occurring is a measurable, proactive change to a process. 2. Avoid reducing RPN with only detection actions because often the

implementation that improves the ability to detect an error involves increased inspection (reactionary) methods. This seldom adds any value to the process. a. 100% inspection is only (at best) 80% effective.

L M N O P Q R S T

b. Reducing RPN with detection does not eliminate the failure mode, nor does it reduce the probability of causes.

U

c. Detection of 10 is not bad if the Occurrence is ranked as 1.

W

3. Techniques for generating risk ideas include

a. Interview Experts (Delphi technique—panel of experts, convergent solutions, for example) b. Group Techniques

i. Brainstorming—different starting point options • Start with prior project review of lessons learned. • Start with literature review (divided among team, bring summary and inferences to team).

V X Y Z

298

Encyclopedia

• Same industry • Similar industry • Trend/Thought leader A B C

• Start with category types of Risk Identification: Internal/External. ii. Project team review of project deliverables

D

• Team define High/Medium/Low Risk

E

• Review WBS for risk items

F

• Review project schedule and estimate for risk items

G H I J K L M N

• Review with subject matter experts/stakeholders • Compare to similar projects iii. Combination (experts and internal brainstorming)

• Independent events • Invite Experts to brainstorming iv. Tools and Techniques

O

• Checklists (by sources of risk)

P

• Flowcharting to show how various elements of a

system relate to Cause-and-Effect risks

Q R S T U V W X Y Z

• Interviewing various stakeholders 4. Risk Identification: Define the source, potential of occurrence, risk

symptoms, and inputs to other processes. Developing a checklist of potential failure modes that is refreshed and improved over time with experience can only help provide a trigger for idea generation. Not all items may be applicable in each situation, but the checklist can serve as a thought-starter. The following list is intended simply for that purpose. a. Internal i. Equipment/Hardware/Technical:

• State of the Art • Physical Properties • Material Properties

Failure Mode s and Effe cts Analysis (FMEA)

299

• Material Availability • Processes (design, operations, maintenance, and so on) • Integration • Testing • Security

A B C

• Safety

D

• Personnel

E

• Business

F

• Others… ii. Software/Technical

• (See level i.) iii. People (labor supply/issues)

• Along the value chain iv. Work Environment—Internal/Non-Technical

G H I J K L M N

• Political

O

• Organizational

P

• Process • Safety • Financial (including cash flow) b. External i. Marketplace

• Competition • Suppliers—along the supply chain (people, processes, technology, tangibles, and intangibles) • Customers • Non-customers • Political • Economic

Q R S T U V W X Y Z

300

Encyclopedia ii. Legal

• Licenses/patent rights • Lawsuits • Contractual agreements/breeches

A B

iii. Weather/Mother Nature/Natural Disasters

C

vi. Government and Regulatory

D E F G

Supporting or Linked Tools Supporting tools that might provide input when developing an FMEA include

H

• Graphical tools (See Also “Graphical Methods,” p. 323)

I

• Process map (See Also “Process Map (or Flowchart)—7QC Tool,”

J K L M N O

p. 522) • VOC and VOB data (See Also “Voice of Customer Gathering

Techniques,” p. 737) • Control chart (See Also “Control Charts—7QC Tool,” p. 217) • Brainstorming (See Also “Brainstorming Technique,” p. 168)

P

• 5-Whys technique (See Also “5-Whys,” p. 305)

Q

• 5M-and-P technique and its variants (See Also “Cause-and-Effect

R S T U V

Diagram—7QC Tool,” p. 173) A completed FMEA provides input to tools such as • Control chart (See Also “Control Charts—7QC Tool,” p. 217) • Brainstorming (See Also “Brainstorming Technique,” p. 168)

W

• 5-Whys technique (See Also “5-Whys,” p. 305)

X

• 5M-and-P technique and its variants (See Also “Cause-and-Effect

Y Z

Diagram,” p. 173) • Control plan (See Also “Matrix Diagrams,” p. 399) • Transition and/or Implementation plan (See Also “Matrix

Diagrams,” p. 399) Figure F-6 illustrates the link between an FMEA and its related tools and techniques.

Failure Mode s and Effe cts Analysis (FMEA)

301

5M and P Technique and its variants

5-Whys Technique

Graphical Tools Control Plan

B

FMEA

Process Map

A

Transition and/or Implementation plan

C D E

VOC and VOB Data

F Control Chart

Brainstorming Technique

G H I

Figure F-6: FMEA Tool Linkage

J K

Variations Design FMEA FMEA examines potential risks associated with the design and development of a new product, service, or process prior to launch or implementation. Its risk categories include • Environmental changes such as competitive and technology threats • Transition and implementation issues (ownership, training, problem and support escalation, documentation, and control plan) • Marketplace acceptance

L M N O P Q R S T

• Stability and quality issues of the new offering (including safety hazards)

U

Healthcare FMEA (HFMEA) HFMEA is used by hospitals, healthcare centers, and clinics involved in patient safety issues and consists of a unique five-step process:

W

Step 1.

Define the Topic

Step 2.

Assemble the Team

Step 3.

Graphically Describe the Process

Step 4.

Conduct the Analysis a. List Failure Modes

V X Y Z

302

Encyclopedia b. Determine Severity and Probability c. Use the Decision Tree d. List all Failure Mode Causes

A B C D E F G H I

Step 5.

Identify Actions and Outcome Measures

This tool was developed by the Department of Veterans Affairs National Center for Patient Safety (www.patientsafety.gov), and its template is supported by HFMEA Severity Ratings, HFMEA Scoring Matrix, and HFMEA Decision Tree. HFMEA scoring criteria uses a 1 to 4 scale. When evaluating each failure mode, the scoring is used to comprehend the impact on the patient outcome, visitor, staff, equipment, or facility and fire. Per the NCPS (National Center for Patient Safety), the standard HFMEA scale is 1. Event Severity Rating:

J

• Catastrophic (4)

K

• Major (3)

L

• Moderate (2)

M N O P Q

• Minor (1) 2. Probability Rating: • Frequent: (4) Likely to occur immediately or within a short

period (several times in 1 year).

R

• Occasional: (3) Probably will occur (1–2 times per year)

S

• Uncommon: (2) Possible to occur (sometime in 2 to 5 years)

T U V W X Y Z

• Remote: (1) Unlikely to occur (sometime in 5 to 30 years)

HFMEA Combined Scoring Figure F-7 provides the Healthcare FMEA risk rating criteria, showing a “combined” approach. HFMEA Blank Template Figure F-8 illustrates a blank HFMEA template. [Source: www.patientsafety.gov]

Failure Mode s and Effe cts Analysis (FMEA) Probability

303

Severity Catastrophic (4)

Major (3)

Moderate (2)

Minor (1)

Frequent (4)

16

12

8

4

Occasional (3)

12

9

6

3

Uncommon (2)

8

6

4

2

B

Remote (1)

4

3

2

1

C

Scores of 8 to 6 require significant contingency planning

A

D

Source: www.patientsafety.gov

E

Figure F-7: HFMEA Rating Criteria

F G

HFMEATM Subprocess Step Title and Number

Action Type Actions or Rational (Control, for Stopping Accept, Eliminate)

Outcome Measure

Management Concurrence

Proceed

Detectability

Existing Control Measure?

Decision Tree Analysis Single Point Weakness?

Haz Score

Probability

Potential Causes

Severity

Scoring Failure Mode: First Evaluate failure mode before determining potential causes

HFMEATM Step 5 - Identify Actions and Outcomes

Person Responsible

HFMEATM Step 4 - Hazard Analysis

H I J K L M N O P Q R S T

Source: www.patientsafety.gov

Figure F-8: HFMEA Blank Template

Process FMEA A Process FMEA examines potential risks or failures associated with an

existing, operating process (typically a transactional or highly-repetitive operational process). It often follows the process flow through all steps in sequential order or analyzed in reverse order. Risk categories include 5Ms-and-P: machines (equipment and technology), materials, methods (process steps), measurements, Mother Nature (environment, surroundings), people (or manpower). (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173)

U V W X Y Z

304

Encyclopedia

This tool can be applied to a project, following the project plan and analyzing its key milestones and deliverables. Risk categories could project plan milestones, deliverables, and the nine knowledge areas of project management: A B

• Scope • Time

C

• Cost/Budget

D

• Quality

E

• Risk Management

F

• Human Resources

G H

• Procurement (including partners and suppliers)

I

• Communications

J

• Integration

K

N

A Process FMEA sometimes is referred to as a Service FMEA (wherein the process is delivering a service). Whichever “label” is given to the FMEA, the approach still involves similar risk categories (5Ms and P and 4Ps) and should suit the particular industry requirements. (See Also “Causeand-Effect Diagram—7QC Tool,” p. 173)

O

Product FMEA

P

A Product FMEA examines potential risks or failures associated with an

L M

Q R S T U V W X Y Z

existing product or services offering (including consultative services, maintenance services, and so on). Risk categories include • Product features and functionality, robustness, manufacturing costs, and warranty and related maintenance, support, safety costs • 5Ms-and-P: machines (equipment and technology), materials, methods, measurements, Mother Nature (environment, surroundings), people (or manpower) • 4Ps: policies, procedures, people, plants (environment, surroundings) (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173) A Product FMEA sometimes is referred to as a Service FMEA (where the offering is the service provided). Whichever “label” is given to the FMEA, the approach still involves similar risk categories (5Ms and P and 4Ps), and should suit the particular industry requirements. (See Also “Causeand-Effect Diagram—7QC Tool,” p. 173)

Systems FMEA Examines potential risks associated with the design and development of systems and subsystems prior to launch or implementation.

5-Whys

305

5-Whys What Question(s) Does the Tool or Technique Answer? What is the root cause of this problem or problematic outcome? The 5-Whys help you to • Quickly focus on discovering the root cause • Understand the relationship between a problem and its different

causes

A B C D E F

Alternative Names and Variations This tool is also known as

G H

• Why-why

I

• 5W

J

Variations on the tool include • 5W2H or 5Ws and 2Hs • Cause-and-Effect diagram; Ishikawa diagram; Fishbone diagram

(See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173) • Fault tree analysis (See Also “Fault Tree Analysis (FTA),” p. 309)

K L M N O P Q

When Best to Use the Tool or Technique Before any action is taken, this tool helps identify the source(s) of problem, rather than the symptom.

R

It is often used in conjunction with the development of a Fishbone diagram when trying to analyze and prioritize multiple problems and their root causes, rather than superficial symptoms.

U

Particularly useful when the problems involve human behavior or interactions.

S T V W X Y

Brief Description The 5-Whys technique asks Why five times to drill down and identify the potential root cause. Applying this method prevents the analysis from stopping with symptoms, which leads to superficial solutions.

Z

306

Encyclopedia

This is a simple technique—easy to use and often gets to the heart of the issue without statistical analysis.

A B C D E

How to Use the Tool or Technique When using the 5-Whys technique 1. Select a potential key causal (often from a Cause-and-Effect diagram or Pareto). (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173) 2. Ensure everyone on the team has a common understanding of what that cause means.

F

3. Ask Why this outcome occurs (or could occur)?

G

4. Select one of the reasons for the Why in number 3 and ask again: Why does that occur? or Why does this situation cause the problem?

H I J

5. Continue asking why until the potential root cause is reached and enough information is gathered, such that it is actionable.

K L N

How to Analyze and Apply the Tool’s Output • Review the information gathered and develop an appropriate action plan to remedy the root cause.

O

• Document the lessons learned so that the error is not repeated in the

M

P

future.

Q R S T U V W X Y Z

Examples A pricing analyst fails to notify management of an error in last month’s report… • Why was the error not reported?

• Because I didn’t know it was my responsibility. • Why did you not know it was your responsibility?

• Because we implemented a new process last month, and I presumed that finance was responsible. • Why did you think finance was responsible?

• Because finance was responsible when we used the former procedure.

5-Whys

307

• Why did you not understand the procedure change?

• The change was not documented in the new procedure, and it was not discussed in the transition training workshop.

Hints and Tips

A

The number “5” is not a magical number. Stop whenever a potential

B

root caused is reached, and the team can act on the information gath-

C

ered. And as information is gathered, record any potential solution

D

ideas, but do not stop the inquiry process until the root cause is

E

reached and understood. Documentation of findings is important; the

F

responses can be recorded as a list, Tree diagram, Fishbone diagram,

G

or in table format. People are seldom the root cause of a problem—

H

often their behavior is caused by something (poorly documented or

I

communicated procedure, response to a special cause situation, and

J

so on). Continue to ask why even if a person or role is identified.

K L M

Supporting or Linked Tools Supporting tools that might provide input when using the 5-Whys include • Histogram (See Also “Histogram—7QC Tool,” p. 330)

N O P

• Pareto chart (See Also “Pareto Chart—7QC Tool,” p. 445)

Q

• Process map (See Also “Process Map (or Flowchart)—7QC Tool,”

R

p. 522) • VOC and VOB data (See Also “Voice of Customer Gathering

Techniques,” p. 737) • Brainstorming (See Also “Brainstorming Technique,” p. 168) • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram—

7QC Tool,” p. 173) • 5MandP technique and its variants (See Also “Cause-and-Effect

Diagram—7QC Tool,” p. 173) A completed 5-Whys provides input to tools such as • Brainstorming (See Also “Brainstorming Technique,” p. 168) • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram—

7QC Tool,” p. 173)

S T U V W X Y Z

308

Encyclopedia • 5MandP technique and its variants (See Also “Cause-and-Effect

Diagram—7QC Tool,” p. 173) • QFD (See Also “Quality Function Deployment (QFD),” p. 543) • Other root cause analysis techniques (See Also “Hypothesis A B C D E F

Testing,” and “Regression Analysis” p. 335 and 571, respectively.) • Concept generation methods (See Also “Failure Modes and Effects

Analysis (FMEA)” “Hints and Tips” section, p. 287) • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287)

Figure F-9 illustrates the link between the 5-Whys and its related tools and techniques.

G H

Cause and Effect Diagram

I J

5M and P Technique and its variants

Histogram

QFD

Pareto Chart

Root Cause Analysis Techniques

K L M N O

5-Whys Technique

Concept Generation Methods

Process Map

P Q

VOC and VOB Data

R S T U V W X Y Z

Brainstorming Technique

FMEA

Figure F-9: 5-Whys Tool Linkage

Variations 5W2H This uses the basic 5-Whys technique, but the following questions are asked instead of the five whys: 1W. Who? 2W. What? 3W. When? 4W. Where? 5W. Why?

Fault Tree Analysis (FTA)

309

1H. How? 2H. How much? The 5W2H technique is often used when analyzing an improvement solution or when planning a new product, service, process or project initiation, or design and implementation.

A

The how and how much questions could be referring to cost or time or how many resources.

B

Cause-and-Effect Diagram; Ishikawa Diagram; Fishbone Diagram The tool helps to organize and begins to analyze potential sources of a problem. (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173)

D

Fault Tree Analysis This tool uses a top-down analytic method and tree diagram structure with unique symbols to dissect a specific problem or problematic outcome into its related causes. (See Also “Fault Tree Analysis (FTA),” p. 309)

Fault Tree Analysis (FTA)

C E F G H I J K L M

What Question(s) Does the Tool or Technique Answer? What are the potential root causes of a single problem or problematic outcome, and how can they be prevented or mitigated? (Typically applied to a design of a system, product, or process that involves human interaction.) Fault Tree analysis helps you to • Quickly focus on discovering the root cause to prevent a single failure • Understand the relationship between a problem and its different

causes, to develop proactive risk mitigation strategies

N O P Q R S T U V

Alternative Names and Variations This tool is also known as • FTA

Variations on the tool include • Failure Modes and Effect Analysis (FMEA) (See Also “Failure

Modes and Effect Analysis (FMEA),” p. 287) • Cause-and-Effect diagram; Ishikawa diagram; Fishbone diagram

(See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173) • 5-Whys (See Also “5-Whys,” p. 305)

W X Y Z

310

A B C D

Encyclopedia

When Best to Use the Tool or Technique Before any action is taken, this tool helps identify the source(s) of problem, rather than the symptom, and categorizes them to assist in developing preventative strategies. This technique is best done early in the design stage, examining the robustness of a design (typically of a system or program). FTA is a narrow analysis, used to study one failure at a time. Often it used in conjunction with the development of an FMEA when trying to analyze and prioritize multiple problems and their root causes, rather than superficial symptoms. It is particularly useful when the problems involve human behavior or interaction.

E F G H I J K L M N O P Q R S T U V W X Y Z

Brief Description The Fault Tree Analysis (FTA) uses a top-down analytic method and Tree diagram structure to dissect and define the relationship of related causes of a specific problem or problematic outcome, using a unique set of classification symbols. The approach starts with a failure (either potential or existing) and probes backward toward the fundamental events or root causes. This FTA approach is frequently referred to as a top-down systems analysis. In contrast, the FMEA analyzes an entire process or system to understand the causes of many potential failure modes and employs a bottomsup method, starting with potential failure modes and then determining the possible consequences of a given failure, if it occurs. An FTA can be used within an FMEA process. Using a vertical “tree” structure, constructed from top-to-bottom, the FTA graphically drills down into what could go wrong or what has gone wrong. Its unique symbols set it apart from other approaches that analyze root causes of failures or potential failures. The FTA unique symbols divide into two categories: 1) Event Symbols and 2) Gate Symbols. The symbols found next are the most commonly used; however, there are additional symbols for more complex situations. The Event symbols represent a dissection of elements describing what has or could occur and are hierarchically displayed from the key problem being analyzed. The Gate Symbols represent logical operators that link higher level and lower level elements in a hierarchical depiction of events. [Symbols produced using Microsoft Office Visio software.] The Fault Tree analysis symbols include

Fault Tree Analysis (FTA)

Event Symbols Event that can be analyzed into more basic causes

Gate Symbols House or ‘Switch’ Event between either occurring or not occurring

Basic Event not a result of other causes (Root cause)

311

Conditional Event used with inhibit gate

AND Gate (Input events occur at same time)

Priority AND Gate (Input events occur in order from left to right)

A OR Gate (Any one event occurs)

Exclusive OR Gate (One and only one input event occurs)

B C D E F

Undeveloped Event due to insufficient data or insignificance

Transfer Symbol, links to another place in diagram

Inhibit Gate (Input event occur when conditional event occurs)

Voting Gate (or Sample Gate) (If “m” out of “n” input events occur)

G H I J

How to Use the Tool or Technique When using a Fault Tree analysis Step 1.

Step 2. Step 3.

Step 4.

Identify the key failure event to be studied, often called a primary or top event, including the boundaries that limit the analysis. Draw a rectangle at the top center and label it with this key failure event. Examine the system, design, or process and identify related events and elements to the key failure event being analyzed, using the appropriate background information and documentation, particularly a Process map or flow diagram if studying a process. Starting with the primary failure event, construct a hierarchical relationship of related events and elements. a. For each event, determine if it is a basic failure, or if can it be analyzed further for its own causes. i. If it is a basic failure, draw a circle around it. ii. If it is not a basic failure, draw a rectangle around it. iii. If appropriate, use the other symbols available to better define this element, such as an undeveloped event.

K L M N O P Q R S T U V W X Y Z

312

Encyclopedia

b. For each event, determine how it is related to the subsequent events that it causes (in the hierarchical flow). Select the appropriate Gate Symbol for each event and related causes. i. The lower-level events are the input events. A

ii. Input events cause output events and are placed above in the hierarchy.

B C

iii. The input and output events are linked by gates and are placed in between them.

D E

c. Repeat Steps a. and b. until all the tree branches depict basic or undeveloped events at its outer-most end.

F G

d. OPTIONAL STEP: Determine the probabilities of each basic or undeveloped event and mathematically calculate the probabilities of each higher-level event and the top event. (Supporting software can assist in this calculation.)

H I J K

Step 5.

L M

Risk Analysis: Analyze the Fault Tree diagram to understand its multiple relationships and define strategies to prevent potential events that lead to failures.

P

a. Use the Gate Symbols to assist in determining the type of relationship and efficiently determine appropriate preventative strategy. (Reference the following Hints and Tips section for suggested strategies for each Gate Symbol.)

Q

b. Focus first on the most likely to occur events.

N O

R S T

Step 6.

Document these strategies in a contingency action plan, describing what to do if a failure occurs and who is accountable for taking action.

U V W X Y Z

How to Analyze and Apply the Tool’s Output Take action to prevent possible risk or respond to the occurrence of one. This is when the value of the FTA comes to light. (See Also “Failure Modes and Effects Analysis (FMEA),” “Hints and Tips” section, p. 287, for additional detail on risk analysis and management.) Also periodically refresh and update the document as part of an evergreen process. • Communicate and distribute the FTA’s contingency action plan to appropriate parties.

Fault Tree Analysis (FTA)

• Incorporate the FTA and its contingency action plan into the control plan. Document any lessons learned after action is taken and update the FTA and its contingency action plan to prevent similar events in the future.

313 A B C D E

Examples Sample FTA Figure F-10 illustrates an FTA about a sales prospect who refuses to see a sales representative. Prospect refuses to see sales representative

F G H I J K L M

Prospect doesn’t want the offering

Prospect Busy

N O P

Calendar is full; rep reschedules

Claims calendar is full; rep cannot reschedule

Prospect doesn’t understand value of offering

Prospect doesn’t need offering

Prospect brought competitive offering

Rep gives up on account

R S T

Doesn’t want offering

Rep schedules with someone else in the organization

Q

U V

Rep didn’t leave literature with value proposition and testimonials

Prospect didn’t read literature with value proposition and testimonials

W

1

2

X Y Z

Figure F-10: Sample Fault Tree Analysis

Hints and Tips Potential Prevention Strategies for Each Gate Symbol AND Gate: Eliminate at least one of the input events (given that they all must occur at the same time) to prevent the outcome from occurring.

continues

314

Encyclopedia

OR Gate: Eliminate as many of the input events as possible (because any one of the input events could occur). Best to prioritize contingency actions on the most likely event to occur first. The more A B C D E F G H I J K L M N O P Q R S T

input events that are prevented, the more the overall outcome risk is minimized. Inhibit Gate: Eliminate either the input event or the related conditional event to prevent the outcome from occurring. Priority AND Gate: Eliminate any of the input events to disrupt the sequence of input events from occurring, thus preventing the subsequent outcome from occurring. Another strategy is to change the sequence of input events, considering the occurrence of the resulting outcome depends on the set sequence occurring. Exclusive OR Gate: Similar to the recommended “OR Gate” strategy, eliminate as many of the input events as possible (because any one of the input events could occur). It’s best to prioritize contingency actions on the most likely event to occur first. Given the condition of this gate dictates “one and only one” input event to occur, and not more than one to cause an output failure event, the strategy also might be to ensure that MORE than ONE input event occurs to prevent the overall outcome event from occurring. Voting Gate: Starting with the most likely input event to occur, eliminate the input event until only “m-1” events remain (one less event than “m”), to prevent the outcome from occurring.

U V W X

Supporting or Linked Tools Supporting tools that might provide input when developing a Fault Tree Analysis include

Y

• Histogram (See Also “Histogram—7QC Tool,” p. 330)

Z

• Pareto chart (See Also “Pareto Chart—7QC Tool,” p. 445) • Process map (See Also “Process Map,” p. 522) • VOC and VOB data (See Also “Voice of Customer Gathering

Techniques,” p. 737) • 5-Whys (See Also “5-Whys,” p. 305)

Fault Tree Analysis (FTA)

315

• 5MandP technique (See Also “Cause-and-Effect Diagram—7QC

Tool,” p. 173) • Brainstorming (See Also “Brainstorming Technique,” p. 168)

A completed Fault Tree analysis provides input to tools such as • Brainstorming (See Also “Brainstorming Technique,” p. 168)

A

• QFD (See Also “Quality Function Deployment (QFD),” p. 543)

B

• Root cause analysis techniques (See Also “Hypothesis Testing” and

“Regression Analysis,” p. 335 and 571, respectively.) • Concept generation methods (See Also Failure Modes and Effects

Analysis (FMEA) in “Hints and Tips” section, p. 287) • FMEA (See Also “Failure Modes and Effect Analysis (FMEA),” p. 287)

Figure F-11 illustrates the link between a Fault Tree analysis and its related tools and techniques.

Histogram

D E F G H I J K

5M and P Technique and its variants

5-Whys Technique

C

L QFD

M N

Pareto Chart Fault Tree Analysis

O P Q

Concept Generation Methods

Process Map

VOC and VOB Data

Root Cause Analysis Techniques

R S T

Brainstorming Technique

FMEA

U V

Figure F-11: Fault Tree Analysis Tool Linkage

W X Y

Variations Failure Modes and Effect Analysis (See Also Failure Modes and Effects Analysis (FMEA), p. 287) • Uses a bottoms-up analytic method to identify potential failures of a

system, product, or process and defines their related causes (effects) and action plan to minimize or eliminate risk.

Z

316

Encyclopedia • FMEA often is applied to an entire system, product, or process, and

the FTA can be used within it.

Cause-and-Effect Diagram; Ishikawa Diagram; Fishbone Diagram (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173) A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

• The tool helps to organize and begins to analyze potential sources of

a problem.

5-Whys (See Also “5-Whys,” p. 305) • This technique asks the question why to drill down and identify the

potential root cause and prevent being satisfied with “symptoms,” which leads to superficial solutions.

Fishbone Diagram—7QC Tool See Also “Cause-and-Effect Diagram,” p. 173.

Flowchart—7QC Tool See Also “Process Map (or Flowchart)—7QC Tool,” p. 522.

Gantt Chart

317

G Gantt Chart A

What Question(s) Does the Tool or Technique Answer? How long will the project take? What are the key milestones, and when should you expect them? A Gantt chart helps you to • Organize the project’s activities. • Display and communicate the project’s planned activities and key

milestones, as well as its status on the actual progress to-date.

B C D E F G H I

Alternative Names and Variations This tool is also known as

J K

• Milestone chart

L

• Project bar chart or (misnomer as “project plan”)

M

• Activity chart

N O

When Best to Use the Tool or Technique The Gantt chart tool is an important project management tool and is used throughout a project. Initially, a Gantt chart is developed during the planning phase of a project and is used to monitor progress throughout the duration of a project. A Gantt chart summarizes and communicates the planned project activities and the progress to complete them, thus is included in review meetings and status reports. It is a good summary tool to document the project schedule and file with the project “close-out” documentation as well.

P Q R S T U V W X

Brief Description The Gantt chart, named after Henry Gantt, is an effective project schedule summary communication tool. It displays project activity as a single horizontal bar chart, typically with time as the horizontal-axis and the planned activities as the vertical-axis. The left-most edge of the bar defines the start date, and right-most edge defines the end date of an activity. The portion of the bar containing unique shading represents the progress to-date, status, or percentcomplete of a given activity. Often a diamond-shape or similar symbol indicates a key milestone. The person held accountable for an individual activity may be documented adjacent to the respective bar on the chart as well.

Y Z

318

Encyclopedia

A Gantt chart is simply constructed and easy to understand. The tool is easy to use, and it quickly conveys the important project management information to an executive level. The critical path can be indicated on the chart by bolding its associated activity-bars. A B C D E F G H

Tools often confused with the Gantt chart are the Activity Network Diagram (AND) and PERT charts. If further analysis of bottlenecks and interdependencies is required, the AND is a more appropriate tool. A traditional Gantt chart lacks information about any interdependency between and among tasks; however, current software packages expand the original design with the important information. The PERT and CPM charts better reflect any activities’ linked relationships. The PERT, AND (or CPM), and Monte Carlo simulations are well-suited tools for predicting variations in expected completion time of a project. (See Also “Activity Network Diagram (AND)—7M Tool,” “PERT (Program Evaluation and Review Technique) Chart” and “Monte Carlo Simulation” p. 127, 453, and 431, respectively, for more details.)

I J K L M N

Example There are several software packages that can be used to build a Gantt chart, such as Microsoft Project, Primavera, or Open Workbench, and one of the newest entries to the market is Minitab Quality Champion® used to plan and manage a project. Figure G-1 displays a simple Gantt chart using Minitab Quality Champion® application software.

O P Q R S T U V W X Y Z Figure G-1: Sample Gantt Chart (from Minitab Quality Champion®)

Notice in the Figure G-1 example that each bar represents a cluster of activities (or phases). The open, or un-shaded, bars represent planned phases, and the shaded portion of the top two bars depicts that portion of

Gantt Chart

319

the project already completed. Hence, this sample project shown in Figure G-1 is close to completing the Measure phase of the five-phase project.

Supporting or Linked Tools Supporting tools that might provide input when developing a Gantt chart include • VOC (See Also “Voice of Customer Gathering Techniques,” p. 737)

A B C

• Activity List

D

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

E

• Activity Network diagram (See Also “Activity Network Diagram

F

(AND)—7M Tool,” p. 127) • Monte Carlo Simulation (See Also “Monte Carlo Simulation,”

p. 431) A completed Gantt chart provides input to tools such as • Activity Network diagram (See Also “Activity Network Diagram

(AND)—7M Tool,” p. 127) • Monte Carlo Simulation (See Also “Monte Carlo Simulation,” p. 431)

G H I J K L M N

• Project plan and status report

O

• Project close-out documentation

P

Figure G-2 illustrates the link between a Gantt chart and its related tools and techniques.

Q R S T

VOC Project Plan and Status Report Gantt Chart

Activity list

U V W

Project Closeout documents

X Y

Brainstorming

Z Activity Network Diagram

Figure G-2: Gantt Chart Tool Linkage

Monte Carlo Simulation

320

Encyclopedia

GOSPA (Goals, Objectives, Strategies, Plans, and Actions)

A

What Question(s) Does the Tool or Technique Answer? How do the activities of a function, program, or project align with the overall organizational strategy?

B

GOSPA helps you to

C D E F G

• Develop a planning methodology used to align actions with organi-

zational direction. • Communicate direction and synergy with organizational goals. • Audit specific activities to ensure linkage with strategic direction.

H I J K

When Best to Use the Tool or Technique Use the GOSPA technique at the beginning of the planning phase of a program or project and monitor its progress against the document throughout execution of a given plan.

L M N O P Q R S T U V

Brief Description GOSPA is a top-down planning technique that starts with the highestlevel organizational goals and defines how individual groups support specific objectives. GOSPA is a simple acronym that stands for Goals, Objectives, Strategies, Plan, and Actions. This planning methodology summarizes an organization’s direction and defines how supporting programs and/or projects relate to it. In the context of Six Sigma for Marketing (SSFM), business, marketing, sales, and customer value chain professionals rely on the GOSPA technique to link their activities to the overall organizational direction to meet a specific goal. GOSPA helps to ensure that the goals of an individual program (or project) support that of the larger organizational goals.

W X Y Z

How to Use the Tool or Technique Develop a GOSPA document using either a matrix or Tree diagram. The guidelines for using the GOSPA method include 1. Goals—Document the organizational goals using directional, qualitative, broad, and general language. Every goal should have at least one objective; typically it may have two to four objectives. • For example, “Become a global leader in…”

GOSPA (Goals, Obje ctive s, Strategie s, Plans, and Actions)

321

2. Objectives—Identify specific goals that are measurable and provide an aggressive but achievable target. • For example, “$7million growth in sales by 2010…”

3. Strategies—Define the specific approach to achieve the objectives. The strategy links the objective to a plan. Consider the various alternative means to achieve the objective and select the best approach. Anticipate potential issues or changes that might emerge and identify an appropriate contingency plan.

A B C

• For example, 1) Develop new products for current market, 2)

D

Introduce product A into new market segment, and 3) Maintain annual expenditures to prior year’s level.

E

4. Plan—Devise a supporting plan to implement a specific strategy that includes specific tangible elements, involving near-term actions and a measurable timeframe. • For example, 1) Initiate a product development team, 2a) Modify product A and select a new color for the cover and its packaging, 2b) Develop marketing collaterals and an advertising campaign that highlight value proposition to the new market segment, 2c) Train the sales people, 3a) Issue policy memo restating year’s level as the current year expenditure targets, and 3b) Communicate a new presidential recognition program to celebrate those departments that successfully implement process improvements. 5. Actions—Expand the plan with details about who is accountable to do what of the specific plan, by when, why, and for how much (budget). Figure G-3 displays the example GOSPA scenario as a matrix. Figure G-4 uses a Tree diagram to illustrate the multi-dimensional hierarchical relationship that could result from an organization with three goals.

F G H I J K L M N O P Q R S T U V W

Hints and Tips An organization should limit its goals to four or five. Ideally a goal represents two to four different objectives. Goals and strategies are qualitative. Objectives, plans and actions are quantitative. To help visualize the hierarchical supporting linkage, map the relationships using a Tree diagram or matrix.

X Y Z

322

Encyclopedia Goal

Objective

Strategy

Develop new products for current market

A B C

Introduce product A into new market segments

D E F

Become a global leader in commercial batteries.

$7 million growth in sales by 2010

Plan (what)

Actions Who

When

Why

How Much

R. Griggs

Completed by August 2008

New technology available

$1,500,000

Modify product A and select a new color for the cover and its packaging

E. Kullberg

Completed by June 2007

Unique segment requirements and differentiation

$250,000

Develop marketing collaterals and advertising campaign the highlight value proposition to the new market segment

F. DeYoung

Completed by June 2007

Unique segment requirements

$250,000

Train the sales people

B. Skea

Completed by July 2007

Unique segment requirements

$300,000

Issue policy memo restating year’s level as the current year expenditure targets

R. Keller

Completed by January 2007

Maintain year spend

N/A

Communicate a new presidential recognition program to celebrate those departments that successfully implement process improvements

N. Minocha

Completed by January 2007

Implement Six Sigma methodology

$100,000

Initate a product development team

G H I

Maintain annual expenditures to prior year’s level

J K L M N

Figure G-3: Example GOSPA Matrix

O

GOSPA Hierarchical Relationship

P Q

G1

G2

G = Goal O = Objective S = Strategy P = Plan A = Action

G3

R S

O1

T U

S1

V W

P1

X

O2

S2

S3 P3

P2

P4

O3

S5

S4

P5

O4

P6

S6

P7

P8

Y Z

A1

A2

A3

A4

A5

Figure G-4: GOSPA Multi-tiered Relationship

A6

A7

A8

A9

Graphical Methods

323

Supporting or Linked Tools Supporting tools that might provide input when using the GOSPA technique include: • VOB (See Also “Voice of Customer Gathering Techniques,” p. 737) • Brainstorming (See Also “Brainstorming Technique,” p. 168)

A

• SMART (See Also “SMART Problem and Goal Statements for a

B

Project Charter,” p. 665)

C

A completed GOSPA matrix provides input to tools such as:

D

• SMART (See Also “SMART Problem and Goal Statements for a

Project Charter,” p. 665) • Project Charter (See

Also “SMART Problem and Goal Statements for a Project Charter,” p. 665)

F VOB

Project Charter

Brainstorming

G H

GOSPA FMEA

I J K

• FMEA (See Also “Fail-

ure Modes and Effects Analysis (FMEA),” p. 287)

E

SMART

Figure G-5: GOSPA Tool Linkage

Figure G-5 illustrates the link between a GOSPA matrix and its related tools and techniques.

L M N O P

Graphical Methods

Q R

What Question(s) Does the Tool or Technique Answer? What does the data look like? Graphical methods help you to

S T U V

• Describe what the data looks like (good communication tool).

W

• Identify trends and stability over time—the patterns or tendencies.

X

• Display the data distribution over its range (the spread).

Y

• Isolate unusual observations (often called “special cause” variation). • Detect relationships among variables. • Support business decisions, by showing the data in a graph or chart.

Z

324

Encyclopedia

Types of Graphical Methods Figures G-6 and G-7 show the various types of graphs, interspersed with a list of graphical methods: • Bar chart—A category or type of graph, wherein the height of a bar A B C D E F G H I J K

represents frequency of occurrence in summary or interval groupings. Bar charts can be individual, multiple (or clustered bars), or stacked bars representing groups of data. • Boxplot (or Box-and-Whisker Plot)—Shows summary frequency in

quartiles. (See Also “Boxplots—Graphical Tool,” p. 165) • Dotplot (or Dot graph) —Shows frequency in detail. (See Also “Dot-

plot,” p. 280) • Histogram (a type of bar chart)—Used with quantitative, continuous

data to show summary frequency of interval or ratio data. (See Also “Histogram—7QC Tool,” p. 330) • Line graph—A category of graphs that shows the “connected” fre-

quency as a single line depicting the area under the curve of a single variable, along some order (time or distance for example).

L M N O P Q

Bar Chart

Boxplot

Dotplot

Histogram

Line Graph & Run Chart

Multi-Vari Chart

R S T U V W X

Figure G-6: Examples of Graphical Methods—Part 1 of 2

Y Z

• Multi-vari Plot—Similar to a scatter plot; displays the relationship

between variables, but this diagrams one variable’s relationship (on the y-axis) with many other variables (depicted on the x-axis). (See Also “Multi-vari Chart,” p. 439) • Pareto chart—A specialized bar chart that shows frequency and

accumulation of nominal data. (See Also “Pareto Charts—7QC Tool,” p. 445)

Graphical Methods

325

• Pie chart—Used to show proportions often in business reports and

presentations (but seldom in Six Sigma). • Run chart—A type of line graph that shows a variable over time

(time sequenced). It also is sometimes called Trend chart, Time Series plot, or Control chart. (See Also “Run Chart—7QC Tool,” p. 611) • Scatter Plot—Shows the relationship between the distributions of

two different numeric variables—whether or not they are correlated. (See Also “Scatter diagram—7QC Tool,” p. 640) • Stem-and-Leaf—Displays the distribution of a sample as a set of

numbers grouped by tens. Also called a stem-plot.

C D E

Carrier Change $928,042

G

Timing ($475,324)

H

Std Cost Credit $143,805

I

Model/Lane Change ($46,909)

Fuel Surcharge $951,749

J

Incorrect Account Used $584,757

Pareto Chart

6 0 0 2 4 0 3 1 0 5

B

F

Volume $564,888

0 1 2 3 4 5 6 7 8 9

A

Pie Chart

K

Scatter Plot

L M

1 0 5 4 0 4 5 2

3 0 7 5 0 5

4 0 8 6 1

5 1

5 1

6 2

6 6

6

6

6

7

7

7

7

7

8

8

9

6 2

7

7

7

7

7

7

7

8

8

8

8

8

8

8

N 9

9

9

O P Q

Stem & Leaf Chart

R

Figure G-7: Examples of Graphical Methods—Part 2 of 2

S T U

When Best to Use the Tool or Technique The old adage holds true, a picture is worth a thousand words. The graphical method is the picture that visually displays the distribution of a set of data or compares multiple sets of data.

V W X Y Z

Brief Description The terms graph and chart often are used interchangeably. The rule of thumb is, “The simpler the better.” Graphs are an effective way to present data, and they assist in communicating key causes and potential recommendations.

326

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

The type of data collected defines the type of graph used to depict the data. Data can be categorized into two primary classifications—either discrete or continuous data. Continuous data provides the most amount of information because it has limitless number of possible values between two given contiguous values. Continuous data is said to have no boundaries between values. It can be divided by two and still make sense. Often times, collecting and measuring continuous data requires more effort (time and money) to study compared with attribute or discrete data. Examples of both kinds of data include • Continuous or variable data—A quantitative type of data that can

take on a range of numeric values and is comprised of “non-counting” intervals such as time, height, weight, length, and temperature. • Discrete or attribute data—A type of data that has clear boundaries

between values. There are several types of discrete data: • Binary—A qualitative or categorical type of data made up of two classifications • With a value rank or order—For example, yes/no, pass/fail, good/bad, on/off, agree/disagree, and good/better/best. • Without a value rank—For example, male/female and heads/tails. • Counts—A type of “quantitative” discrete data containing rank order or ordinal numbering (treated as a numeric scale of intervals). Number of defects (integers), days in the week, age in only years, number of wins or losses, income, and proportions are examples. • Nominal or categorical data—Attribute data that is descriptive, not numeric, with more than two categories, for example, names, phone numbers, colors, type of car, capital cities, and states. This type of attribute data also can be equated to discrete values— for example, to distinguish among people, such as sales representatives, but maintain some anonymity, they could be coded and referred to a discrete number whereby Sally = 1, John = 2, Neeta = 3, and Vladimir = 4. Several tools are listed in Table G-1, which summarizes the rule of thumb guidelines to select which type of graphical method to use when.

Graphical Methods

327

Table G-1: Rule of Thumb Summary of Graphical Methods Purpose

Boxplot

Depict a sample’s freContinuous quency distribution as Quantitative a summary box of quartile data. Used to identify the location of median, the spread (range), and the shape (skewed) of the distribution.

16-49

Depict a sample’s frequency distribution, showing the detail of each data point. Used to identify the location of mean, median and mode, the spread (standard deviation, variance, range), and the shape (bell-shaped, skewed, binomial) of the distribution.

Histogram > Simple….

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

332

Encyclopedia

Figure H-1 displays sample MINITAB screens where the appropriate graph type is selected (“Area 1”) and the appropriate variable data of interest (for example, Yield) are selected (“Area 2”) to produce the final sample histogram found in Figure H-2. A B C D E F G H I J K

Figure H-1: Example MINITAB Histogram Main Screen

L

Histogram of Yield

M

10

N O Q R S

Frequency

P

8 6

4

2

T U

0 76

80

V W X Y Z

84 Yield

88

92

Figure H-2: Example MINITAB Histogram

Based on the Figure H-2 graph, one could expect the process to produce a yield in the mid-80s, with it ranging from the upper-70s to lower90s. If a yield outside of this range occurred, we might investigate whether a special cause was present. The histogram appears to be fairly normally distributed, and we would suspect that as more data were gathered, the graph would become more bell-shaped. (See Also “Control Charts—7QC Tool,” in the “Normal versus Non-normal Data” section for a discussion on statistical tests of normal distributions, p. 217)

Histogram—7QC Tool

How to Analyze and Apply the Tool’s Output A histogram that approximates a single, uni-modal, bell-shaped distribution is said to be a “normal”, as the data are normally distributed. By chance alone, the data points are equally likely to occur on either side of the mean, such that the distribution is symmetrical about the mean. Bellshaped symmetry contains a mean, median, and mode that equal one another. A normal histogram represents a stable, predictable process. A histogram shape with a particularly large “spike” in one of the bars typically represents a special cause variation and should be investigated. Special cause events should be identified and eliminated. A binary outcome produces a bimodal distribution, such as “yes/no” or “fail/pass.” Bimodal distribution also could represent a segmented data set that might best be stratified, such as a two-shift operation. A bimodal histogram also may represent a stable process. A skewed distribution represents a particularly long tail on either side of the mode—an asymmetrical curve. A negatively (or left) skewed curve features a long tail toward the “negative” side of the peak—off-center. The opposite is true for a positively skewed curve, which has a tail trailing off to the right of the peak. Distributions with a few extreme data points produce a skewed-shaped graph. An example often sited is the average home price in the United States with a few homes priced over $1 million. If a process variation exceeds the upper and lower specification limits (that is, USL and LSL), then the process requires improvement modifications to reduce the variation. Inspection or audits may artificially cause the process to “squeeze” within the specification limits, thereby producing a truncated histogram with its tails being “artificially” cut off. Figure H-3 features various histogram shapes.

333

A B C D E F G H I J K L M N O P Q R S T U V

Normal Histogram

W

Histogram with special cause LSL

USL

X Y Z

Bimodal Histogram

Negatively Skewed Histogram

Truncated Histogram

(may also be polymodal)

(skewed left)

(e.g. after inspection)

Figure H-3: Examples of Histogram Shapes

334

Encyclopedia

Hints and Tips Histograms plot numeric data, and their bars are contiguous or adjacent to one another because their x-axis scale is continuous. A B C D E F G H I J K L M N O P Q R S

Bar charts plot categorical data and are depicted with “gaps” in between the bars because the x-axis contains the categories—non-continuous data. A Pareto is a special type of bar chart. The histogram’s visual approach allows estimation of • Central tendency—mean, median, mode • Spread—range, standard deviation, variance • Shape—bell-shaped or not (skewed or bi-modal); unusual variation or truncation • Kurtosis—a measure of the amount that the middle of the curve is flattened or peaked. A bell-shaped (so called “normal”) histogram characterizes a stable process in that the process is predictable. If the histogram appears to be distributed normally, it should be double-checked statistically for more information. If using MINITAB, its Normal Probability Plot or Descriptive Statistics applications provide the necessary statistics to verify statistical normality. If the data are normal, a larger number of statistical analytical tools become options as a “next step.” (See Also “Control Charts—7QC Tool,” in the “Normal versus Non-normal Data” section for a discussion on statistical tests of normal distributions, p. 217)

T U V W X Y Z

Supporting or Linked Tools Supporting tools that might provide input when creating a histogram include • Data gathering plan to collect the appropriate metrics (See Also

“Data Collection Matrix,” p. 248) • Performance charts and dashboards • Monte Carlo simulation, the probability outcome distributions dis-

played as a histogram (See Also “Monte Carlo Simulation,” p. 431)

Hypothe sis Te sting

335

A histogram can provide input to tools such as • Hypothesis testing, examining the frequency of sample of data (See

Also “Hypothesis Testing,” p. 335) Figure H-4 illustrates the link between a histogram and its related tools and techniques.

B

Data Gathering (metrics) Performance Charts and Dashboards

A C D

Histogram

Hypothesis Testing

E F G

Monte Carlo Simulation

H I

Figure H-4: Histogram Tool Linkage

J K

House of Quality (HOQ) See Also Quality Function Deployment (QFD), p. 543

Hypothesis Testing

L M N O P Q

What Question(s) Does the Tool or Technique Answer? Is this population (or sample) of data different from another by chance alone or because of an outside influence? Hypothesis testing helps you to • Determine if one population (or sample) of items is statistically dif-

ferent from another. • Understand if there is a statistical difference between two sets of

data, either due to common cause (or chance alone) or because something has influenced the process. • Use as a Root Cause Analysis technique. • Infer whether two sets of data are similar or different; determining if

there is a relationship between them or not. Identify if one set of data is related to or is different from another set of data.

R S T U V W X Y Z

336

Encyclopedia • Make decisions…

• Decide if the “improvement” made to the product or process produced a statistically different result and is it worth the investment to continue to implement it. A B C D E F G

• Determine if the different defect types are independent of when in the production process they occur or if the defect categories are dependent on the process variation from day-to-day. • Make inferences…

• Make a prediction about a population based on sample data. For example, based on the feedback from a sample audience, make changes in an advertising campaign to better segment the marketing message.

H I J K

Alternative Names and Variations This technique uses several different statistical tests. Some of the more common ones include

L

• Chi-Square Test

M

• Student t-Test, or “t-Test”, or Paired t-Test

N O P

• Z-test • F-Test or ANOVA (Analysis of Variance)

Q

(Table H-1 provides a more complete list of the different tests.)

R

Variations on the tool include

S T U

• A type of “inferential” statistics, wherein collected data is organized,

analyzed, and interpreted so as to make predictions about a population or relationship.

V W X Y Z

When Best to Use the Tool or Technique Hypothesis testing generally happens at two different times. First, after data has been collected, a hypothesis emerges as to the root causes (the critical variables) behind why the process or product performance is what it is. More questions arise to understand what the key process input and output variables (KPIV and KPOV) are. Second is after a process or product modification has been made and the changed state is hypothesized to be improved from the prior state.

Hypothe sis Te sting

Brief Description Hypothesis testing primarily answers the question of cause-and-effect relationship—a Root Cause Analysis technique. Sir Ronald Fisher, an early 1900s scientist and mathematician, developed several statistical principles to conclude causal relationship with a degree of confidence (or significance). His work established the alpha-level (α-level) standard (a default of 0.05) to represent the point beyond which the hypothesis gets accepted if supported by sufficient evidence. The larger the sample, the more confident one can conclude about the hypothesis. However, the practicality of time and economics often restricts the sample size. Hypothesis testing takes this balance into account by determining the amount of acceptable error. Hypothesis tests use information contained in sample data to make conclusions about process or population characteristics. A formed theory, or hypothesis, may question whether two sets of data are related to one another; thus, it is part of the same population. If the theory tries to show they are different, generally it is because of chance alone or because one set of data had something influence it, causing a change. For example, a theory about whether soybeans grown in Asia are statistically different from soybeans grown in the United States. Do they have different genetic profile? Do the soil, climate conditions, and growth enhancers from the different regions influence how human digestion breaks down and uses its chemical components? Hypothesis testing can help answer these questions. Another example is a theory comparing two different processes with different timing and procedure for harvesting and cooking soybeans to see which process yields the most statistically nutritious by-products. Which process is better? Were the improvements worth the investment for a national rollout? Does the new process produce a better yield? Again, Hypothesis testing can help to answer these questions. Hypothesis testing follows similar logic to that of the U.S. judicial system when drawing conclusions about evidence. A defendant is presumed innocent until proven guilty. The prosecutor carries the burden of proof to produce sufficient evidence (data) to reject the presumption of innocence. The presented evidence must be “beyond a reasonable doubt” to be considered valid. If sufficient evidence exists, a “guilty” verdict results. However, if the defendant is not declared “innocent,” the final verdict merely states that there was “insufficient evidence” to convict. In terms of Hypothesis testing, the presumption of innocence is called the Null hypothesis.

Hypothesis Testing Terminology Hypothesis testing utilizes a variety of special terms to describe its principles, including the following: • Hypothesis—A theory requiring proof to declare if true or false.

337

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

338

Encyclopedia • Hypothesis test—One of the statistical tests to declare if a theory is

true or false. The appropriate test to use is based on both the type of data available and kind of decision required. Each test has a unique formula to evaluate the test frequency distribution relative to the calculated test statistic (or p-value). A B C D E F

If the area under the curve beyond the test statistic is smaller than the significance level (alpha), then the test statistic probably did not come from the known distribution. Figure H-5 shows the Alternative Distribution’s test statistic (its mean or µa ) to the right of the upper confidence limit indicating that p-value is less than the alpha. Examples of common statistical tests include z, t-test, F and Chi-square. (See Also “Analysis of Variance (ANOVA)—7M Tool,” for more details, p. 142)

G H I J

Note Alpha equals the shaded area to the right of the upper confidence limit.

K L M N O P Q R S T

• Test statistic—A calculated number that summarizes sample data

and used to analyze an hypothesis. Compare the test statistic to the appropriate hypothesis test’s critical value (or “t-critical”) to decide to reject or not reject the Null hypothesis. Examples of a statistic include average (or mean), variance, and proportion. • Null hypothesis (H0)—A statement of what typically is desired to

be disproved through sample data. In general, the statement about two sets of data (or a sample data set versus a population) are said to exhibit “no difference due to chance alone,” denoted by the symbol “H0” and illustrated in Figure H-5.

U V

Ho: Null Distribution

Ha: Alternative

1-

1-

(Confidence Level)

(Power)

Distribution

W X Y Z Type II Error ( )

Type I Error ()

a

0

Upper Confidence

Figure H-5: Generic Hypothesis Testing Schematic

Hypothe sis Te sting

339

• Alternative hypothesis(Ha)—A statement of what is desired to be

concluded through sample data. The alternative statement is dependent on the Null hypothesis. Hence, a “true” alternative hypothesis statement comes about only if the Null hypothesis is proven false. It implies that the two sets of data (or a sample data set versus a population) exhibit a difference due to a “real effect,” not chance, denoted by the symbol “Ha.” • If the Null is true, the correct wording is that there is “insufficient evidence to reject the Null hypothesis” or we are “unable to reject” it. • If the Null is false, the correct wording is, “there is sufficient evidence to reject the Null hypothesis.” Figure H-5 illustrates the two distributions contrasting a “generic” Null and alternative hypothesis.

• Confidence interval—The probability range within which the popu-

lation data (parameters) exists. This region can be described two ways, either a single limit (an upper or lower) or a two-sided limit (both an upper and lower). Outside these limits is defined as the rejection region or critical region, wherein the “test variable” is said to not be part of the population. Sometimes these limits are referred to as the t-critical as they demarcate the rejection boundaries. Figure H-5 illustrates a one-tail upper limit. A commonly used confidence interval is 95%, wherein 95 times out of 100 expect to be right, versus five times out of 100 expect to be wrong. This means that if sample data (or test statistic) were drawn from a population distribution, there is a 95% confidence that the test statistic will be contained within this range. • One versus two-tailed tests—Hypothesis statements are worded as

pairs—the Null and alternative hypothesis—to represent the probability of all the possible outcomes. In general, the Null hypothesis contains an “equal to” statement, and the alternative statement is the theory to be proven. One-tail test—Seeks to answer if the test statistic is less than or greater than the known distribution. An alternative hypothesis stated as a “greater than” test is a right-tail test. Conversely, a “less than” alternative hypothesis statement is a left-tail test. • Right-tail test—H0: Ha mean < H0 mean; Ha: Ha mean > H0 mean. Example of a written word statement: H0: The average speed of Product B is less than or equal to that of Product A; Ha: The average speed of Product B is greater than that of Product A.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

340

Encyclopedia

• Left-tail test—H0—Ha mean > H0 mean; Ha: Ha mean < H0 mean. Example of a written word statement: H0—The average speed of Product B is greater than or equal to than that of Product A; Ha: The average speed of Product B is less than that of Product A. A B C D

Note This left-tail test scenario dictates that bigger is better.

E F G H I J K

Two-tailed test—Merely asks if the test statistic is different from the known distribution; it is indifferent to whether it is larger or smaller. Hence, the alternative hypothesis statement is worded as a “not equal to” and as a two-tailed test because either tail would satisfy the hypothesis. • Two-tailed test—H0—Ha mean = H0 mean; Ha: Ha mean ↑ H0 mean.

L

Example of a written word statement: H0: The average speed of Product B is not different from that of Product A; Ha: The average speed of Product B is (detectably) different from that of Product A.

M N O P Q

Note

R

This two-tailed test scenario dictates that nominal is best.

S T U V W

Figure H-6 illustrates one versus two-tailed tests. One-tail test (Right-tail)

Two-tailed test

X Y Upper Confidence Limit

Z

Lower Confidence Limit

Upper Confidence Limit

Rejection Region

Rejection Region

Fail to Reject H 0

Fail to Reject H 0

t crit ical (= α)

- t crit ical (= α/2)

Figure H-6: Generic One versus Two-tailed Hypothesis Test

t crit ical (= α /2)

Hypothe sis Te sting

341

• Confidence level—Defines making a correct test decision (correctly fail-

ing to reject the Null hypothesis) because the Null is true. The “doublenegative” wording is awkward English, but this language adheres to the probabilistic-nature of hypothesis testing, similar to the U.S. Judicial system. It is determined by the formula one minus alpha (or 1- α). Often the default is set at 95% as an industry standard. Figure H-5 illustrates a confidence level of a generic hypothesis test schematic. • Significance level—The pre-defined level of certainty required to

“test” such that the observed effect (a test statistic or “test factor”) actually caused the response or is different from the known distribution, and not by chance alone, denoted by the term alpha (α). This significance level (alpha) determines the probability that a test statistic is part of the test known distribution—the distribution representing Null hypothesis. This probability relates to the p-value. Often the default alpha = 0.05 (or 5%), but sometimes 0.01 (1%) or 0.10 (10%) are used.

A B C D E F G H I J

• If the hypothesis test requires a one-tail test (wherein only one confidence level is used), then the default alpha is 5%.

K

• If the hypothesis test is a two-tailed test (wherein both the upper and lower confidence intervals are used), the default alpha is evenly split between the two tails + 2.5% (0.05/2 = 0.025).

M

Confidence level, significance level, and confidence interval are related. With the significance level (alpha) set at 5% and the confidence level is 95% (one minus alpha), then the data that falls within the 95% confidence interval are said to have occurred by chance alone (no special cause effecting the outcome).

Figure H-5 illustrates an alpha in a “generic” hypothesis test schematic. • P-value—The p-value serves the decision criteria to interpret a

hypothesis test. A “goodness of fit” statistic represents the probability that a test statistic is part of the known distribution (the Null hypothesis). It is calculated by the area of the distribution beyond the test statistic. The smaller the p-value, the more confidence exists that the effect is real and not by chance alone. If the p-value ≤ 0.05, then reject H0. Hence, the test statistic is located in the tail of the distribution beyond the confidence limit (or t-critical), in the rejection region. This provides sufficient evidence to suggest that the test statistic is not part of the known distribution and the H0 is false. The likelihood of being wrong by rejecting the Null hypothesis is less than 0.05. • The Type I Error is no bigger than 0.05. (Most equate the pvalue with the Type I Error.)

L N O P Q R S T U V W X Y Z

342

Encyclopedia

If the p-value > 0.05, then do not reject H0. Hence, the test statistic is located inside the confidence interval of the known distribution, and not in the rejection region. There is insufficient evidence to suggest that H0 is false, and the test statistic appears to be part of the known distribution. A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

• Perhaps with more evidence (that is, data), rerunning the test may indicate that the H0 is false. However, the additional data (or increased sample size) is needed. (Type II Error is too large or Power is too low.) • Power—Defines making a correct test decision—correctly rejecting

the Null hypothesis because the Null is false. Power represents the ability to detect a difference when one exists. It is determined by the formula one minus beta (or 1- β). Often the default is set at 80%. Figure H-5 illustrates how Power is associated with the Alternative Hypothesis distribution of a generic hypothesis test schematic. • Errors in Hypothesis testing—Hypothesis testing involves both

samples and probabilities; therefore, at times incorrect conclusions are drawn. Two types of hypothesis errors exist, and both represent a different type of risk. The appropriate risk levels are selected prior to testing the hypothesis with sample data—namely the significance level (alpha) and beta. • Type I Error arises when incorrectly rejecting the Null hypothesis, when it is actually true. Thus based on the test statistic, the conclusion rejects the Null hypothesis, but in truth, it should be accepted. Type I Error equates to the alpha (α) or significance level, whereby the generally accepted default is 5%. This error type is also known as producer’s risk because the consequence of this incorrect decision of thinking a part or product is bad causes it to be scrapped or reworked, when in fact the item actually is acceptable. Alternatively, the confidence level for making the correct decision by failing to reject the Null hypothesis, when in fact the Null is true, is calculated by one minus alpha. This can be found in the upper left corner of both the quadrants found in Figures H-7 and H-8. Hence, total probability (p) of correctly deciding that a Nullhypothesis is true (that is, 100% or 1) equals the probability of correctly failing to reject the Null (1 minus alpha) plus the probability of incorrectly rejecting the Null (alpha)—or 1 = (1α) + (α) = confidence level plus the Type I Error.

Hypothe sis Te sting

343

• Type II Error occurs when the test decision incorrectly accepts the Null hypothesis. Based on the test statistic, the final decision fails to reject the Null when it is actually false. Type II Error also is called “beta”(β), and the default is typically set at 20%. This error type is also known as consumer’s risk because the manufacturer ships what it thinks is a good product, but in truth, the customer receives a bad (or defective) product. Alternatively, the Power of making the correct decision to reject the Null hypothesis, when in fact the Null is false, is calculated by one minus beta. This can be found in the lower right corner of both the quadrants found in Figures H-7 and H-8. Hence, total probability (p) of correctly deciding that a Null hypothesis is false (that is, 100% or 1) equals the probability of correctly rejecting the Null (1 minus beta) plus the probability of incorrectly accepting the Null (beta)—or 1 = (1- β) + (β) = Power plus the Type II Error. • Interdependency and Risk Level—In general, the intention is to minimize both alpha and beta. However, given a constant sample size, alpha and beta are inversely related, such that minimizing alpha increases beta. As Figure H-5 illustrates, if the upper confidence level setting moves left to shrink the Type II Error (or beta), the Type I Error (or alpha) increases.

A B C D E F G H I J K L M N O P

Note

Q

Increasing the sample size for a given alpha-level decreases beta.

R

The sample size is impacted by the two error types (alpha and beta),

S

the standard deviation and whether there is a detectable difference

T

when one exists (Power, for example). (See Also Sampling, p. 618)

U V

The scenario (or situation) of the theory helps to prioritize between the two errors. Setting an acceptable level of risk is determined by the seriousness of the error. If the product affects the customers’ health, the default of 20% probably is unacceptable, and this Type II Error would be set much lower. Figure H-7 illustrates the various possible decision outcomes versus the truth, including the two error types and the consequences, using the U.S. Judicial system, and Figure H-8 uses a product scenario.

W X Y Z

344

Encyclopedia

Not Guilty The Truth

A

Not Guilty

B C

Guilty

Correct Decision – Innocent goes free (Level of Confidence)

Wrong Decision – Criminal goes free Type II Error, 

D E

Jury’s Decision Guilty

Wrong Decision – Innocent to jail Type I Error,



Correct Decision – Criminal goes to jail (Power)

Figure H-7: US Judicial System Example of Decision Outcomes

F G H I J K L M N O P

Figure H-8 represents a similar decision outcomes quadrant as Figure H-7 but includes the distribution curves similar to those found in Figure H-5 to show how the test statistic (identified by an “X” on the x-axis) relates to the curves and the different decision options. If the calculated test statistic (X) falls to the right of the “t-critical” (or upper confidence limit) in the tail, then the test results produce a p-value that falls into the rejection region, thereby rejecting the Null hypothesis. Using a generic two-product example for a one-tail test, the scenario theorizes that (the new) Product B works better (faster) than Product A. Knowing that the Alternative Hypothesis should represent the theory, the pair of hypothesis statements for Figure H-8 can each be written one of two ways as a one-tail test:

Q Test Decision Product is bad Product is good

R S T

The Truth

U

Product is good

V W X Y Z

Product is bad

Correct Decision – Good

Wrong Decision –

Product shipped (Level of Confidence)

Good product reworked

Wrong Decision – Bad

Correct Decision –

product shipped

Bad product reworked (Power)

Type II Error, 

Type I Error, 

Producer’s Risk

Consumer’s Risk

Figure H-8: Generic Product Example of Decision Outcomes

Option 1: • Null Hypothesis (Ho): Product A > Product B • Alternative Hypothesis (Ha): Product A < Product B

Hypothe sis Te sting

345

Option 2: • Null Hypothesis (Ho): Product A—Product B = 0 • Alternative Hypothesis (Ha): Product B—Product A > 0 Ho: Product A – Product B = 0

Ha: Product B – Product A > 0

Test Decision Product B works faster (Ha) Product A & B the same (H 0) The Truth

Products A&B the same

Correct Decision – Products A&B the same (Level of Confidence)

B C D E

1- 

Power 1- 

Confidence Level

A =0

x

Type II Error, !

1- 

Power 1- 

Confidence Level

x

 A =0

Test t Critical statistic

Test Statistic

t Critical

Wrong Decision – Products A&B the same

Product B works faster

Wrong Decision – Product B faster Type I Error,  (Producer’s Risk)

A

(Consumer’s Risk)

Correct Decision – Product B faster (Power)

F G H I J

1-  Confidence Level

 A=0

Power 1- 

x

Test t statistic

Critical

1- 

Confidence Level

 A=0

Power 1-  t Critical

x Test Statistic

Figure H-9: Generic Two-Product Example of Decision Outcomes with Distribution Curves

K L M N O

In this two-product scenario, bigger is better, indicating that Product B works faster than A. Presume that the hypothesis test set the significance level at 95% for this one-tail test, with the risk levels set at their respective defaults—alpha at 5% and beta at 20% . Hence, if the hypothesis test results show that the p-value < 0.05, then reject H0 and take action to ship the new product (Product B) because it works faster than Product A. In Figure H-9, the upper left quadrant illustrates when the hypothesis test concludes with 95% confidence that the Null is true; and the p-value is greater than or equal to 0.05. In the lower right quadrant, the hypothesis test provides sufficient evidence to reject the Null, the p-value is less than 0.05, and the test conclusion is correct. Power is set at 80% (that is, one minus beta of 20%) to detect a distinguishable difference between the two products. In both the upper right and lower left quadrants of Figure H-9, the risk of making a wrong decision is illustrated. The upper right quadrant shows the 5% probability (that is, 1 minus the 0.95 significance level) of making a Type I Error of incorrectly rejecting the Null hypothesis, when the test results-produced p-value is less than 0.05; however, in truth the products worked the same. The lower left quadrant shows the 20% probability of making a Type II Error of incorrectly accepting the Null

P Q R S T U V W X Y Z

346

Encyclopedia

hypothesis when the test results produced p-value is greater than 0.05; however, in truth Product B worked faster than A.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Different Hypothesis Tests Hypothesis testing is actually the “category” of statistical tools, and depending on the type of data and the sample size available, determines which statistical test gets applied. Table H-1 summarizes the different types of Hypothesis Tests that presume normally distributed data. If the data are not normal, transformation is required to use these statistical tests. (See Also “Control Charts,” section on “Transformation of Nonnormal Data (plus Normal Probability Plots,” p. 444) Table H-1 provides not only the different tests by data type and sample size, but also it summarizes each test common application, its applicable formulas to calculate the test statistic, and the degrees of freedom if applicable. The term degrees of freedom (df) is a statistical expression that represents the amount of freedom (or “float”) the data has to “represent” the population. It is the number of measurements that are independently available to estimate a population parameter. As data is collected, degrees of freedom (df) are earned. As statistical information is calculated (or a parameter is estimated), degrees of freedom are “spent” on describing the population from the sample data. The mean is a calculated statistic, and it uses up one degree of freedom, resulting in (n-1) degrees of freedom, where “n” represents the sample size. If a population average is represented by four numbers, there is freedom for the first three numbers to be whatever they want to be, but the fourth number must be “dedicated” (as a calculation) to achieve the same population average. Hence, that fourth number does not have the freedom to be whatever it wants to be, and so a four-number sample (n=4) has three degrees of freedom to describe the average (for example, df = n-1; 4-1 = 3). Manual calculations require the appropriate hypothesis test formula to compute and compare the specific sample data set’s test statistic with the test’s critical value. Manual calculations also require referencing of the statistical test’s distribution table to derive the appropriate critical value for the specific scenario. These distribution tables are provided as a reference toward the back of the book. (See the “Statistical Distribution Tables,” in Appendix A, p. 939.) However, statistical software packages, such as MINITAB, have automated the calculations and negate the need for the formulas and distribution tables. Table H-1 summarizes the commonly used Hypothesis tests, including the type of data required for the test, the recommended sample size, its appropriate application, its test statistic formula, and degrees of freedom formula.

Table H-1: Reference Summary of Common Hypothesis Tests Data Type and Sample Size

Hypothesis Test

t-Test (or Student t)

Test Statistic Formula

Application

Continuous or Variable. Sample size is less than 30 (n 1-Sample t…. • Select the sample data from its worksheet; set the test mean in the 1-

Sample t main screen as shown in “Area 1” of Figure H-10. • Set risk levels in the “Options” screen for the Confidence level, leave

the Confidence at the default of 0.95, and select the “greater than” test for the Alternative hypothesis statement, as shown in “Area 2” of Figure H-10. Select OK on both the Options and main screens to run calculations.

J K L M N O P Q R S T U V

Figure H-10: MINITAB One Sample t-Test Example

W X

Conclusions Using MINITAB Sample Output

Y Z

• The output displayed in the MINITAB session window provides the

p-value. Figure H-11 annotates MINITAB’s output to highlight the conclusions. • The p-value of 0.269 is greater than the set alpha of 0.05; hence, we

cannot reject the Null hypothesis. We conclude that there is not enough evidence to say that the new hybrid gets better mileage than the manufacturer’s claim.

Hypothe sis Te sting

355

A B C D E Figure H-11: MINITAB One Sample t-Test Session Window Example

F G

• The sample’s 95% Lower Bound of 51.8333 is less than population’s

52.5 MPG mean, so the sample mean of 52.884 is likely different from 52.5 because of random chance alone.

H I J

Manual Calculations

K

• To determine the t-test critical value, look up alpha in a t-Distribution

statistics table.

L M N O

Note

P

The example is a one-tail test, with alpha at 0.05, and the degrees of

Q

freedom (n–1) is 24. The critical value found in the t-Distribution Table

R

is 1.711. (See Also “t Distribution Critical Values,” p. 939 in Appendix

S

A, “Statistical Distribution Tables” of the Appendix.)

T U

• Using the Student t formula and the following statistics

from sample data to calculate the test statistic: sample mean (x-bar) = 52.884; population mean (52.5); standard deviation = 3.0701; n = 25.

t=

−µ s n

• The resultant test statistic equals 0.625, which rounds to 0.63 and

matches MINITAB’s output. Conclusions Using Manual Calculations • The test statistic of 0.63 does not fall into the rejection region; it is

less than the critical value of 1.711. Fail to reject the Null hypothesis.

V W X Y Z

356

Encyclopedia • We conclude that with 95% confidence, based on our experience

with 25 tanks of gas, that the mileage for our new hybrid car is no different from the manufacturer’s claim of 52.5 MPG when driving 50% in the city and 50% on the highway. A B

Figure H-12 compares a p-value and test statistic drawn on a frequency tdistribution for this example to show how the same conclusion is reached regardless of which statistic is used.

C D E F

Using P-VALUE to draw conclusions

G

Rejection Region Fail to Reject H 0

H I J K L M N O P Q R S T U V W X Y Z

X

t crit ical

p-value @ 0.269

Using Test statistic to draw conclusions

(0.05 for alpha)

Rejection Region Fail to Reject H 0

X

t crit ical (1.711 from t-Distribution Table)

Test statistic @ 0.63

Figure H-12: Compare P-value and Test Statistics to Draw Conclusions

Paired t-Test Example Scenario: An advertising firm wants to compare two different collaterals to see if its new version (called Material 2) more effectively communicates the message to its target audience. It decides to survey five customers to compare effectiveness of the new versus the old version. The data collected are the number of negative comments about the two pieces. Material 1 refers to the old content. Consider sources of variability. In addition to the two different training materials, this scenario has other sources of variability, which are not of interest. The different learning styles and backgrounds represent a potential source of variability among the students. Effective planning isolates these variability sources of interest (testing the training materials on two different groups) for further study. The concept of blocking in a Hypothesis test removes unwanted sources of variation (called noise) from the study. Blocking improves the signal-to-noise ratio, thereby improving the ability to detect a difference if it exists. Conducting a Paired t-Test accounts for blocking,. The test compares two sample groups, wherein the sample size of each group is less than 30.

Hypothe sis Te sting

357

Note As the sample size approaches 30, the t-distribution approaches a normal distribution. A B

Other examples of paired data include

C • Comparing scores on tests both pre-and post-training sessions

D

• Comparing the operators’ efficiency ratings both before and after

E

training

F

• Comparing machine throughput both before and after a software

upgrade

G H

• Comparing the durability of different (product) materials under

varying conditions

I J K

Hypothesis Statements and Conditions

L • H0: Material 2 is the same as Material 1; no distinguishable differ-

ence (µ2 = µ1).

M N

• Ha: Material 2 is different from Material 1 (µ2 ≠ µ1).

O

• Theory to question is a two-tailed test; wherein “nominal is best.”

P

• Set confidence level at 95%; alpha at default 5%; beta at default 20%.

Q

• Sample size (n) is 25. The following data was collected:

R S T

Subject

Material 1

Material 2

Skip

6

7

Kathy

12

7

W

Eduardo

15

10

X

Neeta

8

4

Kevin

7

4

MINITAB Commands • After the data is entered into a MINITAB Worksheet, select the fol-

lowing sequence from its drop-down menu: Stat > Basic Statistics > Paired t….

U V

Y Z

358

Encyclopedia • Select the sample data from its worksheet in the Paired t main screen

as shown in “Area 1” of Figure H-13. • Set risk levels in the Options screen; leave the Confidence level at

A B C

the default of 0.95, leave the Test mean at the default of 0.0, and select not equal for the Alternative hypothesis statement, as shown in “Area 2” of Figure H-13. • Select “OK” on both the “Options” and main screens to run

calculations.

D E F G H I J K L M N O P Q R

Figure H-13: MINITAB Paired t-Test Example

Conclusions Using MINITAB Sample Output

S

• The output displayed in the MINITAB session window provides the

T

p-value. Figure H-14 annotates MINITAB’s output to highlight the conclusions.

U V W X Y Z

Figure H-14: MINITAB Paired t-Test Session Window Example

Hypothe sis Te sting

359

• The p-value of 0.045 is less than the set alpha of 0.05; hence, we reject

the Null hypothesis. We conclude that there is sufficient evidence to say that the new advertising collateral is different from the original. Manual Calculations • To determine the t-test critical value, look up alpha in a t-Distribu-

tion statistics table. (See Also “t Distribution Critical Values,” p. 939 in Appendix A, “Statistical Distribution Tables.”)

A B C D

Note

E

This example is a two-tailed test with alpha at 0.05 split between the

F

two tails (0.05/2 = 0.025), and the degrees of freedom (n-1) is 4. The

G

critical value found in the t-Distribution Table is 2.776. (See Also

H

“t Distribution Critical Values,” p. 939 in Appendix A, “Statistical Dis-

I

tribution Tables.”)

J K

• Calculate the difference between the pair of data per subject and cal-

culate the average difference for the set of data.

L M N

Subject

Material 1

Material 2

Difference (d)

Sd = (di—d-bar)2/n-1

O

Skip

6

7

-1

17.64/4 = 4.41

P

Kathy

12

7

5

3.24/4= 0.81

Eduardo

15

10

5

3.24/4= 0.81

Neeta

8

4

4

0.64/4 = 0.16

Kevin

7

4

3

0.04/4 = 0.01

V

Sum:

16

6.2

W

Calculation:

(16/5=3.2 d-bar) (Sq. root of 6.2 = 2.489979 Sd )

R S T U

• Using the Paired t-test formula and the following statis-

tics from sample data to calculate the test statistic: sample difference mean (d-bar) = 3.2; standard deviation of difference = 2.48998; Sd n = 5.

Q

t

X Y Z d Sd / n

(d1 d ) 2 n 1

360

Encyclopedia • The resulting test statistic equals 2.87368, which rounds to 2.87, and

matches MINITAB’s output. Conclusions Using Manual Calculations A B C D E F G

• The test statistic of 2.87368 falls into the rejection region; it is greater

than the critical value of 2.776. Therefore, reject the Null hypothesis. • We conclude that with 95% confidence, based on our experience

with five customers, that the new advertising collateral is different from the old version. Figure H-15 compares a p-value and test statistic drawn on a frequency t-distribution for this example to show how the same conclusion is reached regardless of which statistic is used.

H I J K

Using P-VALUE to draw conclusions

L

Rejection Region Fail to Reject H 0

X

M

t crit ical (0.05 for alpha)

N O P Q R S T U V W X Y Z

p-value @ 0.045

Using Test statistic to draw conclusions

Rejection Region Fail to Reject H 0

X

t crit ical

Test statistic @ 2.87

(2.776 from t-Distribution Table)

Figure H-15: Compare P-value and Test Statistics to Draw Conclusions for Paired t-Test Example

Chi-Square (χ2) Test Example (pronounced “ky” square) Scenario: A marketing department conducts a survey of premier customers to understand their preference for a new service offering it just announced. The survey was distributed by email and paper. A total of 200 customers replied. Some of the respondents chose to respond by email, and some replied by paper using the “snail-mail” approach. Marketing questioned whether the customers’ response differed depending on

Hypothe sis Te sting

361

which of these two methods (paper or email) they used (Test for Independence). Attribute data: If response data is attribute (counts of occurrences or defects) and we wish to make comparisons of these proportions across multiple groups or strata, then we use a Chi-Square test for independence. Examples of typical Chi-Square scenarios include • Product (or services) preference compared with gender type, age, or

income category. • Defect types compared with categories of time (such as Week 1, 2,

and so on) or compared with suppliers. • Cycle time compared with machine type, shift, type of process, or

job type.

A B C D E F G H

Hypothesis Statements and Conditions

I

• H0: The number of positive responses (“yes”) (or the number of

non-responses “no”) is independent of paper and email communication approaches (p2 = p1).

J K L M

Note

N

In the preceding formula, “p” represents “proportions.”

O P Q

• Ha: The number of positive responses (or the number of non-

responses) differs by paper and email; they are not independent (p2 ≠ p1).

• Theory to question is a test for independence, hence use a right-tail

Chi-Square test. • Set confidence level at 95%; alpha at default 5%; beta at default 20%. • Sample size (n) is 200. The survey results are Response

Paper

“Yes”

R S T U V W X

Email

Total

45

62

107

“No”

55

38

93

Total

100

100

200

Y Z

362

Encyclopedia

Note The collected data is arranged in a table, often called a contingency table, cross-tab matrix, or row and column analysis. A B C D E F G H

MINITAB Commands • After the data is entered into a MINITAB Worksheet, select the fol-

lowing sequence from its drop-down menu: Stat > Tables > ChiSquare Test (Table in Worksheet)…. • Select the sample data from its worksheet in the Chi-Square Test

main screen and select OK to run the calculations, as shown in Figure H-16.

I J K L M N O P Q R S T U V W X Y Z

Figure H-16: MINITAB Chi-Square Test Example

Conclusions Using MINITAB Sample Output • The output displayed in the MINITAB session window provides the

p-value. Figure H-17 annotates MINITAB’s output to highlight the conclusions.

Hypothe sis Te sting

363

A B C D E F G

Figure H-17: MINITAB Chi-Square Test Session Window Example

H • The p-value of 0.016 is less than the set alpha of 0.05; hence, we

I J

reject the Null hypothesis. • We conclude that with 95% confidence, there is sufficient evidence to

K

say that customer responses in the email communication were statistically different from those who selected the paper “snail” mail approach. The response and communication method selected are not independent; there is a relationship. The customers who responded by email were more likely to respond “yes” to preferring the new service than the customers who chose to reply by paper.

L

• The shape of the Chi-Square distribution depends on the degrees of

• The Chi-Square test for

independence generally is a right-tail test (α), to

N O P Q

Manual Calculations

freedom. As the sample size gets larger and the degrees of freedom are 15 or higher, the shape approaches normal distribution. Otherwise, the Chi-Square distribution is a right (or positive) skewed curve with asymmetrical tails, as illustrated in Figure H-18.

M

Chi-Square ( 2) Distribution

 /2 2

For a two-tailed Goodness of Fit

Rejection Region

R S T U V W X

Rejection Region

Fail to Reject H0

 21-  For a left tail Goodness of Fit

Y Z

 2 For a Right tail Goodness of Fit & Test of independence

Figure H-18: Illustration of a Chi-Square Distribution

364

A

Encyclopedia

look for a large difference in the formula’s numerator (for example, the squared difference between the observed minus the expected frequencies). Although the Null hypothesis wording states “no difference,” that the observed versus the expected proportions are independent, and the alternative states “a difference” that the comparison indicates a relationship or dependency, only a right-tail test is used.

B C D E F G

Note The appropriate alpha-level for a Goodness of Fit Chi-Square test, comparing the variances between a sample and that of the known population, depends on the Null and alternative hypothesis statements.

H I

• To determine the Chi-Square critical value, look up alpha with the corre-

J

sponding degrees of freedom in a Critical Values of the Chi-Square-Distribution statistics table. The degrees of freedom (df) formula is “number of rows minus one times the number of columns minus one,” or (r-1)(c-1).

K L M N O P Q R S T U V W X

• For this example, use a right-tail test, with alpha at 0.05. • The df calculation is (2 rows–1)(2 columns–1) = (1 x 1) = 1, which matches MINITAB’s Session window output. • The critical value found in the Chi-Square Distribution Table for alpha of 0.05 and df of 1 is 3.84.] (See Also “Chi-Square (χ2) Distribution Critical Values,” p. 939 in Appendix A, “Statistical Distribution Tables.”) • Calculate the expected value (or frequency) for each cell of the set of data, using the following formulas: • row probability = row total/grand total • column probability = column total/grand total

Y

• cell probability = row probability x column probability

Z

• expected cell frequency = cell probability x grand total The shaded cells in the contingency table represent those containing the actual observed data. • Double-check the math by ensuring that the summed rows equal the grand total and the summed columns equal the grand total.

Response

“Yes” responses

Paper Actual

Paper Expected (probability)

Email Actual

Email Expected (probability)

45

(0.535 × 0.5 = 0.2675) × 200 = 53.5

62

(0.535 × 0.5 = 0.2675) × 200 = 53.5

χ2 for Yes

(45-53.5)2/53.5 = 1.35

χ2 for No

(0.465 × 0.5 = 0.2325) × 200 = 46.5

38

(55-46.5)2/46.5 = 1.55376

Total responses

Total χ2

2.7 (93/200 = 0.465) × 200 = 93.0

93

(38-46.5)2/46.5 = 1.55376

(100/200 = 0.50) x 200 = 100

100

(107/200= 0.535) × 200 = 107

107

(62-53.5)2/53.5 = 1.35

(0.465 × 0.5 = 0.2325) × 200 = 46.5

55

“No” responses

Total Expected (probability)

Total

3.108

(100/200 = 0.50) x 200 = 100

100

2.904

(200/200 = 1.00) x 200 = 200

200

2.904

5.808 Hypothe sis Te sting 365

A

B

C

D

E

F

G

H

I

J

K

L

M

N

O

P

Q

R

S

T

U

V

W

X

Y

Z

366

Encyclopedia `• Using the Chi-Square (χ2) Test for Independence formula, calculate

the test statistic. 2

E)2

(O E

A B C D E F G

where O = Observed (or actual) frequency, and E = Expected frequency • The resultant Chi-Square (χ2) test statistic equals 5.808, which matches MINITAB’s session window output. Conclusions Using Manual Calculations • The Chi-Square test statistic of 5.808 is greater than the critical value

H

of 3.84; therefore, it falls into the rejection region. Reject the Null hypothesis.

I

• We conclude that with 95% confidence, based on our experience with

J K L

200 respondents, that the number of positive responses (or the number of non-responses) differs by paper and email communication method. The response and communication method are not independent.

M N

Hints and Tips

O

All hypothesis tests are constructed in a similar manner. The Null

P

hypothesis states that there is no difference between the parameters

Q

of interest. The Alternative hypothesis states that there is a difference

R

between the parameters of interest.

S

Rule of thumb: The Null hypothesis (H0) usually is worded as “equal

T

to” (including less than and equal to or more than and equal to). Hence,

U

the alternative hypothesis (Ha) generally contains “not equal to.”

V

Rule of thumb: The underlining hypothesis, what is hoped to be true,

W

generally worded as the alternative hypothesis (Ha). So generally, start

X

by stating the alternative (Ha) first and then state the Null hypothesis

Y

(H0) afterward.

Z

The Difference between Directional versus Non-Directional Tests Non-directional or two-tailed test—If you want to determine if two things are different, then the Alternative hypothesis is considered nondirectional. The alternative hypothesis will use non-equivalence language (mathematically a “not equal” sign is used (≠ or ).

Hypothe sis Te sting

367

Directional or one-tail test—If you are interested in stating that something is smaller than or larger than a quantity, then the Alternative hypothesis is called directional. The alternative hypothesis will consist of a comparison that is either less than or greater than a quantity of interest (mathematically this equates to using a < or > symbol).

A B C D

Hints and Tips

E

Decision Criteria—If this p-value is smaller than the pre-defined level

F

of significance (typically α = 0.05), then the Null is rejected.

G

The p-value is the actual Type I Error based on the sample data.

H

A signal-to-noise ratio is calculated with a probability (p-value)

I

assigned to it, designating the likelihood that this size of ratio could

J

occur by random chance.

K L

Supporting or Linked Tools Supporting tools that might provide input when developing an hypothesis test include • Data Gathering techniques (See Also “Data Collection Matrix,”

p. 248) • Performance charts and dashboards

M N O P Q R S

• Graphical tools (See Also “Graphical Methods,” p. 323)

T

• Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram—

U

7QC Tool,” p. 173) • Process map (See Also “Process Map (or Flowchart)—7QC Tool,”

p. 522)

V W X

• Hypothesis

Y

• QFD (See Also “Quality Function Deployment (QFD),” p. 543)

Z

• Solution selection matrix (See Also “Solution Selection Matrix,”

p. 672) • Brainstorming (See Also “Brainstorming Technique,” p. 168)

A completed hypothesis test provides input to tools such as • QFD (See Also “Quality Function Deployment (QFD),” p. 543)

368

Encyclopedia • Solution selection matrix (See Also “Solution Selection Matrix,” p. 672) • Brainstorming (See Also “Brainstorming Technique,” p. 168) • Control Plan (See Also “Matrix Diagrams—7M Tool,” p. 399, for a

brief discussion on control plan.) A B C D

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),”

p. 287) Figure H-19 illustrates the link between an hypothesis test and its related tools and techniques.

E F G H I

Data Gathering (metrics) Performance Charts and Dashboards

QFD

Solution Selection Matrix

Brainstorm

J K

Graphical Tools

L M N O P

Control Plan Hypothesis Testing

Cause-Effect Diagram

Process Map

Q R

Hypothesis

S T U V W X Y Z

Figure H-19: Hypothesis Test Tool Linkage

FMEA

Interrelationship Diagram—7M Tool

369

I Interrelationship Diagram—7M Tool A

What Question(s) Does the Tool or Technique Answer? How do the various cause-and-effect ideas relate to one another in a complex situation? An Interrelationship diagram helps you to • Understand the relationship between topics of a complex situation • Decipher the relationship of intertwined topics to identify the area of

greatest impact for improvement

B C D E F G H I J

Alternative Names and Variations This tool is also known as • Relationship diagram or digraph • Network diagram

K L M N O

Variations on the tool include

P

• Matrix relations diagram

Q R S

When Best to Use the Tool or Technique The tool is best used on more complex issues wherein the exact causal relationship or inner-dependencies are difficult to discern.

T U V W

Brief Description The type of diagram maps the various links, using directional arrows to indicate source of a cause (at the base of the arrow) and impact (at the point of the arrow). Developing an Interrelationship diagram involves a team, preferably a cross-functional one, using creative problem-solving to construct the diagram. This tool builds on work done in other tools such as Affinity diagrams, Tree diagrams and Cause-and-Effect diagrams. Upon completion, the tool maps the connections of the various causeand-effect relationships to highlight the hub of greatest activity. The team then analyzes the network of relationships to identify the key causes.

X Y Z

370

A B C D E F G

Encyclopedia

This tool is a member of the 7M Tools, attributed in part to Dr. Shewhart, as seven “management” tools, or sometimes referred to as the 7MP, for seven management and planning tools. These 7M tools make up the set of traditional quality tools used to analyze quantitative data. The 7M toolset includes: 1) Activity network diagrams or Arrow diagrams, 2) Affinity diagrams, 3) Interrelationship digraphs or Relations diagrams, 4) Matrix diagrams, 5) Prioritization matrices, often replacing the more complex Matrix data analysis, 6) Process decision program charts (PDPC), and 7) Tree diagrams. The Quality Toolbox, by Nancy Tague, presents the 7M tools ranked from those used for abstract analysis to detailed planning: Affinity diagram, Relations diagram, Tree diagram, Matrix diagram, Matrix data analysis (commonly replaced by a more simple Prioritization matrix), Arrow diagram, and Process Decision Program Chart (PDPC). (See Also “References,” in Appendix C, p. 981)

H

L

How to Use the Tool or Technique The Interrelationship diagram is built much like an Affinity diagram. The best development technique involves materials such as sticky notes or index cards to capture one idea per sticky note or card. The procedure involves the following steps:

M

Step 1.

I J K

N

Identify the complex topic of interest and gather supporting or input documents including those with text or verbal data. Examples of different inputs include: written reports, Affinity diagrams, Tree diagrams, and Fishbone (or Cause-and-Effect) diagrams.

O P Q R S T U V W X

Prepare to build the diagram.

Step 2.

Construct topic cards. Document one idea or topic on either a single piece of paper (or index card) or sticky note. Accomplish this by referencing supporting documents and brainstorming additional ideas. If a Fishbone or Tree diagram is referenced, select the most granular or detailed level of information. Place only one idea per piece of paper (index card or sticky note).

Y Z

Note As a short cut, if the data already exists in a text document, simply cut the document into strips, such that one strip of paper contains only one written idea.

Interrelationship Diagram—7M Tool

Step 3.

371

Group related ideas. Working with one documented idea at a time, place the idea card or sticky on the working surface (a flip chart or table). Try to place the topic near a related idea. If no related themes exist, place it on the working surface with enough empty space around it to allow room for related cards to be added and for arrows to be drawn. Repeat this step until all cards (or stickies) are placed.

A B C D

Hint

E

S huffling the cards (or stickies) and randomly addressing each idea

F

sometimes triggers creative thinking about related topics.

G H

Step 4.

Draw the relationship arrows. Review and discuss how an idea is related to another. Draw the relationship arrow by starting with the causal topic as the base of the arrow and connecting it to the effected topic (cause→ effect). Repeat this step until each card has been reviewed and all the relationship arrows drawn.

J K L M N

a. There are no restrictions on number of arrows coming into or out of any one card; however, there should be no bidirectional (or two-ended) arrows.

O

b. Take the time to reflect and review the diagram. Often the revisions are as insightful as the original draft.

R

c. Option: Once the diagram is finalized, record the number of incoming and outgoing arrows on a lower corner of each card (or sticky) to assist with the prioritization step (for example, “1/5” for one incoming and 5 outgoing arrows). d. A large number of arrows stemming from a topic indicates a root cause or driver. Several arrows pointing to an item indicates a key outcome. Step 5.

I

Prioritize and assign accountabilities. Identify those topics with the most outgoing arrows and with the most incoming arrows. Draw conclusions about those item(s) requiring the most attention. Select the top priority

P Q S T U V W X Y Z

372

Encyclopedia

items (usually the top 2 or 3), encircle the cards (or stickies), assign each top priority to one person accountable for resolution, and write the person’s name on the card.

A B C D

Hint Highest priority may involve severity of impact, rather than highest number of inbound or outbound arrows. Thus, double-check the prioritization list and gain agreement on those requiring immediate action.

E F G H I J

Figure I-1 illustrates a completed Interrelationship diagram to look at the Cause-and-Effect relationship of trying to improve new business sales. Notice that each item contains the number of inbound and outbound arrows documented in the lower portion of the box. Those items with the highest number of arrows (either outgoing causes or incoming effects) also have bolded outline and assigned person accountable to take action.

K L

Length of time in territory (1/1)

Tenure selling (1/1)

M

Sales rep ability to interaction well with customer (5/2) B Skea

Sales selling skills (1/3)

Product training (1/3)

N O

Number of Meetings with Decision makers (5/1) C. Kiley

Number of Prospects (2/1)

Sales territory mix (2/2)

Number of Proposals written (2/2)

Number of Proposals written (2/1)

P Q R S

Sales product mix to sell (1/2)

Sales has access to customer decision maker (1/1)

Sales annual targets (units / revenue) (1/2)

U

Improve New Business Sales

V

(Arrows lr / Out, Person Accountable)

Z

Sales ability to negotiate directly with customer (1/3)

Productivity (Units sold / year / sales rep) (3/2)

Product fits customer need (3/1)

Sales understands product and its benefits (2/2)

Customer perception (5/1) L. Niland

Productivity (Revenue / year / sales rep) (1/1)

Product feature / functionality (2/2)

Product promotion (3/1) Product price (6/1) D. Ide

W Y

Sales understands customer need (1/1)

Number of sales reps in territory (1/2)

T

X

Sales relationship with customer (1/1)

Customers purchased product (2/1)

Sales manager (0/2)

Product complexity (1/1) Advertising (2/2)

Value Proposition (6/1) G Ruckus Market share (3/1)

Product Relations (1/1)

Company reputation (1/2)

Competitor (0/7) N/A

Figure I-1: Sample Interrelationship Diagram

Product quality (1/1)

Product support (2/2)

Interrelationship Diagram—7M Tool

373

Notice that Competition is identified as one of the themes with the highest number of outgoing arrows in Figure I-1. However, since the item is not actionable, no person is assigned to it.

Supporting or Linked Tools Supporting tools that might provide input when developing an Interrelationship diagram include • Written report • Affinity Diagram—7M Tool, p. 136

A B C D E

• Tree Diagram—7M Tool, p. 712

F

• Fishbone or Cause-and-Effect diagram (See Also “Cause-and-Effect

G

Diagram,” p. 173) • Brainstorming (See Also “Brainstorming Technique,” p. 168)

A completed Interrelationship diagram provides input to tools such as • Brainstorming (See Also “Brainstorming Technique,” p. 168)

H I J K L

• QFD (See Also “Quality Function Deployment (QFD),” p. 543)

M

• Root cause analysis techniques (such as statistical Hypothesis testing

N

and Regression) (See Also “Hypothesis Testing,” p. 335, and “Regression Analysis,” p. 571)

O P

• Concept generation techniques

Q

• FMEA (See Also “FMEA,” p. 287)

R

Figure I-2 illustrates the link between an Interrelationship diagram and its related tools and techniques, showing how some provide input to, while others utilize the information from a completed Interrelationship diagram.

S T U V W X Y Z

374

Encyclopedia Written Report

Brainstorming

Affinity diagrams Interrelationship Diagram

A B

QFD

Root Cause Analysis Techniques

Tree Diagram

Concept Generation Methods

Fishbone

FMEA

C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure I-2: Interrelationship Diagram Tool Linkage

KJ Analysis

375

K KJ Analysis A

What Question(s) Does the Tool or Technique Answer? What matters to your customers? What are the natural groupings, categories, or affinities of a large number of topics, ideas, and quotations, and how can you best translate the verbal input into requirements? A KJ Analysis helps you to

B C D E F

• Discover and clarify customer requirements

G

• Organize and categorize large volumes of language-based data (that

H

is, text or verbal input) • Show a “parent-child” relationship between more detailed ideas to

their higher-order theme, also called Affinity groupings • Translate the input into customer requirements

I J K L M

Alternative Names and Variations This tool is also known as • Affinity diagram or chart

Variations on the tool include • House of Quality (HOQ) or Quality Function Deployment (QFD)

N O P Q R S

• Fishbone diagram (rarely)

T

• Tree diagram or matrix (rarely)

U V

When Best to Use the Tool or Technique Successful improvement efforts start with a focus on customer needs. The KJ Analysis starts with Voice of the Customer (VOC) input, which is gathered for the purpose of either developing an offering concept for the strategic Portfolio Renewal process or at the early stages of the tactical Offering Development process. (See Also “VOC Gathering Techniques,” p. 737, “Six Sigma for Marketing (SSFM),” in Part I , p. 67, and “Listening to the Customer First-hand; Engineers Too,” in Part III, p. 851.) A KJ Analysis organizes a large number of ideas, quotations, and facts into themes. This categorization allows the content to be translated into customer requirements. This approach often times translates the implicit

W X Y Z

376

Encyclopedia

into explicit needs. It helps reduce complexity and connect related topics. This tool analyzes customer interview data or survey results and summarizes it for purposes of communication and devising responding actions, such as specifications, root cause analysis, and solution generation techniques. A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Brief Description The tool was developed in the early 1960s, by a Japanese anthropologist, Jiro Kawakita, which is where its name comes from, “KJ.” [Recall that a Japanese name reverses the order of the given and sir name relative to an English name.] Dr. Kawakita was the founder of the Kawayoshida Research Center. Kano Model: NUD Versus ECO The KJ Analysis translates customer input into NUD requirements [pronounced “nuˇ d”]—what is New, Unique, and Difficult. NUD requirements add value to the customer and differentiate an offering from its competition. Within this context, the NUD customer requirements are defined as • New—Something the customer has not asked for or thought of

before; it is completely new. • Unique—Something different that is not currently being provided in

the marketplace today (either by your organization and/or by a competitor). This could be a substitution product or service. • Difficult—Something that is very challenging to offer (not necessar-

ily new or unique but often is). NUD requirements delight customers. If you do not satisfy your customers’ needs, your competition will. NUD requirements put pressure on competition because these attributes require more lead-time, know-how, and investment to replicate or mirror. Per the Kano model, customer delighters satisfy customer needs better than those offerings that provide expected or basic quality, features, and functionality. The Kano model defines three types of product or service quality that provide customer satisfaction: 1. Fitness to Latent Expectations and delighting quality—Where a

neutral effect if not present, but extremely powerful if present. 2. Fitness to Use and performance quality—Where more is better,

which is often the focus of VOC efforts. 3. Fitness to Standards and basic quality—These items must be pres-

ent, yet have a neutral effect from the customer. They may be assumed and often are not clearly documented.

KJ Analysis

377

The attributes of new, unique, and difficult describe how best to produce a delighter that excites and stimulates customer buying behavior. Figure K-1 illustrates how the Kano model of customer satisfaction relates the three levels of quality—delighting, performance, and basic. A

Customer Satisfaction

B C D E

Fitness to Latent Expectations = Delighter

F Corporate Execution

Fitness to Use = Linear Satisfier

Fitness to Standard = Must Have

G H I J K

Figure K-1: Kano Model of Customer Satisfaction

The idea is to avoid working on ECO customer requirements [pronounced “eˇ?k-y-”]—those that are Easy, Common and Old requirements. These attributes describe the basic quality curve on the Kano model—the mere minimum expected by the customer. The ECO requirements provide little to no value to the customer but must to be present in a product or service. Plus, competitors can easily replicate or respond inkind. The ECO requirements are defined as

L M N O P Q R S

• Easy—Simple to provide with low effort, low cost, and low risk.

T

• Common—It has been offered before with a repeatable history of

U

success; then it is not unique. • Old—Something we have experience fulfilling, and it does not rep-

resent a critical requirement. A good source of NUD requirements stems from the context within which a customer currently works. This data often comes from images in the form of descriptive sentences or observations. An important aspect and distinguishing feature of a KJ Analysis is the processing of customer input as “images.” Images or pictures that describe the context provide the best source of data. This image data provides insight into the

V W X Y Z

378

Encyclopedia

customer’s “latent” needs (intangible or implicit). From a customer environment perspective, there are three kinds of images: 1. A picture of some event, situation, process, or condition that the customer describes during an interview discussion. A B C D E F G H I J K L

2. An actual event, situation, process, or condition that the interviewer witnessed in the customer’s environment. 3. A picture of some event, situation, process, or condition that forms in the interviewer’s mind during an interview discussion. The KJ produces three main document outputs: Image KJ Document, Translation Document, and a Requirements KJ Document. The Requirements KJ document is similar to a CTQ (Critical-to-Quality) in that it translates the customer requirements into the team’s understanding and places them in a House of Quality tool. The KJ Analysis is similar to but technically more complex than an Affinity diagram. It comprehends the implicit needs revealed through story-telling and observations. Both were developed by Jiro Kawakita. Hence, this article discusses the more in-depth KJ Analysis technique. (See Also “Affinity Diagram,” p. 136)

M N O P Q R S

How to Use the Tool or Technique This technique is performed usually by a small group made up of members of a product development team assigned to work on analyzing the VOC data. The team produces three main document outputs when conducting a KJ Analysis: 1. Image KJ Document

T

2. Translation Document

U

3. Requirements KJ Document

V W X

The KJ Analysis procedural steps to build these three documents are as follows:

Y

Image KJ Analysis Procedure This procedure describes the development of an Image KJ Document.

Z

Step 1.

Gather VOC. Identify the topic and then collect and/or capture the text or verbal data. Examples of different means of gathering such input are interviews or surveys. a. Identifying the “topic” may simply involve agreeing what aspect of the VOC Interview requires the team to dissect and analyze.

KJ Analysys

379

b. Ideally seek “imagery” input from the customers because the verbal data is richer. Imagery data paints a scenario, sets the context within which the customers are doing something, and uses something to describe their requirements. c. VOC Interview gathering techniques are best done in the customers’ own environment. Observations of the work setting assist the customers in recalling the current state and visualizing what they would like to keep the same; start doing differently or stop doing all together. d. Inquiry should focus on “in-context” analysis to gain additional insights on how and why things are done. This inquiry can effectively be achieved through both further verbal probing or simply observing the surrounding environment in total. Step 2.

Prepare the data by putting one idea, thought, or suggestion (either written or verbal input) on either a single piece of paper, index card, or sticky note. (Only one idea per piece of paper, index card or sticky note.) This one idea per card allows ideas to be moved around, mixed-and-matched, and grouped more easily, rather than placing several ideas on one piece of paper and deciding later that one of the ideas may be grouped better with another set of ideas.

A B C D E F G H I J K L M N O P

Note

Q

If the data already exists in a text document, a shortcut may be to

R

simply cut the document into strips, such that one strip of paper con-

S

tains only one written idea.

T U

Step 3.

Sort the input by content into themes or categories. Review each individual piece of paper and begin to place them in piles. Those ideas in the same pile relate to one another (or have an “affinity” for one another). If using sticky notes, follow the same procedure, but place the ideas with a common theme on a single flip chart sheet or section of the wall. Continue this step until all the individual ideas are placed in a group. a. Seek clarification on an idea if unclear by discussing among team members and reviewing additional customer input. Edit the idea, as needed.

V W X Y Z

380

Encyclopedia

b. Eliminate redundancy by combining an idea onto one piece of paper or sticky, such that within a group, one piece of paper represents one diverse idea. c. If an idea relates to more than one group, create a duplicate piece of paper and place it in all the appropriate groups.

A B

d. A few ideas may not fit well into the other groupings or themes; they may be independent ideas or loners. If so, place them in a “Miscellaneous” grouping.

C D E

Step 4.

F G H I

a. Remove all slips of paper that are absent of a red dot.

J

b. Count the red dot slips. If 20 to 30 remain, proceed to the next step. If more than 30 remain, then repeat this step by adding a second red dot to a slip, keeping only those with two red dots.

K L M N O P Q R

Identify important ideas using red dots. With either a red pen or red dot sticker, have the team indicate which ideas seem important. Only one red dot per statement; if someone has already put a red dot on a note, it has been identified as important. This helps identify the idea easily in later steps.

Step 5.

Organize the group and structure it all into a Ladder of Abstraction. Start by reviewing each idea within a grouping or affinity then again sort them into further subsets. Identify those that are similar and hierarchically group them. Figure K-2 illustrates a generic Ladder of Abstraction structure.

S T U V

Figure K-2: Ladder of Abstraction

W X Y Z

a. Double-check for redundancy and eliminate an unnecessary duplication. b. A sub-set classification example includes images and statements (verbatim comments). Use the image ideas; hold the statements for the Requirements KJ document described next.

KJ Analysys

381

i. Images—Describe the customer needs using

imagery by painting a scenario or story about tangible items. ii. Statements (verbatim)—Describe the customer

needs using less tangible language. c. Ladder of Abstraction. Arrange the input within each group into hierarchical clusters and arrange it into a rational Tree diagram, wherein the content relates to one another. The hierarchy is determined by sorting or ranking the data within each group by importance or the ability to meet the objective of the study. i. The first level of (or lowest level of detail in) the Ladder of Abstraction is called the black-level, named so for easy reference in future steps. d. Look across groups and determine if any groups are related to one another. Determine if there is a strong connection; if so, draw a dotted-line between the two main groups. Determine if any inputs are misplaced and rearrange. Step 6.

Name the themes by first identifying a category heading or name for the highest-level of commonality in a given group, and record it on another slip of paper (index card or sticky note). Start at the lowest-level of detail and continue this until all the groups have a label, except for the Miscellaneous group. a. Working from the lowest level of detail up, create a category name or sentence that states a high-level image that represents what is contained in each group. These statements should be only one level of abstraction higher. They are abstractions, not summary statements. i. Separating out the inherent qualities or properties. ii. Do not add new facts during this activity. iii. The name or theme title slips are referred to as the red-level in the Ladder of Abstraction. b. If the Ladder of Abstraction has multiple levels of subgroups within a big category, continue the naming process by labeling those higher-level groupings. i. This level constitutes combining lower-level (or redlevel) subgroups.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

382

Encyclopedia ii.The highest-level of title statements in a Ladder of

Abstraction is referred to as the blue-level. c. Reexamine if the Miscellaneous ideas relate to any of the named groupings; if so, move them to that category. These statements do not require a category name or statement. A B

Step 7.

C D E F

Diagram the relationships. Start with the Ladder of Abstraction structured in a hierarchical flow, similar to a Tree diagram, with the blue-level themes at the upper-most level and the corresponding red-level flowing from the higher-level, followed by the black-level. Using arrows, indicate which blue-level groups or themes support or conflict with another group by placing the appropriate symbol between the two blue groups.

G H I J

Note Convention calls for the relationship arrows to be drawn in blue.

K

a. Directional arrows (→) indicate a supportive relationship.

L

b. Crossed-lines (an “X”) indicate a conflicting relationship.

M

Figure K-3 illustrates at a high-level the flow of steps Two through Seven in conducting a KJ Analysis.

N O P Q R S T U V W X Y Z

Step 8.

Draw conclusions to complete the Image KJ document. Summarize the boundaries of the groups and draw conclusions about the structure. a. Draw boundaries around each blue grouping in its entirety. b. Write a conclusion statement about the relationships of the major (blue) themes, place them above the Relationship Arrows of the Image KJ document, and have the team sign their names on a bottom corner.

At this point, the Image KJ document has been completed and represents the scenes or images described by the customer or observed by the team.

KJ Analysis Preparing data

Sorting data into Themes

383

I.D. Important Ideas — Red Dots Misc Misc Misc

Organizing data into Ladder Abstraction

Naming the Themes Misc

Misc

Topic

Misc

A

Diagram Relationship Misc

Misc

Topic

Misc

Misc Misc

Misc

Blue Level Red Level Black Level

B C D E F

Black-level stickies: most basic, lowest level of detail. Black-level sticky with Red Dot: indicates important idea. Misc

Misc.: miscellaneous, independent or Lone Wolf ideas. Red-level stickies: theme title for a set of black-level ideas. Blue-level stickies: highest level category name for a set of red-level themes.

Figure K-3: KJ Analysis Flow of First Few Steps

G H I J K L

Requirements KJ Analysis Procedure (and Translation Document) This procedure describes the development of both the KJ Requirements document and a KJ Translation document.

M

Step 1.

P

Prepare to conduct the KJ Analysis and develop the KJ Translation document. Use the held statements (verbatim) that were sorted in Step 5 of the previous Image KJ procedure. a. Link each statement to at least one image. (A single statement may be liked to numerous images.) As explained earlier, KJ statements refer to those written items that describe the customer needs using less tangible language. b. Create several blank Customer Requirement Translation Worksheets, one per statement. The template should feature three sections: 1) Customer Statements (Voice), 2) KJ Images (from the KJ Image document), and 3) Translated Customer Requirements, as illustrated in Figure K-4.

N O Q R S T U V W X Y Z

Note If working in a team setting, use flip chart sheets for all to view.

384

Encyclopedia Project: Reqt. # Customer Statements (Voice):

A

KJ Images (from KJ Image document):

B C D

Translated Customer Requirements:

E F G H I

Figure K-4: Sample KJ Customer Requirement Translation Worksheet

J K L M N

Note Steps 2 through 5 mirror the same procedure as for the earlier Image KJ procedure described at the beginning of this section.

O P

Step 2.

Q R S

a. Double-check for redundancy and eliminate any unnecessary duplication.

T U

b. Sub-set classification example includes images and statements (verbatim comments). Use the image ideas; hold the “statements” for the Requirements KJ document described next.

V W X Y Z

Organize the group and structure into a Ladder of Abstraction. Start by reviewing each “held” statement within a grouping or affinity and then sort them again into further subsets. Identify those that are similar and hierarchically group them.

Step 3.

Ladder of Abstraction. Arrange the input within each group into hierarchical clusters and arrange it into a rational Tree diagram, wherein the content relate to one another. The hierarchy is determined by sorting or ranking the data within each group by importance or ability to meet the objective of the study.

KJ Analysys

385

a. The first level of (or lowest level of detail in) the Ladder of Abstraction is called the black-level. b. Look across groups and determine if any groups are related to one another. Determine if there is a strong connection; if so, draw a dotted line between the two main groups. Determine if any inputs are misplaced and rearrange. Step 4.

Name the themes by first identifying a category heading or name for the highest-level of commonality in a given group and record it on another slip of paper (index card or sticky note). Start at the lowest-level of detail and continue this until all the groups have a label, except for the Miscellaneous group. a. Working from the lowest level of detail up, create a category name or sentence that states a high-level image that represents what is contained in each group. These statements should be only one level of abstraction higher. They are abstractions, not summary statements. i. Separate out the inherent qualities or properties.

B C D E F G H I J K L

ii. Do not add new facts during this activity.

M

iii. The name or theme title slips are referred to as the

N

red-level in the Ladder of Abstraction. b. If the Ladder of Abstraction has multiple levels of subgroups within a big category, continue the naming process by labeling those higher-level groupings. i. This level constitutes a combining of lower-level (or

red-level) subgroups. ii. The highest-level of title statements in a Ladder of

Abstraction is referred to as the blue-level.

Step 5.

A

O P Q R S T U V

c. Reexamine if the Miscellaneous ideas relate to any of the named groupings; if so, move them to the appropriate category. These statements do not require category names or statements.

W

Diagram the relationships. Start with the Ladder of Abstraction structured in a hierarchical flow, similar to a Tree diagram, with the “blue-level” themes at the upper-most level and the corresponding red-level flowing from the higher-level, followed by the black-level. Using arrows, indicate which bluelevel groups or themes support or conflict with other groups by placing the appropriate symbol between the two blue groups.

Z

X Y

386

Encyclopedia

Note Convention calls for the relationship arrows to be drawn in blue.

A

a. Directional arrows (→) indicate a supportive relationship.

B

b. Crossed-lines (an “X”) indicate a conflicting relationship.

C D E F G H

Note After this step, the procedure involves unique steps from the earlier Image KJ procedure.

I J

Step 6.

K L M N O

a. Three Color Dot Stickers—Use red, blue and green sticky dots (or any other three colors) to vote for the most important red-level customer requirements. Provide one of each color sticky dot to each team member. Everyone votes using the dot-stickers, whereby a red dot equals three points, a blue dot equals two points, and a green dot equals one point. Sum up the total votes for each statement to determine the top three vote getters.

P Q R S T U

b. Weighed Voting—Each team member receives a total of six points to distribute among the statements any way he/she deems appropriate. Sum up the total votes for each statement to determine the top three vote getters.

V W X Y Z

Rank the top three red-level customer requirements. Using a voting system, identify the top three red-level requirements. This ranking system is the team’s opinion on rank and importance. The team should validate the results at a later time with the customer (via a survey, for example). Select one of the two approaches that follow to rank the red-level ideas:

Step 7.

Draw conclusions to complete the Requirements KJ document. Summarize the boundaries of the groups and draw conclusions about the structure. a. Draw boundaries around each blue grouping in its entirety. b. Write a conclusion statement about the relationships of the major (blue) themes and place them above the Relationship Arrows of the Image KJ document and have the team sign their names on a bottom corner.

KJ Analysys

387

Note At this point, the Requirements KJ document has been completed. The last step is to complete the Translation document. A

Step 8.

Fill in the Translation Worksheet for each of the top three redlevel customer requirement statements. Using one customer statement per form, write a clear and non-prescriptive customer requirement statement, characterized as a “solution-free statement.” Use factual language. Whenever possible, use continuous variables language (for example, use “75 degrees,” which is less ambiguous than “hot.”) This more precise language enables a crisper translation into actionable terms that an offering designer finds meaningful. a. Customer Statements describe the customer’s voice articulating their desires. Typically these statements start with “I want…” or “I need…” b. KJ Images are the references to the Image KJ document. They are the stories and observed behaviors that describe the context, the setting, and the scenario and lead to the latent expectations of the customers. These ideas are the innovation stimulators. (Reference the black-level ideas to enliven the story.)

Step 9.

Screen Customer Requirements. Create a matrix or a spreadsheet with three columns titled 1) Customer Requirements, 2) NUD Designation, and 3) Kano Designation. a. Customer Requirements column—List the black-level requirements. b. NUD Designation column—Indicate if the requirement is “N” for new; “U” for unique, “D” for difficult, or “SR” for a standard requirement (meaning ECO—easy, common, or old). c. Kano Designation column—Identify what Kano category the requirement falls under—”B” for basic; “LS” for linear satisfier; or “D” for delighter, described earlier in Figure K-1.

B C D E F G H I J K L M N O P Q R S T U V W X Y Z

388

Encyclopedia

d. Transfer this information into the Translated Customer Requirements section of the Translation document, thus completing the KJ Translation document.

A B C D E F

How to Analyze and Apply the Tool’s Output After the KJ Analysis is complete, the content from its three documents (Image, Requirements document, and the Translation document) can be transferred to a Quality Function Deployment (QFD) or House of Quality (HOQ). Rarely is this level of detail ported to other tools such as a Fishbone or Tree diagram.

G H I J K L M

Supporting or Linked Tools Supporting tools that might provide input when developing a KJ Analysis include • VOC Data Gathering Tools and Techniques, such as surveys and

interviews (See Also “Voice of Customer Gathering Techniques,” p. 737)

N

• Written reports

O

• Brainstorming (rarely) (See Also “Brainstorming Technique,”

P Q R S T U

p. 168) A completed KJ Analysis provides input to tools such as • House of Quality (HOQ) and/or Quality Function Deployment (QFD) (See Also “Quality Function Deployment (QFD),” p. 543) • Tree diagram and matrix (rarely) (See Also “Tree Diagram,” p. 712)

V

• Fishbone (rarely) (See Also “Cause-and-Effect Diagram,” p. 173)

W

• Simple matrix (rarely) (See Also “Matrix Diagrams,” p. 399)

X Y Z

Figure K-5 illustrates the link between the KJ Analysis and its related tools and techniques, showing how some provide input to, while others utilize the information from a completed KJ.

KJ Analysys

389

HOG or QFD

VOC Data Gathering (Survey, Interviews)

Matrix Written Report

KJ Analysis

A Tree Diagram

B C

Brainstorming

D Fishbone

Figure K-5: KJ Analysis Tool Linkage

E F G H

Variations Affinity Diagram The Affinity diagram involves less detail and analysis than what is described in this entry. (See Also “Affinity Diagram,” p. 136)

I J K L M N O P Q R S T U V W X Y Z

390

Encyclopedia

M Market Perceived Quality Profile (MPQP) A B C D E F G

What Question(s) Does the Tool or Technique Answer? How does the market perceive the quality of the product and/or services offerings versus competition? An MPQP helps you to • Estimate the market’s perception of current product and services offerings to evaluate and develop a response and go-forward strategy

H I J K L

When Best to Use the Tool or Technique Early in the process when gathering requirements from customers, use the MPQP tool to understand current gaps in quality compared to competition’s offerings, based upon existing data and internal opinions.

M N O P Q R S T U V W X Y Z

Brief Description The Market Perceived Quality Profile is a technique that identifies and measures market perceived quality of an offering, relative to meeting customers’ requirements and expectations. Quality, as defined by the marketplace, can represent several dimensions—feature/functionality, durability, value, ease of use, and total cost of ownership. MPQP estimates and profiles your company market perception relative to competition. The tool uses data about both the market and market segment to evaluate the value that your company and the key competitors deliver. Presuming that customers prefer to buy on value, the tool calculates value as a function of quality relative to price. It dissects quality into product and/or services performance and features and then contrasts them to their respective market price. Customers perceive quality by how much an offering conforms to their expectations or requirements. The magnitude of conformance of both implicit and explicit expectations determines how much they are willing to pay for it. To start, MPQP utilizes a set of default customer quality criteria to determine value as surrogate Voice of the Customer (VOC) requirements as to the critical quality parameters. However, it should be adapted, expanded, and refined to meet individual market needs and reflect actual VOC input, if available. The default set of quality characteristics varies depending on whether the offering is a product, service, or both.

Market Perceived Quality Profile (MPQP)

391

Dimensions of Quality For product quality, MPQP utilizes David Garvin’s eight dimensions of quality to represent customer’s criteria. They are as follows: • Performance—Primary functional characteristic of a product or service

A

• Features—Secondary characteristics of a product or service

B

• Reliability—Frequency with which a product or service fails

C

• Conformance or consistency—Ability to fulfill specifications or comply with standards • Durability—Product resilience and life

D E F G

• Serviceability—Speed, courtesy, and competence of repair or adjustment

H

• Aesthetics—Form, fit, and finish

J

• Perceived Quality—Reputation For service quality, MPQP utilizes Berry’s ten characteristics to represent customer’s criteria. They are as follows:

I K L M N

• Reliability—Consistency of performance

O

• Responsiveness—Timeliness of service

P

• Competence—Possession of required skills and knowledge

Q

• Access—Approachability and ease of contact • Courtesy—Politeness, respect, consideration, and friendliness

R S T

• Communication—Listening to customers and keeping them informed in a language they can understand

U

• Credibility—Trustworthiness, believability, honesty

W

• Security—Freedom from danger, risk, or doubt

X

• Understanding the customer—Efforts to understand the customer’s needs • Tangibles—Physical evidence of the service The MPQP calculates a gap analysis to estimate where your company’s score is relative to its competitors in the market perceived quality. The MPQP is used to help identify opportunities for growth and threats that

V

Y Z

392

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

present a potential risk to your current and future near-term offerings. The tool starts with internal VOC data (actually it is the voice of your marketing experts plus secondary market research data). Then the estimated-MPQP profile is validated through actual customer interviews. This market validation process is essential to confirm your competitive advantage and customer requirements.

Competitive Summary Market validation confirms the competitive marketplace drivers as quality-based or price-based. The quality-based markets favor dominant traits in technology and performance leadership. The price-based markets characteristically prefer the low cost production of commodity products. The MPQP tool to summarize the competitive nature of the marketplace is the Competitive Position Matrix, illustrated in Table M-1. In general, the competitive summary determines which market player possesses the dominant theme relative to quality or performance sensitivity. Once the competitive landscape is defined, the question is raised as to what gaps need to be closed with respect to quality performance, based on the customer expectations.

Customer Requirements Market validation also confirms the VOC requirements as to the critical quality parameters and their relative rank order. During this phase, any customer’s new, unique, and difficult (NUD) requirements should be solicited as input to the design of future offerings. In turn, these NUD requirements and the MPQP customer quality criteria feed into the future offering’s Quality Function Deployment (QFD). (Moreover, the MPQP gap can be used in the competitive column of the QFD tool.) Responding to the NUD requirements will increase the likelihood of reaching the full entitlement of that offering’s business case. (See Also “Kano Model: NUD versus ECO,” section of “KJ Analysis,” p. 375 and Quality Function Deployment (QFD), p. 543, for more detail.) Ensure that the customer perspective is broad enough to cover not only the current customer base, but prospective customers. Data also should be collected from non-customers. Within the served market, identify key product and service attributes that affect the purchase decision (purchase criteria, dimensions of performance, needs). As previously mentioned, start by using the default Dimensions of Quality (either Garvin’s eight or Berry’s ten). The sample customer data needs to reflect the market segmentation within the served market. For these market segments, identify the quality attributes (dynamics) that help characterize the served market. In addition, the sample customer data need to reflect the functional disciplines representing the target audience. These functions drive the

Market Perceived Quality Profile (MPQP)

393

actual purchasing decision within the customer account. Identify the different functions that consume or appreciate the quality attribute(s) represented in the customer’s purchasing decision group (for example, purchasing, operations, and design). Determine the unique attributes of quality that this group values.

Completed MPQP A completed MPQP helps define opportunities to improve your competitive position. As a result, subsequent strategies should focus on improving your company’s performance ratings, shifting focus to the criteria where your company has a sustainable competitive advantage, and focusing on market segments representing a new or sustainable competitive advantage. A completed MPQP (as shown in Table M-2) helps a company answer the following key business questions: • What are the major product and service attributes that affect the customers’ decisions to purchase from your company rather than a competitor?

A B C D E F G H I J K

• Who are your major competitors?

L

• How do customers perceive your company’s product/service offering versus your competitors’?

M

• Can the Key Customer Criteria suggest meaningful ways to segment the market? Are market dynamics active in these areas of quality? (You may use the appropriate default set—Garvin’s eight product or Berry’s ten service dimensions of quality.)

O

• Can you identify areas where your company can improve its estimated market perceived quality profile? • Which quality attributes represent NUD opportunities? These attributes could come from one of the following sources: • Baseline—All providers perform well; no competitive edge.

N P Q R S T U V W

• Competitive—Differences in performance determine competitiveness and thus opportunities or threats.

X

• Secondary—Currently a competitive attribute, but catch up efforts and/or declining weight might take away the top performer’s competitive edge.

Z

• Dynamic Opportunities—Will become a Competitive attribute when some provider pulls ahead and/or more weight shifts to this quality attribute.

Y

394

Encyclopedia

How to Use the Tool or Technique The MPQP process is comprised of two parts—the Competitive Position Matrix and the actual Market Perceived Quality Profile (MPQP) Matrix.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

The first component summarizes the competitive landscape using a Competitive Position Matrix, to assess a company’s relative position on quality and value to its competitors. This matrix provides input into the MPQP. The second component is the actual Market Perceived Quality Profile (MPQP). To simplify the calculations, use a software application such as Microsoft Excel to build the matrix and embed the various formulas. Step 1.

Complete a competitive summary by using the Competitive Position Matrix. a. Select your company’s leading competitors relative to a particular offering (product and/or service). b. Enter your company’s percent of market share and that of your leading competitors in the second row of the matrix—labeled “Market Share.” c. Enter the price index of your leading competitors’ products and/or services as a percentage relative to your company’s price in the third row of the matrix—labeled “Price Index.” The relative price index is expressed as a percentage compared to your company’s priced offering (that is, Competitor A’s price is 110%, or 90% of your price). d. Enter the direct cost of the leading competitor’s offering as a percentage relative to your company’s direct costs in the matrix’s fourth row—labeled “Direct Costs.” This cost index will be an estimate based on benchmarking research or marketing data. e. Identify whether your competitors’ technology and/or performance quality positions are ahead, equal, or behind relative to your company’s offering and record that position in row five—labeled “Technology/Performance Quality.” Table M-1 provides a snapshot of the competitive landscape comparing four dimensions: 1) market share, 2) price, 3) direct cost, and 4) quality and/or technology performance.

Market Perceived Quality Profile (MPQP)

395

Table M-1: Competitive Position Matrix Our Company

Competitor A

Competitor B

Competitor C

Market Share Price Index (% relative to ours)

100%

Direct Costs Index (% relative to ours)

100%

A

Technology / Performance Quality (relative to ours)

Equal

B

Step 2.

C

Define your estimated competitive gaps by completing the Market Perceived Quality Profile (MPQP) matrix, as illustrated in Table M-2.

D

a. Create an eleven column L-shaped matrix, preferably in Microsoft Excel. Label the columns as follows:

G

i. Key Customer Criteria ii. Relative Importance, Criteria Weight iii. Relative Importance, Share of Importance iv. Quality Performance Rating, Our Company v. Quality Performance Rating, Competitor A (per the leading competitors selected in Step 1.a.) vi. Quality Performance Rating, Competitor B (per the leading competitors selected in Step 1.a.)

E F H I J K L M N O P Q

vii. Quality Performance Rating, Competitor C (per the leading competitors selected in Step 1.a.)

R

viii. Weighted Ratings, Our Company

T

ix. Weighted Ratings, Quality Performance Leader x. Weighted Ratings, Share Leader xi. Gap (Our Company versus Quality Performance Leader) (See Also “Matrix Diagrams—7M Tool,” p. 399, for details about different shaped matrices.) b. Select the set of appropriate Key Customer Criteria and list them down the first column of the MPQP matrix. You may use actual VOC data or start with and/or modify the appropriate Garvin’s eight dimensions for product

S U V W X Y Z

396

Encyclopedia

quality or Berry’s ten dimensions for service quality, which were previously mentioned. In the last row of this list, record the name “Total” in the final cell to serve as the row heading for the total scores that will be calculated. A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

c. Estimate the Customer Criteria stack rank order as relative weights from 0 to 100 points, with 1 as low and 100 as high, for each and document them in the second column—labeled “Relative Importance, Criteria Weight.” Complete this task until each Customer Criteria is stack ranked. Avoid duplicate (or tie) scores to better differentiate one from another. Add the scores in the column, and document it in the final row of that column, labeled “Total.” d. Calculate the percent importance contribution of each Customer Criteria relative to the total and record it in the third column—labeled “Relative Importance, Share of Importance.” The calculation formula is: the individual Key Customer Criteria Weight divided by the Total Criteria Weight, times 100%. Complete this calculation for each cell in the third column containing a Key Customer Criteria Weight. e. Using the same leading competitors selected in Step 1.a., determine the relative Performance Rating for them and your company for each Key Customer Criteria, using the scale of 1 to 10, with 1 as low and 10 as high. Record the rating score in the appropriate columns four through seven—labeled “Quality Performance Rating for Our Company or Competitor A, B, or C.” Avoid duplicate (or tie) scores for a given Key Customer Criteria to better differentiate one from another across the market players. Complete this relative comparison rating for each cell in the columns four through seven aligned with a Key Customer Criteria Weight. f. Determine your company’s Weighted Ratings for each Key Customer Criteria by multiplying its performance rating by the share of importance percentage and recording it in the eighth column—labeled “Weighted Ratings, Our Company.”

Market Perceived Quality Profile (MPQP)

397

The calculation formula is: the individual Share of Importance times the Quality Performance Rating, Our Company. Complete this calculation for each cell in the eighth column aligned with a Key Customer Criteria Weight. g. Determine the Quality Performance Leader’s Weighted Ratings for each Key Customer Criteria by selecting the market player with the highest Performance Rating (from columns four through seven); multiplying it by the share of importance percentage; and recording it in the ninth column—labeled “Weighted Ratings, Quality Performance Leader.” The calculation formula is: the individual Share of Importance times the highest Quality Performance Rating found across columns four through seven. Complete this calculation for each cell in the ninth column aligned with a Key Customer Criteria Weight. h. Identify the market share leader from the data gathered in Step 1. Determine the Weighted Ratings Share Leader scores for each Key Customer Criteria by using the market share percentage as a constant and multiplying it by the Share of Importance, and recording the calculation in the tenth column—labeled “Weighted Ratings, Share Leader.” The calculation formula is: the individual Share of Importance times the market share percentage as a constant, times 100%. Complete this calculation for each cell in the tenth column aligned with a Key Customer Criteria Weight. i. Calculate the Offering Quality Gaps by subtracting the Quality Performance Leader’s Weighted Rating score from your company’s score and recording it in the appropriate cell of the eleventh column—labeled “Gap.” A rating of either a positive or negative score is possible. If your company is the Quality Performance Leader for a given Key Customer Criteria, the gap score will be zero. The calculation formula is: the Individual Weighted Ratings, Our Company minus the corresponding Weighted Ratings, Quality Performance Leader for the appropriate Key Customer Criteria. Complete this calculation for each cell in the eleventh column aligned with a Key Customer Criteria Weight.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

398

Encyclopedia Table M-2: Market Perceived Quality Profile (MPQP) Sample Matrix Relative Importance (Avoid ties)

Key Customer Criteria

A B C D E F G H I

Criteria Weight (1-100)

Quality Performance Rating

Gap

Weighted Ratings

(Rating scale 1 to 10; avoid ties or duplicates)

Quality Share Competitor Competitor Competitor Share of Our Our A Importance Company B C Company Perf. Leader Leader

(Our Company versus Quality Perf. Leader)

Product or Service Performance

90

30.0%

8

9

6

7

2.4

2.7

13.5

Proprietary Features

45

15.0%

9

6

5

4

1.4

1.4

6.8

0.0

Reliabilility

60

20.0%

5

6

4

7

1.0

1.4

9.0

-0.4

-0.3

Conformance or consistancy

15

5.0%

6

7

8

9

0.3

0.5

2.3

-0.2

Durability

30

10.0%

3

5

4

8

0.3

0.8

4.5

-0.5

Serviceability

25

8.3%

7

8

5

6

0.6

0.7

3.8

-0.1

Appearance or aesthetics

20

6.7%

8

9

6

7

0.5

0.6

3.0

-0.1

Quality Reputation & intangibles

15

5.0%

6

9

4

8

0.3

0.5

2.3

-0.2

300

100.0%

Total:

Table M-2 uses an example wherein the market share leader possesses a 45% share. This percentage is used to calculate the Weighted Ratings, Share Leader scores found in column ten.

J K L M

Supporting or Linked Tools Supporting tools that might provide input when developing an MPQP matrix include

N

• Benchmarking (See Also “Benchmarking,” p. 160)

O

• KJ Analysis (See Also “KJ Analysis,” p. 375)

P Q R S T U V W X Y Z

• Porter’s 5 Forces (See Also “Porter’s 5 Forces,” p. 464) • SWOT (See Also SWOT (Strenghs-Weaknesses OpportunitiesThreats), p. 699) • VOC gathering techniques (See Also “Voice of Customer Gathering Techniques,” p. 737) A completed MPQP matrix provides input to tools such as • Fishbone (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173) • FMEA (See Also Failure Modes and Effects Analysis (FMEA), p. 287) • GOSPA (See Also Goals, Objectives, Strategies, Plans, and Actions (GOSPA), p. 320) • QFD (See Also Quality Function Deployment (QFD), p. 543) • Solution Selection techniques (See Also “Solution Selection Matrix,” p. 672 and “Pugh Concept Evaluation,” p. 534) Figure M-1 illustrates the link between the MPQP matrix and its related tools and techniques.

Matrix Diagrams—7M Tool

399

Bencharking Fishbone KJ Analysis FMEA Porter’s 5 Force

A B

MPQP GOSPA

SWOT

C D E

QFD VOC Data Gathering (Survey Interviews)

Figure M-1: MPQP Tool Linkage

F G H I J K

Additional Resources or References • Gale, B. T., Managing Customer Value: Creating Quality and Service That Customers Can See, New York: The Free Press, 1994; ISBN: 0-02-911045-9. • Garvin, D. A., Managing Quality, New York: The Free Press, 1988; ISBN: 0-02-911380-6.

L M N O P Q R

Matrix Diagrams—7M Tool

S T U

What Question(s) Does the Tool or Technique Answer? How do these two (or more) groups relate to one another? Matrix diagrams help you to • Communicate how groups relate to one another and often the strength of that relationship • Collect, organize, plan, manage, prioritize, focus on, and decide about a specific topic

V W X Y Z

400

Encyclopedia

Alternative Names and Variations This tool is also known as • Chart A B

• Checklist (See Also “Checklists—7QC Tool,” p. 204) • Table

D

Variations on the tool include a specialized matrix used to collect, organize, plan, manage, prioritize, and focus on a specific topic. Often these specialized matrices are known by a unique name.

E

Examples of such specialized matrices include

C

F G H I J K

• Cause-and-Effect Prioritization matrix (See Also the “Cause-and-Effect Prioritization Matrix,” p. 188) • Communication plan (See Also the Communication Plan “Variation,” section, p. 405 for an example.) • Control plan (See Also the Control Plan section, p. 406 for an example.)

L

• CTQ matrix (See Also “Critical to Quality (CTQ),” p. 242)

M

• Decision Authority Matrix (See Also the Decision Authority Matrix section, p. 408 for an example.)

N O

• FMEA (See Also Failure Modes and Effects Analysis (FMEA), p. 287)

P

• Launch plan (See Also the Launch Plan (or Implementation Plan, Pilot Plan) section, p. 409 for an example.)

Q R S T U V W X Y Z

• Prioritization matrices (See Also “Prioritization Matrices—7M Tool,” p. 470) • Pugh Concept Evaluation (See Also “Pugh Concept Evaluation,” p. 534) • RACI (See Also “RACI Matrix (Responsible, Accountable, Consulted Informed),” p. 554) • Real Win Worth (RWW), uses three set of specific matrices. (See Also “Real Win Worth (RWW) Analysis,” p. 560) • Quality Function Deployment (QFD) or House of Quality (HOQ) (See Also “Quality Function Deployment (QFD),” p. 543) • SIPOC (See Also “SIPOC (Supplier-Input-Process-Output-Customer),” p. 663) • Solution Selection Matrix (See Also “Solution Selection Matrix,” p. 672) • Transition plan (See Also the Transition Plan section, p. 410 for an example.)

Matrix Diagrams—7M Tool

401

When Best to Use the Tool or Technique The matrix is such a flexible tool that it can be used at any time during a project or process. If it is to be used to communicate the current (prior to any improvements) state, then it should document the improved state as well.

Brief Description The matrix is a management and planning tool that evolves depending on the complexity of the situation. It organizes data or knowledge to examine the relationship between two or more groups of topics. Fundamentally, it is a grid-structured tool comprised of rows and columns. The simplest matrix evaluates two groups, where one is plotted horizontally as column headings (for example, the X-axis), and the second is plotted vertically as row headings (for example, the Y-axis). The strength of relationship is indicated at the intersection of a row and column, in the grid’s cell. The cell data typically is comprised of numbers, letters, or symbols. Generally, the associations being explored in a matrix can include: objectives and methods (for example, current approach, results, or possible solution), causes and effects, categories and respective results, and tasks and people. The preceding list of matrix variations align to these association categories as follows: • Objective and Method: • Communication Plan • Control plan • CTQ matrix

A B C D E F G H I J K L M N O P Q R

• Launch plan

S

• Prioritization matrices

T

• Pugh Concept Evaluation

U V

• Quality Function Deployment (QFD) or House of Quality (HOQ)

W

• Solution Selection matrix

Y

• Transition plan • Cause-and-Effect: • Cause-and-Effect Prioritization matrix • FMEA

X Z

402

Encyclopedia

• Category and Result: • Checklists • Real-Win-Worth (RWW) A B

• SIPOC • Tasks and People:

C

• Decision Authority Matrix

D

• RACI

E F G H I J K L M N O P Q

The shape of a matrix depends on the number of groups or topics being compared. As a comparison category gets added, the dimension of the matrix changes from a simple two-dimension framework comparing two groups to comparisons between and among multiple groups. Hence, there are six different types of matrices: • L-shape—Comparing two groups. Simplest of matrices, showing a one-for-one relationship of two topics, where the rows and columns of a grid form an sideways “L.” This is the most common matrix shape and depicts how Topic A relates to Topic B. • T-shape—Comparing three groups, one topic relative to two dependent topics. This matrix compares two topics to a common third topic. The common topic is plotted along the X-axis that divides the other two related topics—one plotted on the upper Yaxis, and the second plotted on the lower Y-axis, thereby forming a sideways “T.”

R S

Note

T

The common topic relates to two dependent topics; however, the two

U

dependent topics are not compared to one another. Topic A relates to

V

Topics B and C.

W X Y Z

• Y-shape—Comparing three groups, showing a circular relationship among the three. This matrix combines two L-shaped matrices along the Y-axis to create a three-dimension grid, shaped similar to a house (with a roof). Topic A relates to Topic B, which relates to Topic C, which relates to Topic A.

Matrix Diagrams—7M Tool

• C-shape—Comparing three groups, showing a criss-crossed relationship among the three topics. The matrix combines two L-shaped matrices along the Y-axis to represent two topics, and a third topic is plotted along the Z-axis, as the third dimension, to form a cubed structure. (Often one of the L-shaped matrices may be at a 45-degree angle to the other.) This relationship is often difficult to draw, thus often left to computer software to create. (Software packages tend to be specific to the context, such as engineering or healthcare.) Topics A, B, and C simultaneously relate to one another.

403

A B C

• X-shape—Comparing four groups—in which two sets of topics are compared. This matrix essentially combines two T-shaped matrices with a common Y-axis but different X-axes. Topic A relates to Topics B and C, and Topic D relates to Topics B and C, but it does not examine the relationship between Topics B and C.

D

• Roof-shaped—Compares one set of topics internally to one another. This matrix forms a triangle that indicates the interrelationship of related topics to one another, such as potential features (or components) within a product (or service). Matched topics with a relationship may indicate a positive reinforcing and supportive, or negative, counter-balancing effect. These cells contain a symbol to indicate the strength of the relationship [that is, circle, concentric circles (or doughnut), triangle, square, arrow]. Blank cells indicated no relative impact on one another. It usually accompanies an L-shaped or Tshaped matrix and commonly caps a Quality Function Deployment (QFD) or House of Quality (HOQ) matrix. Topic A relates to Topic B within the same group.

H

Figure M-2 illustrates the different shaped matrices. Matrix diagrams are a member of the 7M Tools, attributed in part to one of the first quality thought-leaders, Dr. Shewhart, as seven “management” tools, or sometimes referred to as the “7MP” or seven management and planning tools. These 7M Tools make up the set of traditional quality tools used to analyze quantitative data. The 7M Toolset includes: 1) Activity network diagrams or Arrow diagrams; 2) Affinity diagrams; 3) Interrelationship digraphs or Relations diagrams; 4) Matrix diagrams; 5) Prioritization matrices, often replacing the more complex Matrix data analysis; 6) Process decision program charts (PDPC); and 7) Tree diagrams. The Quality Toolbox, by Nancy Tague, presents the 7M Tools ranked from those used for abstract analysis to detailed planning: Affinity diagram, Relations diagram, Tree diagram, Matrix diagram, Matrix data analysis (commonly replaced by a more simple Prioritization Matrix), Arrow diagram, and Process Decision Program Chart (PDPC).

E F G I J K L M N O P Q R S T U V W X Y Z

404

Encyclopedia

Location A1

Location A2

Location A3

Team C1

Team C2

B C D

Location A1

Location A2

Location A3

Problem B1

1 ic B Top B2 ic p To

Problem B2

T-shaped Matrix

C-shaped Matrix = simultaneous 3-way relationship among Location A2, Topic B1, and Team C1

Cause C1

+

Cause C2 Team D1

Team D2

Location A1

Location A2

Location A3

-

Problem B1 Problem B2

I

X-shaped Matrix

J L

Team C2

1 nA atio Loc

Cause C2

H

K

Team C1

2 nA atio Loc

Cause C1

F G

Y-shaped Matrix

L-shaped Matrix

E

Location A2

Problem B2

Problem B2

A

Location A1 Problem B1

Problem B1

Part A1

Part A2

Part A3

Roof-shaped Matrix (Part A1 positively related to A3 and negatively related to A2)

Figure M-2: Different Types of Matrices

M N O P Q R S T U V W

How to Use the Tool or Technique Matrices that serve a specific function and are so named for that purpose have a unique set of guidelines or procedures, for example one that prioritizes potential root causes is called a Cause-and-Effect Matrix. A separate entry in this reference book documents many of those unique matrix tools (Please see “Cause-and-Effect Prioritization Matrix,” “CTQ Matrix,” “RACI,” “SIPOC,” and so on). A list of such matrices with the exact entry name and page numbers to reference is documented in the previous “Alternative Names and Variations” section. When creating a generic matrix, the procedure is as follows: Step 1.

Define the purpose of the matrix—its objective/what relationship is to be studied.

Y

Step 2.

Identify the topics or groups to be compared.

Z

Step 3.

Select the appropriate matrix format and symbols to be used.

Step 4.

Complete the matrix diagram.

X

Matrix Diagrams—7M Tool

405

Additional Resources or References • U.S. Government web sites with instructions and diagramming help: http://www.usbr.gov/pmts/guide/toolbox/matrixdi.html , and http://www.ed.gov/inits/americareads/resourcekit/MakingInfo/ title2.html • Visual Matrix, http://www.visualmatrixpms.com/ ?gclid=CJXz_9mj6okCFSIOgQodjinUGw.

A B C

Variations

D E

Communication Plan To plan and organize the execution of a project or process communication plan, a customized matrix tool can identify the key message, target audience, type of communication vehicle, timing of delivery, and who should develop and distribute the final communiqué. A communication plan can take on several dimensions. Its objective can show the relationship between the sender (or deliverer) of a message and the receiver. Another objective could demonstrate the relationship between the message and different recipients. A more complex plan could examine the relationship among what needs to be communicated to whom; why is it important; how the message will be communicated (in what format and venue) and when; who needs to develop and deliver the message and by when; and how much the development and delivery will cost. A communication plan also may consider how the sender can test for understanding with the receiver (or a question-and-answer venue) or how the receiver can notify the sender that the message was received. These multiple dimensions may dictate multiple formats, such as a written report, presentation, or matrix. Sometimes a matrix can suffice alone or be incorporated within a report as a planning and management tool. Figure M-3 shows a sample communication plan template for a marketing department to develop and deliver key customer messages. Figure M-4 shows a sample of a project team’s communication plan to show how they will work together.

F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

Completion Date (Target); Actual

Team Member 8

Team Member 5

Team Member 6

Delivery

Completion Date (Target); Actual

Frequency (When)

Team Member 4

Method (How)

Team Member 3

Objective (Why they care)

Team Member 2

Target Audience (To Whom)

Project Manager

Communication Message (What; Topic)

Team Member 1

Development

Team Member 7

406

A B C D E

Objective: (D)irective; ( I )nformative; ( O)ther Method: (E)Mail; ( R)eport; ( M )emo; ( V)oicemail; (CC) Conference Call; ( Mtg) Meeting ; ( B)rochure Frequency : (D)aily; (W)eekly; ( M)onthly; (Qtr ) Quarterly; ( A)nnually

Figure M-3: Example of a Communication Plan Template for Messages

J K L M N O P Q R S T U V

Team Member 6

Team Member 5

Frequency

Team Member 4

Method

Team Member 3

Communications Meetings (Oral) Sub-team Project Team Status Management Review Interim Project Review Post Implementation Customer/Client

Team Member 2

I

Team Member 1

H

Project Manager

G

Project Customer

F

Reports (Written) Meeting Agenda Meeting Minutes Action Items Project Schedule Variance Reports Project Status Report Method : ( E)Mail; (R)eport; (M)emo; ( V)oicemail; (CC) Conference Call; ( Mtg ) Meeting Frequency : ( D)aily; ( W)eekly; (M)onthly; ( Qtr ) Quarterly; (A)nnually

Figure M-4: Example of a Communication Plan Template for a Project Team

W X Y Z

Control Plan To prepare for transition from an improvement project environment to the process players or to simply manage a process, a customized matrix tool can identify key control plan elements. The document might include the key process parameters to be monitored, their respective metrics, target values, source of the data, and who is accountable to gather, monitor, control, and provide status on each factor.

Matrix Diagrams—7M Tool

Similar to a communication plan, a control plan can take on multiple dimensions. It should document how to sustain any improvements, plus manage and adapt the ongoing operations of a process. It defines the technical method of control, and documents the response plan to adjust to any changes in either a written report, standard operating procedure, and/or matrix. The control methods may include a data collection matrix, statistical Control charts, dashboards, or operations review meetings. The response plan involves risk identification and response. Similar to a communication plan, a control plan may contain all these elements in a written report and/or be referenced in a single matrix or several smaller matrices. This type of matrix should be specifically customized for the organization’s purpose. Figure M-5 provides an example of a Control plan using a matrix structure.

407

A B C D E F G H

Control Plan Process Name: Process Owner:

Int/Ext

Prepared by:

Page:

Approved by:

Date:

of

J K

Process Management Components Component

I

Accountable

Key Metrics

Frequency Cycle

Key Customer

Standard Work Policies and Procedures

L

Critical Process Parameter Data Gathering Coordination

M

Planning and Preparation for Operations Review of process performance (calendar, agenda, logistics)

N O

Conduct Operations Reviews Integration of Risk Mitigation Plan (Data and Action Plan integration)

P

Communication Plan (within Process, in-company, external)

Q

Integration of Risk Mitigation Plan Process Player Training Plan (new members and ongoing)

R

Communication Plan (within Process, in-company, external)

S T

Critical Process Parameters Process Step

CTQ Characteristic

CTQ Metric KPOV

KPIV

Target +/or Control Limit USL

Measurement Sample Size Frequency Method

Who Measures

When Decision Rule/ Recorded Corrective Action SOP Reference

USL

U V W X

Figure M-5: Example of a Control Plan Using a Matrix Format

(See Also “Data Collection Matrix,” p. 248; “Control Charts—7QC Tool,” p. 217; Failure Modes and Effects Analysis (FMEA), p. 287; and “Poka-Yoke,” p. 462)

Y Z

D E F G H I J K L M N

This matrix tool identifies the roles or people accountable for making decisions in a process or organization. The Decision Authority Matrix reflects and reinforces the type of decision-making style or culture of the organization— consensus, democratic, participative, or authoritarian. This kind of documentation is helpful on complex, new, or cross-organizational processes. As its name states, a Decision Authority Matrix identifies one, and only one, decision-maker who is accountable for making the decision and its consequences. For complex scenarios, it also defines the decision-making process and who else is providing appropriate background information and context, including those responsible for any analytical activities. Complex processes may identify a person accountable for integrating the relevant data, synthesizing it, and making a recommendation to the decision-maker. At times, decision-making may be delegated or escalated to another role or person, depending on the level of risk and financial impact. Figure M-6 provides an example of a Decision Authority Matrix template.

O

S T U V W

Deliverable/Task Description Decision 1 Decision 2 Decision 3 Decision 4

Member D

R

process

Member C

for

Member B

Q

D = Decision-maker (only 1 entry) R = Recommend C = Concur P = Perform E = Explain

Decision Authority Matrix Member A

P

D

R D

P

E E D

R

P

C D

Member H

C

Member G

B

Decision Authority Matrix To resolve any ambiguity, gaps, or redundancy for a given decision-making process, a customized matrix tool can define the standard procedure. The tool should identify the key players involved in the process, their deliverables, and ultimately who has the final vote. The matrix defines the type of decision-making process pertinent to a given topic, such as authoritative, collaborative, participative, or consensus.

Member F

A

Encyclopedia

Member E

408

P C R P

P

P

C

P

P

R

Figure M-6: Example of a Decision Authority Matrix Template

X Y Z

This tool is often compared with an RACI Matrix, since both tools define the responsibilities for those involved in a process. Although the names are similar, the Decision Authority Matrix distinguishes itself from two other kinds of matrices—a Decision Matrix and a Prioritization Matrix. A Decision Matrix may refer to a specific tool or a category of matrices used to evaluate and prioritize a list of options, often called by a specialized name—Pugh Matrix or Solution Selection Matrix.

Matrix Diagrams—7M Tool

409

A Prioritization Matrix is a category of three different types of matrices used to narrow options by comparing groups. (See Also “RACI Matrix (Responsible, Accountable, Consulted, Informed),” p. 554; “Pugh Concept Evaluation,” p. 534; “Solution Selection Matrix;” p. 672; and “Prioritization Matrices—7M Tool,” p. 470) A

Launch Plan (or Implementation Plan, Pilot Plan) To plan and organize the execution of a product and/or services offering launch to the marketplace, a customized matrix tool can identify the key elements. It should identify the key deliverables, respective metrics, respective target audience (customer), timing of delivery, and who should develop and distribute the final output. In preparation to launch a product or services offering, a launch team uses a launch implementation plan to introduce the offering to the marketplace and the customer value chain. Such a plan can apply to both a pilot, the implementation of an improvement, or the launch of a new offering. This plan documents the readiness requirements, deliverables and process to plan and manage the launch. The plan components may include the following:

B C D E F G H I J K L

• Schedule or timeline of activities, events, deliverables, and milestones.

M

• Work Breakdown Structure (WBS); a Tree diagram or matrix defining the key tasks, deliverables, task owners, task customer, timeframe to complete the task.

O

• Control plan documenting the key process and performance metrics, the score card (or dashboard) to monitor the metrics. • Standard operating procedures, as applicable.

N P Q R S T

• Communication plan defining what will be communicated to whom, when, and how.

U

• Process map or flow (the former and enhanced process). (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

W

• SIPOC to changes in any one of the following areas: suppliers, inputs, high-level process, outputs, and/or customers. (See Also “SIPOC (Supplier-Input-Process-Output-Customer),” p. 663)

Y

• RACI matrix to highlight those involved in the launch assigned tasks, deliverables, and/or responsibilities. (See Also “RACI Matrix (Responsible, Accountable, Consulted, Informed),” p. 554)

V X Z

410

Encyclopedia

• Training plan of the initial and ongoing training and potentially any other human resource concerns (such as compensation or bonus impacts).

A B C D E F G

• Risk mitigation plan identifying potential risks and the corresponding contingency and response plan to each prioritized risk. This plan should include who is accountable for taking action if the risk occurs and any key triggers to detect if the risk occurs. (See Also Failure Modes and Effects Analysis (FMEA), p. 287) • Budget allocated to the launch activities. Figure M-7 illustrates a Launch plan matrix set up similar to a status report as an example. However, this type of matrix should be specifically customized for the organization’s purpose.

H

XXX Launch Plan Key Deliverables

I J K

Task #

N O P

Target Completion

Status

Accountable

Start Date

Completed Date

Responsible Individuals

12/30/20XX

Marketing

1.00

Features & Functionality Documented

1.10 1.10.1

L M

Deliverable Launch XX in Direct Sales + Dealer Channel

Objective:

Draft 1.0 Completed & Reviewed w/C. Magee

2/18/20XX

Completed

G. Testa

1/24/20XX

2/28/20XX

N, Minocha, M. McCandless

1/31/20XX

Completed

G. Daniel

1/24/20XX

1/31/20XX

N, Minocha, M. McCandless

G. Daniel

1/31/20XX

2/2/20XX

2/2/20XX

2/3/20XX

2/2/20XX

2/28/20XX

2/2/20XX

2/28/20XX

2/2/20XX

2/28/20XX

1.10.2

Draft 2.0 Completed & Reviewed w/C. Magee

2/2/20XX

Completed

1.10.3

Draft Reviewed w/M. George

2/3/20XX

Completed

G. Testa

Draft Reviewed w/T. Garrett

2/17/20XX

Completed

G. Testa

1.10.5

Draft Reviewed w/D. Washington

2/17/20XX

Completed

1.10.6

Field Service level of involvement determined

G. Testa G. Testa

1.10.4

Value Proposition Defined by Target Audience

Completed 2/28/20XX

In-process

Target Audience Identified (by market segment)

2/17/20XX

Completed

D.Garcia D.Garcia

1.30.2

Value Propositions drafted

3/30/20XX

Completed

D.Garcia

1.30.3

Value Propositions tested with market segment

5/30/20XX

In-process

D.Garcia

1.30 1.30.1

J. King W. Jewett, S. Jones

2/2/20XX

2/28/20XX

W. Jewett, S. Jones

2/20/20XX

1/30/20XX

W. Jewett, S. Jones W. Jewett, S. Jones

Figure M-7: Example of a Launch Plan Template

Q R S T U V

A well-executed launch implementation plan of a poor idea is more likely to be successful than the other way around. And a well-developed launch implementation plan becomes the control plan. Similar to a communication plan, a launch plan may contain any or all of these elements in a written report and/or referenced in a single matrix or several smaller matrices.

W X Y Z

Transition Plan To plan and organize the hand-off of the project work to the ongoing process owner and players, a customized matrix tool can identify the key elements of a transition plan. Similar to a Launch plan, a Transition plan prepares the final project recommendations for ongoing execution by those working in the process, and allows the project team to disband while maintaining or sustaining the desired project results.

Matrix Diagrams—7M Tool

411

In preparation to closing out a project, a project team uses a transition plan to hand off its improvements to the ongoing process players in the organization. This plan documents the changes made to the current process and may include the following: • Training plan of the initial and ongoing training • Process map or flow (the former and enhanced process) (See Also “Process Map (or Flowchart)—7QC Tool” p. 522) • SIPOC to changes in any one of the following areas: suppliers, inputs, high-level process, outputs, and/or customers (See Also “SIPOC (Supplier-Input-Process-Output-Customer),” p. 663) • RACI matrix to highlight changes in the process players’ assigned tasks, deliverables, and/or responsibilities (See Also “RACI Matrix (Responsible, Accountable, Consulted, Informed),” p. 554) • Financial plan or budget to document any initial or ongoing funding requirements of the improvements Similar to a communication plan, a transition plan may contain all these elements in a written report and/or referenced in a single matrix or several smaller matrices. This type of matrix should be specifically customized for the organization’s purpose. Figure M-8 provides an example of a Transition plan using a matrix structure.

Int/Ext

Prepared by: Approved by:

Page: Date:

of

Component

Accountable

Key Metrics

Frequency Cycle

Key Customer

Process Map (revised as necessary)

E F G H I J K L M N O Q S U

RACI Document (responsibilities revised as necessary) Critical Process Parameter metrics (revised as necessary)

V

Other Improvement Revisions Improvement Implementation Risk Mitigation Plan (Integrated with ongoing process)

W

Change Management Plan (integrated with ongoing process)

X

Communication Plan of Improvements (within Process, in-company external) Process Player Training Plan (current process players, and added to new members training)

Y

Operations Review of Hand-off (after XXXX $ of observed cycles) Project Close-Out document (after transition completed and approvals gained)

Z

Revised Critical Process Parameters CTQ Metric KPOV

D

T

Standard Work Policies and Procedures (revised as needed)

CTQ Characteristic

C

R

Process Management Components

Process Step

B

P

Transition Plan Process Name: Process Owner:

A

KPIV

Target +/or Control Limit USL

Measurement Sample Size Frequency Method

USL

Figure M-8: Example of a Transition Plan Template

Who Measures

When Decision Rule/ Recorded Corrective Action SOP Reference

412

Encyclopedia

Measurement System Analysis (MSA)

A B C

What Question(s) Does the Tool or Technique Answer? How accurate is the measurement system? Is the process truly performing the way the data seems to be reporting, or is the measurement system inaccurate? MSA helps you to

D

• Determine the reliability of the measurements

E

• Understand the integrity of the data

F G H I J K

Alternative Names and Variations This tool is also known as • This tool is also known as theMeasurement System Evaluation method

L M N O P

When Best to Use the Tool or Technique Before any data is collected, evaluate the measurement system (the calibration of device(s) and utilization procedure) for accuracy to ensure that it is not a source of variability and that in fact any observed variation is due to the product or process of interest—ensuring data integrity.

Q R S T U V W X Y Z

Brief Description In Six Sigma, variation is the enemy, particularly if it is discernable by the customer. Often times unknowingly, the measurement system can introduce variability (noise) in the measurement data. Thus, observed variation may contain two sources of dispersion—the product (or part) itself and how it is measured. The objective is to minimize those controllable factors that exacerbate variation in data. The total variability observed equals the sum of product (or part variability) plus the variability in the measurement system of the product (or part). The equation is written as: Total standard deviation (σTotal2) = Product (or Process) standard deviation (σProduct2) + Measurement System standard deviation (σMeasurement System2). A measurement system is comprised of the measurement device (or tool) and its calibration, the procedure to take a measurement with the device, and the person or

Measurement System Analysis (MSA)

413

machine performing the process with the device. The utmost data integrity represents isolating any detected variation as attributed to the product (or part) of interest and not the measurement system. The Measurement System Analysis (MSA) is a set of tests designed to evaluate how much variability is in the measuring device. The objective is to ensure that the measurements reflect the true nature of what is being measured and avoid a poor (noisy) measurement system. The measurement results equals the true value plus the error of a measuring instrument, represented by the equation: Measurement standard deviation (σMeasurement2) = True standard deviation ( σTrue2) + Error ( σError2). However, measurement systems have the potential to add a large amount of variability to the data of interest. A poor measurement system could cause an organization to reject products that are actually good, simply due to inaccurate and unreliable measurements—thereby potentially rendering the process incapable of meeting target. If a poor measurement system exits, it may be unable to detect subtle differences that are important. The measurement system needs to be discriminating enough (have enough resolution) to detect differences of interest. It is important to conduct an MSA to validate the accuracy of the measurement system before determining the baseline capability of the process. Ideally, measurements are accurate and reliable and do not introduce any variability; if not, calibrate the measurement system before proceeding with the baseline capability test. The MSA tries to avoid the results of the old adage of “garbage in, garbage out” from occurring—so that good output is properly accepted and bad output is properly rejected. (See Also “Process Capability Analysis,” p. 486) MSA applies to any kind of data—continuous (variable) or discrete (attribute). As the data type changes, so does the measurement instrument. A tangible item can be evaluated for its size (inches or metrics), weight (pound or grams), or temperature (Celsius or Fahrenheit)— continuous measures. The same item also can also be evaluated for its attributes, such as color, type or name category, or customer satisfaction—discrete measures (count, rank, yes/no, and so on). The appropriate MSA test changes to align with the data type, which is no surprise, given that the measurement instrument differs with the data type as well. (See Also “Graphical Methods,” p. 323 for a more detailed discussion on different types of data.)

Accuracy and Precision Two terms characterize variability—accuracy and precision. Accuracy describes how close to the target; and precision measures the degree of repeatability and reproducibility.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

414

Encyclopedia

Which scenario is preferable—one in which the output is precise but off-target or one that occasionally hits the target (occasionally accurate) but contains a high amount of variability? The answer is precision, but off-target is easier to correct than highly variable, occasionally accurate output, as shown in Figure M-9. A B C D E F G H I

Figure M-9 illustrates a target (a bulls-eye in the center of the circle) that both Lori and George are trying to hit. George’s work may be viewed as more accurate than Lori’s, considering he has hit the target once, but also the output contains a lot of variability. George may need more training or change how he is using the equipment to hit the target; multiple components of his work may need to improve. Lori’s results have missed the target but collectively are more tightly clustered than George’s. Hence, her results are more precise but not accurate. Lori’s work is easier to correct. Lori’s may simply need to aim differently but keep the process that she is using the same. Accuracy is described relative to hitting a target. Precision describes the amount of variability in a set of data. Both an accurate and precise outcome would be output centered directly on the target.

J K

Lori

George

L M N O P Q R S T U

Figure M-9: Accuracy Versus Variability Illustration

V W X Y Z

Accuracy defines the precision with which a target is reached and is measured as the difference between the outcome and a target or standard value. If a sample of items were taken, and each were measured for accuracy, the average of those accuracy measurements for each sample item is known as bias. Figure M-10 illustrates the bias for two distributions.

Measurement System Analysis (MSA)

415

Bias Illustration Standard Value Poor Accuracy

Better Accuracy

Bias

x1

A B

x2

C

Average of multiple measurements of different items

D

Figure M-10: Bias Illustration

E

Precision defines the variation in repeated measurements of the same item. There are two major ways to measure precision—repeatability and reproducibility. Repeatability is calculated by repeated measurements of the same item, by the same person, with the same measuring device. A measurement system is repeatable if the variation of the repeated measurement is tight. Reproducibility is when the variation source is the averages from repeated measurements (or factors) made by different people (or machines) on the same item. This variation between factors is also called operator or technician error. Figure M-11 depicts both repeatability and reproducibility. Repeatability

Better Repeatability

Poor Repeatability

Same item repeatedly measured, by same person, with same measuring device.

Reproducibility

F G H I J K L M N O P

Poor Reproducibility

Q R

Better Reproducibility

Key; Operator 1- solid, Operator 2- dotted, Operator 3- dashed

Same item repeatedly measured, by different people (or machine), with same measuring device.

Figure M-11: Repeatability and Reproducibility Illustration

The operator error depicted in Figure M-11 represents two sources of variation—the differences from one operator’s technique to the next and the interaction between the operator and the part. The latter variation source occurs when an operator is inconsistently handling the part when measuring it, resulting in different measurements, perhaps from different sections of the part or improper measurement of the part. Often times operator error is corrected with training and clarification of the standard operating procedure (SOP). (See Also “Attribute MSA,” p. 425)

S T U V W X Y Z

416

A B C D E F G H I J K L

Encyclopedia Accuracy In summary, continuPoor Good ing with the dart board analogy, it integrates well the concepts of accuracy and precision. It is easier Good to adjust the aim of a dart thrower who consistently hits the same spot than it is to correct someone’s Precision throw whose darts hit all over the place. Thus, it is easier to correct poor Poor accuracy than poor precision. Relative to a target, accuracy describes centering (how tight, consistent), and precision refers Figure M-12: Accuracy Versus Precision to the spread. Figure M-12 examines the relationship between good versus poor accuracy and good versus poor precision.

M N O P Q R S T U V W X Y Z

Stability Over Time Every measurement is a combination of various sources of variability. Ideally, any observed variation would be due solely to any part-to-part variation, and not to the measurement system itself. Part-to-part variation represents the measurement differences from one physical part to another. The ability to discern minor changes in part-to-part variation, or small part characteristics, means the measurement system has a highdegree of resolution or discrimination. The measurement system, as previously discussed, can be a source of variation as well. Stability describes a measurement system that lacks change in bias or precision over time when repeatedly measuring the same items. Hence, a stable measurement system is said to “not drift.” The measurement device may have an inherent bias across its operating range, which is defined by its measurement scale. For example, a device may introduce variability at either extreme of its scale—bias relative to a true value. This type of variability is called linearity. Figure M-13 illustrates both stability and linearity.

MSA Study A Measurement System Analysis (MSA) looks to reduce or eliminate any variability coming from the measurement system to obtain stability.

Measurement System Analysis (MSA)

417

Figure M-14 summarizes the different variation components in a Tree diagram. The MSA focuses on the entire right-half of the Tree diagram, under the Measurement System Variability branch, indicated by the rectangles with rounded corners. Stability — bias and preision over time

Linearity — bias versus true value Undesired Behavior

Undesired Behavior

Measured Value

A B C

Measured Value

Desired Behavior

D E

Desired Behavior Time

True Value

Figure M-13: Illustration of Stability and Linearity

F G H

Observed Measurement Variation

I J

Part-to-Part Variation

K

Measurement System Variation

L M

Reproducibility

Repeatability

N O P

Operator*Part Interaction Variation

Operator-to-Part Variation

Figure M-14: Components of Variation

MSA for Variable Data—Gage R&R Recall that the Measurement System Analysis (MSA) is a set of tests designed to evaluate how much variability is in the measuring device. One test focuses on variable data, known as the Gage Repeatability & Reproducibility (Gage R&R, Gauge R&R, or GR&R). This test examines the total variation (TV), the product variability, and the measurement system variability[also known as measurement system error or precision (P)]. Total variation (TV) equals the product variation plus the measurement variation. The Gage R&R contains three tests—% Contribution, P/TV, and P/T:

Q R S T U V W X Y Z

418

Encyclopedia

• % Contribution

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

The ratio of variances of the measurement system standard deviation (σMeasurement2) to the total standard deviation ( σTotal2 ) times 100% defines the %Contribution test for each component of variation. This test is the most common or popular of the three tests because it easily identifies the culprit contributing the largest amount of variation. The Acceptance Criteria for this test is 30%

Inadequate

D

>5

Excellent

E

=4

Marginal

F

80%); greater than or equal to 90% is excellent ( > 90%). A

Operator error typically can be improved by tactics such as • Training • Improving the Operational Definition

B C D E

• Changing the process, or environment.

F

• Sense multipliers (devices to improve human senses)

G

• Masks/templates (block out unimportant information) • Checklists

H I J

• Automation

K

• Reorganization of work area

L

• Visual aids

M N

Attribute MSA Example Example Scenario: In an administrative department, the measurement system for the invoice process appears to have an increased number of errors, which is causing delays in paying the customers. Randomly collect 30 sample invoices and conduct an Attribute MSA.

O

Following the Attribute MSA procedure previously described, create a data collection sheet in the MINITAB Worksheet to collect the results of the test. The data can be stacked or unstacked. Figure M-20 shows stacked data, wherein there are two appraisers who evaluate the same invoice twice. (Recall that when this test is being conducted, that the sequence of invoices must be random.) Figure M-20 shows the appraiser results in the Result column, C4-T (meaning Column 4-Text), and the Reference Standard in the Expert column, C5-T.

S

MINITAB needs the result data (provided in text) translated to numeric data (a.k.a. coded data) to run the analysis. Label the next column—for this example, label column C6 as Coded Results. To do so, select Data > Code > Text to Numeric and then enter which column contains the result text data that requires coding (Area 1, Code data from

P Q R T U V W X Y Z

428

A B C D E F G

Encyclopedia

columns), which column the newly coded numeric data will be recorded (Area 2, Into columns), and the original to new values, one for one [Area 3, Original values (that is, red light blue)]. In this case, those invoices determined as good will be coded as 1, and the bad as 2. Click OK, and MINITAB will populate the coded data in column C6, as illustrated on the right-half of Figure M-20. Now the data is ready to conduct an Attribute Agreement Analysis among the appraisers. From the MINITAB main screen, select the following commands from its drop-down menu: Stat > Quality Tools > Attribute Agreement Analysis… Figure M-21 displays a sample MINITAB main screen where the appropriate attribute test data is selected to produce the final sample graphical display showing the Within Appraiser and Appraiser versus Standard results.

H I J K L M N O P Q R S T

Figure M-20: Example of MINITAB’s Code—Text to Numeric Main Window and Resulting Worksheet

U V W X Y Z

The overlapping confidence intervals, shown in the graph of Figure M-21, indicate no significant difference between appraisers in repeatability (Within Appraisers) and between appraisers in bias (Appraiser vs. Standard). Hence, there is no consistent bias, which is good. Now examine whether there is any specific discrepancy between and within the appraisers’ judgments. To do this, conduct a Gage Run Chart by selecting the following commands from its drop-down menu: Stat > Quality Tools > Gage Study > Gage Run Chart…

Measurement System Analysis (MSA)

429

A B C D Figure M-21: Example MINITAB Attribute Agreement Analysis Main Screen and Graphical Output

Figure M-22 displays a sample MINITAB main screen where the appropriate attribute test data is selected. This test requires the coded (numeric) results versus the text attribute data. Figure M-22 also exhibits the final graphical display of the “Gage Run Chart of Coded Results by Invoice, Appraiser” results for each of the 30 invoice samples in this example. In the Gage Run Chart graph, notice the three highlighted scenarios for the different disagreement combinations—for just one appraiser, between both, and within both.

E F G H I J K L M N O P Q R S T U V

Figure M-22: Example MINITAB Gage Run Chart Main Screen and Graphical Output

W X Y

Supporting or Linked Tools Supporting tools that might provide input when conducting an MSA include • Data collection sheet (See Also “Data Collection Matrix,” p. 248) • Process map (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

Z

430

Encyclopedia

A completed MSA provides input to tools such as • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173) • FMEA (See Also Failure Modes and Effects Analysis (FMEA), p. 287) A B C D

• Graphical methods (See Also “Graphical Methods,” p. 323) • Idea generation techniques to improve a poor MSA (See Also “Brainstorming Technique,” p. 168)

E

• Process capability study (See Also “Process Capability Analysis,” p. 486)

F

• Statistical analytical tools (See Also “Statistical Tools,” p. 684)

G H

Figure M-23 illustrates the link between an MSA and its related tools and techniques.

I J

Cause and Effect diagram

K L

FMEA

M N O P Q R

MSA Process Map

S U

Root cause analysis techniques

V X Y Z

Idea generation techniques

Process capability

T

W

Graphical methods

Data collection sheet

Figure M-23: MSA Tool Linkage

Monte Carlo Simulation

431

Monte Carlo Simulation What Question(s) Does the Tool or Technique Answer? What are the probabilities and risks associated with several possibilities (how about scenarios, outcomes, or decisions)?

A

The Monte Carlo simulation helps you to

B

• Simulate the range of outcome possibilities for scenarios to aid in the decision-making process • For example: scheduling of a project, program, or process; forecasting financials; and managing an offering portfolio • Understand the variability in a process or system • Identify problems within a process, system, or system design • Manage risk by understanding the cost/benefit relationship (See Also “Selecting Project Portfolios using Monte Carlo Simulation and Optimization,” in Part III of this book, p. 921)

C D E F G H I J K L M

When Best to Use the Tool or Technique Construct a Monte Carlo simulation at the beginning of a project or program to explore a range of possibilities and their associated risks and to inform the selection of a path forward with a degree of certainty. This tool is useful for any problem where variation and uncertainty cause risk or defects. It also can be used to solve problems that are computationally intensive by sampling on the computer.

N O P Q R S

Common applications for a Monte Carlo simulation include models for

T

• Business process

U

• Engineering design

V

• Financial planning and forecasting

W X

• Manufacturing production system

Y

• Offering Portfolio management (See Also “Selecting Project Portfolios using Monte Carlo Simulation and Optimization,” in Part III, p. 921)

Z

• Sales planning and forecasting

432

A B C D E F G H I J K L M N O P Q R S T U V W X Y

Encyclopedia

Brief Description A Monte Carlo simulation produces different combinations of possible outcomes for different key parameter distributions. It models variation and uncertainty to help make better decisions. A simulation is the process of experimenting with a model to measure performance and behavior of a system’s inputs. Modeling is an iterative process that tries to replicate or represent a real system. Often times the model is a simple representation of a complex system or process. A model built via a spreadsheet is easy to use and is flexible; however, it is limited to only a one number answer (deterministic), and it is difficult to audit the model, its data and assumptions. The single-point estimates (even if they include the most-likely, best- and worst-case scenarios) lack probability information. In contrast, a simulation model uses random numbers to measure the effects of uncertainty. Monte Carlo gets its name from the city in Monaco that is famous for its roulette wheel as a kind of random number generator. In the 1940s, the Los Alamos National Labs used Monte Carlo to simulate the impact of a nuclear bomb. Interest in generating random numbers dates back to the 1800s when mathematicians were attempting to approximate pi. However, the real power behind simulation modeling came about with increasing computer capabilities. With today’s technological capabilities, computer software packages such as Decisioneering®, Inc.’s Crystal Ball® software, which works on top of Microsoft Excel, have simplified the user interface and run the calculations quickly in the background. Monte Carlo simulation recognizes that independent variables (Xs) in a math model can vary over a range. The Y = f(X) equation describes a dependent variable (Y) as a function of independent variables (Xs). The independent variables can have a range of values. This range can be characterized by a probability distribution. By sampling the values of X parameters and making repeated calculations of the result, the model develops probabilities of the variability to forecast the impact on the dependent Y variables and the resulting risk. (See Also “Y = f(X),” and “Regression Analysis,” p. 758 and p. 571, respectively.) Simulation is an inexpensive means to evaluate decisions prior to action and reveals critical sensitivity points of a system. The results are sensitive to the accuracy of the input data. Simulation does not solve the problem; it simply aides in the decision making process.

Z

How to Use the Tool or Technique For purposes of demonstration, the following procedure develops a Monte Carlo simulation model using Crystal Ball software from Decisioneering. Crystal Ball software runs as an add-in to Microsoft Excel, so both software application packages are required.

Monte Carlo Simulation

433

The procedures to run a Monte Carlo simulation are as follows: Step 1.

Develop a system flow diagram or algorithm. a. Collect the input. Establish the functional relationships (the math model) that exist between the various input assumption terms to produce the forecasted output response.

Step 2.

Record the data in a Microsoft Excel spreadsheet with Crystal Ball software open. (For example, list each project task and record the time associated with the minimum, maximum, and median (or most likely) estimates.) The spreadsheet should have tasks listed down the first column, as the row headings, as shown in Figure M-24.

A B C D E F G H I J K L M N O P Q R S

Figure M-24: Sample Crystal Ball Software Spreadsheet

Step 3.

Use Crystal Ball software to model assumptions and forecast probability distributions. a. Set up the problem by defining the input assumptions. i. Define critical X variable assumptions, by clicking and selecting a cell with the data, and going to the Define drop-down menu and select Define > Define Assumptions. Enter the appropriate assumptions, as shown in M-25. ii. Select and enter the distribution for your assumption. Repeat this step for each component or element in your model.

T U V W X Y Z

434

Encyclopedia

b. Select Define Forecast (Y variable) from the Define dropdown menu (Define > Define Forecast). i. Specify the items being calculated in the forecast formula. Repeat this step for each forecast formula in your model. A

ii. Option in the Define Forecast: highlight the cell containing the total median time estimate, and enter any of the conditions of interest—LSL (Lower Spec Limit), USL (Upper Spec Limit), and/or Target, as illustrated in Figure M-25.

B C D E

iii. More than one forecast can be generated by defining several forecast formulas within the same spreadsheet.

F G H I J K L M N O P Q R

Figure M-25: Sample Crystal Ball Software Define Assumptions and Forecast Windows

S T

Step 4.

U

Run the simulation and analyze the results. a. Choose Run Preferences from the Run drop-down menu (Run > Run Preferences).

V W X

b. Under the Trials tab, type in the number of trials you would like to run—must be a 200 minimum.

Y

c. Run the model by selecting Run > Start Simulation.

Z

Step 5.

Improve the model and/or make decisions. a. Interpret the results using the Monte Carlo outputs— descriptive statistics, frequency chart, and sensitivity chart for the X and Y variables evaluated in the model.

Figure M-26 illustrates the various outputs from a Crystal Ball software simulation and includes a cumulative frequency chart just underneath the histogram. Both graphical charts contain the target indicated by

Monte Carlo Simulation

a line and a text box containing the value, if defined in the Define Forecast drop-down menu. It identifies the optimal statistical test and provides the common statistical data associated with the simulation to the right of the resultant histogram. Place the cursor in the upper-right box of the output screen with the statistical tests and then hit the space bar on your computer’s keyboard to open a second table of statistics containing the Normality test at the top, followed by the various metrics, including process capability. Repeat tapping the space bar to open a third window that contains the percentile distribution. (See Also “Normal Probability Plots, Normal versus Non-normal Data,” p. 227 for a discussion on Normality test; and “Process Capability Analysis,” p. 486)

435

A B C D E F G H I J K L M N O P Q R

Figure M-26: Sample Monte Carlo Simulation Output Using Crystal Ball Software

S T U

Examples Simulation of a Project Management Schedule The scenario involves a project team responsible to develop a product concept. The process starts with gathering the voice of the customer up through the development of specific product requirements. In such a complex, cross-functional project, the schedule could contain some competing critical paths, as the team works toward driving a launch schedule. Traditional PERT analysis looks at the variation of tasks on the critical path but fails to recognize that the critical path can change. Crystal Ball software looks at all paths through the activity network and understands how other paths can influence total time. (See Also “Activity Network Diagram (AND)—7M Tool,” and PERT (Program Evaluation and Review Technique) Chart, p. 127 and p. 453, respectively.)

V W X Y Z

436

Encyclopedia

The team mapped out a 12-step project plan and listed the activity steps down the first column in their Excel spreadsheet, as shown in Figure M-24.

A B

Click inside the first cell with the median task assumptions (that is, cell B4), and from the Crystal Ball software main toolbar or the Excel Define menu, select Define Assumption… from the drop-down menu. The screen to define that cell’s distribution assumptions will open, as displayed in Figure M-27.

C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure M-27: Example Crystal Ball software Define Assumption Dialog

To set up a skewed distribution for the various task durations, enter the different Minimum, Likeliest, and Maximum times in the appropriate boxes and then click Enter and OK. Repeat this step until each of the activity steps have been set up. Use your best judgment on the minimum, likeliest & maximum time assumptions you are entering into Crystal Ball software. For this example, the assumptions can be found in Figure M-25. Define the FORECAST Cell (that is, B17 in Figure M-25) to sum up the time for the critical path for the project plan tasks by using the Excel Summation formula. In this case, enter the summation formula =SUM(B4:B15) to sum the twelve tasks listed from cells B4 to B15. Define the Forecast by selecting Define Forecast… from the Excel Define menu or using the Crystal Ball toolbar icon. Fill in the Forecast Name & Units and then click OK, as illustrated in Figure M-28. Now Crystal Ball software is ready to run the simulation. Select Run Preferences… from the Excel Run menu or by using the Crystal Ball toolbar icon. Select the Trials tab in the upper-left corner of the Run Preferences screen to enter the number of trials to simulate. In this case, enter 10,000 and keep the default of 95% for the confidence level, as shown in Figure M-29. Then select OK.

Monte Carlo Simulation

From the Excel Run menu, select Start Simulation to run the simulation or just click the Run icon on the Crystal Ball toolbar. The Forecast window will fill with the sample data being produced during the 10,000 trials being run and eventually will produce a distribution graph similar to the one found in Figure M-26.

Figure M-28: Example Crystal Ball Software Define Forecast Screen

To view the statistics associated with the Forecast window, select View > Statistics to get the window displayed in the upper-right of Figure M26, or Figure M-31 (or hit the space bar on your keyboard). It provides information about the output distribution, such as the mean, median, mode, and Figure M-29: Example Crystal Ball Software Run Preferstandard deviation. Given ences Screen the forecasted standard deviation, the process capability can be calculated. Also in the Forecast window, select Forecast > Open Sensitivity to see the various sensitivities in the model and which tasks contribute the largest variance, as illustrated in Figure M-30. This information may help identify steps requiring corrective action and specify specific failure modes in an FMEA. (See Also Failure Modes and Effects Analysis (FMEA), p. 287) In this example, the simulation indicates that the project plan can deliver its requirements within the upper specification limit of 175 days (a 95% confidence level). However, as shown in Figure M-34, the process capability of about 1.0 indicates the project is barely capable of completing by the target of 160 days (with a mean of 160.64). The project plan is tight, and all the task estimates need to hit the “most likely” estimate.

Note Figure M-31 represents only the statistical portion of the simulation Forecast output and zooms in on the upper-right corner of Figure M-26 after hitting the space bar once.

437

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

438

Encyclopedia

A B C D E F G H I

Figure M-30: Example Crystal Ball Software Forecast Window and Sensitivity Chart

J K L M N O P Q R S T U V W X Y Z Figure M-31: Example Crystal Ball Software Statistical Portion of the Final Output

Multi-vari Chart

439

As a result, the project team might examine the critical path of its project plan to explore opportunities for change based on the simulation’s results. Perhaps some activities can occur in parallel. Moreover, an expectation setting discussion with the project sponsor would also be appropriate.

Multi-vari Chart

A B C

What Question(s) Does the Tool or Technique Answer? Across multiple sources of variability, which one contributes the most?

D

A Multi-vari chart helps you to

E

• Visually analyze variance from multiple sources

F G

• Understand the differences in the means and variability together in one graphical analysis tool

H

• Visually stratify multiple sources of variation to examine possible patterns and largest contributors

J

I K L

Alternative Names and Variations This tool is also known as

M N

• Multi-vari plot

O

• Multi-vari study

P

When Best to Use the Tool or Technique Before conducting a full analysis of variance, conduct a Multi-vari study to visualize which variable contributes the most variation. Look for patterns, trends, or interactions among the multiple sources to determine where to focus further attention to minimize the variation or identify areas not needing further investigation. (See Also Analysis of Variance (ANOVA)—7M Tool, p. 142)

Q R S T U V W X

Brief Description The Multi-vari chart is a good executive communication tool of a relatively complicated comparative analysis. It is a powerful graphical tool that draws the relationship between one variable with many other

Y Z

440

Encyclopedia

variables, to examine the differences in their means and variability on a single plot. The variation can either be controllable or uncontrollable (or noise). The tool uses continuous and discrete data, with an average of 15 items per sample variable. A B C D E F G H I J K

The variation may come from 1) within an item, 2) part-to-part, or 3) time-to-time (over-time). The Multi-vari chart examines all three of these types of variation. There are two types of Multi-vari studies—nested and crossed. The nested design looks at variation of one item within another item. A nested design is plotted over time; therefore the data adheres to the order within which it was produced. The tool can be used to examine the stability of a process if used along a time scale, and the length of each line plotted represents the range for that sample data. It exhibits three types of “within variation” for a product or process: • Positional—Variation within a part (or batch). Positional variation has three components: • Cylinder—End-to-end.

L

• Batch—Top-to-bottom, or side-to-side.

M

• Flat piece—Across width, or front-to-back.

N O P Q R S T U V W X Y Z

• Cyclical—Variation from consecutive piece-to-piece (or batch-tobatch). • Temporal—Variation occurring over time (that is, shift-to-shift or day-to-day). A crossed design manipulates the three sources of variation as independent variables. This design analyzes the two-way interaction of each of them to an output variable—the impact of the independent factor on the dependent one. The chart diagrams these multiple variables, known as input (Xs) against a single output variable (Y) to help identify which might be attributed to the largest source of variation. The independent variables typically are tested at two levels (that is, high/low or on/off). And the design set-up should be orthogonally balanced, as in a Design of Experiment (DOE). Different from the nested design, the crossed-design charts are not time-based. (See Also “Y=f(X),” and “Design of Experiment (DOE),” p. 758 and p. 250)

Multi-vari Chart

441

How to Use the Tool or Technique Using MINITAB, the procedure to conduct a nested or crossed design is the same; the data determines which test MINITAB runs. For purposes of an example, data for a crossed design will be used. Enter the data into the MINITAB Worksheet. Select the following commands from its drop-down menu: Stat > Quality Tools > Multi-Vari Chart…. Within the main screen of MINITAB’s Multi-Vari Chart, select the column containing the appropriate data for the response (output or dependent variable) and any of the independent factors of interest, shown in Figure M-32. Select the Options button and notice that MINITAB’s default includes connecting the means for as many factors identified as of interest. To see the granularity of the variation, sometimes it is helpful to see the individual data points; therefore, select Display individual data points in the Options screen, shown in M-32. Click OK in both the Options and main screens to produce the Multivari graph, shown in Figure M-33.

A B C D E F G H I J K L M N O P Q

Figure M-32: Example MINITAB Multi-vari Chart Main and Options Screens

R S

Figure M-33 plots three work shifts along the horizontal axis. The vertical axis (Y-axis) displays the yield that the shifts produce as the response. The data points represent two different locations—Plant A (indicated by an open circle) and Plant B (indicated by a circle with an embedded cross). The spread in the data is similar between plants and across shifts. There is very little difference among the shifts within a location. There is a consistent difference of about 10% in Yield between the two locations, as indicated by the slope of the three solid lines connecting the means of the two locations by shift. Based on the chart, further exploration seems warranted as to why Plant A consistently produces a higher yield than Plant B, regardless of shift.

T U V W X Y Z

442

Encyclopedia

Multi-Vari Chart for Yield2 by Location- Shift 95

Location Plant A Plant B

90

A

85

C D E

Yield2

B

80

75

F G H I J

70

65 1

2

3

Shift

K L

Figure M-33: Example MINITAB Multi-vari Chart

M N O P Q R S T U V W X

Supporting or Linked Tools Supporting tools that might provide input when developing a Multi-vari chart include • Data gathering and data collection sheets (See Also “Data Collection Matrix,” p. 248) • Performance charts and dashboards A completed Multi-vari chart provides input to tools such as • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173)

Y

• FMEA (See Also Failure Modes and Effects Analysis (FMEA), p. 287)

Z

• QFD (See Also “Quality Function Deployment (QFD),” p. 543) • Statistical analysis tools (See Also “Statistical Tools,” p. 684) Figure M-34 illustrates the link between a Multi-vari chart and its related tools and techniques.

Multi-vari Chart

443

Cause-Effect Diagram Data Gathering (metrics)

FMEA Mulit-vari Charts

Performance Charts and Dashboards

A QFD

C Statistical Analysis Tools

Figure M-34: Multi-vari Chart Tool Linkage

B D E F G H I J K L M N O P Q R S T U V W X Y Z

444

Encyclopedia

N Normal Probability Plot A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

See Also “Control Charts—7QC Tool,” in the “Normal versus Non-normal Data” section, p. 217.

Pareto Chart—7QC Tool

445

P Pareto Chart—7QC Tool A

What Question(s) Does the Tool or Technique Answer? What are the vital few items with the biggest impact? Which 20% of items produce 80% of the impact (the 80/20 rule)? A Pareto chart helps you to • Prioritize and select the biggest problem areas or largest areas of opportunities

B C D E F G

• Analyze the frequency of an event (occurrences or number of items) in a process and identify the biggest contributors

H

• Communicate a snapshot summary of how 80% of the problem comes from 20% of the causes

J

I K L

Alternative Names and Variations • This tool is also known as a Pareto diagram • Variations on the tool include Weighted Pareto Analysis

When Best to Use the Tool or Technique A Pareto chart is a tool to focus attention on priorities when trying to make decisions. It is a good communication tool describing the data in a simple and easy to read bar chart. It highlights the vital few major contributors with the largest impact.

M N O P Q R S T U V

Brief Description The Pareto chart is used to prioritize which contributors make the biggest impact on a problem, or which represent the largest areas of opportunity. These diagrams communicate the 80/20 rule, which states that 80% of an effect comes from 20% of the causes. For example, 80% of the revenue comes from 20% of the customers; or alternatively 80% of the customers contribute only 20% of the revenue. Another example, 80% of the customer complaints come from only these few causes (20%), or 80% of the repair time is spent on these few problem areas (20%). Essentially the 80/20 rule means that focusing on the vital few yields larger gains than the trivial many.

W X Y Z

446

A B C D E F G H I J K L M N O P Q R S T U V W X Y

Encyclopedia

The term “Pareto” stems from an economist, named Vilfredo Pareto (1848–1923). He was born in Paris but fled to Italy in search of political freedom. Pareto studied the distribution of wealth and devised mathematical models representing his findings that 80% of wealth was distributed among only 20% of the population. His publications supposedly were attributed to initiating Fascism in Italy. In the 1940s, Dr. Joe Juran, a renowned quality thought-leader, used a cumulative frequency chart to illustrate the concept of the vital few versus the trivial many in his Quality Control Handbook. To describe these curves, he wrote a caption underneath them stating, “Pareto’s principle of unequal distribution….” Juran applied Pareto’s economic analysis of unequal income and wealth distribution to a broader context. Interestingly, the actual concept of these cumulative curve diagrams first was attributed to M.O. Lorenz in 1904, but Juran is recognized as having popularized not only the use of these charts, but also the more universal application of the Pareto principle. The Pareto chart serves as a good executive management communication tool and is easy to construct. It is a specialized bar chart displaying the frequency and accumulation of nominal (count or attribute) data. The horizontal axis lists the categories of interest. Structurally, the Pareto plots are from left-to-right organized by height from tallest to shortest. The bars’ height indicates the frequency or count, as indicated along the vertical axis. By convention, a connecting line from the center-top of each bar depicts the cumulative percent portion of each bar’s count, adding from left-to-right until 100% total is reached. Some Pareto charts lack this connecting line. However, if a cumulative line exists, the left vertical axis represents the count, and a right-axis shows the cumulative percentage. Typical sample size used to construct a Pareto is 30 or more. The Pareto chart is a member of the 7QC Tools, (or seven Quality Control tools), attributed to Dr. Kaoru Ishikawa. The 7QC Tools sometimes are called the seven basic tools because they were the first set of tools identified as the core quality improvement tools. Ishikawa’s original 7QC toolset includes: 1) Cause-and-Effect diagram; 2) Check sheet (or checklist); 3) Control charts; 4) Histogram; 5) Pareto chart; 6) Scatter diagram; and 7) Stratification. More recently, the 7QC toolset is modified by substituting the Stratification technique with either a flowchart (or Process map) or a run chart (or Time Series plot).

Z

How to Use the Tool or Technique Pareto charts can be constructed manually or by using various software packages such as Microsoft Excel or MINITAB.

Pareto Chart—7QC Tool

447

For illustration purposes, the example scenario examines defects of painting metal components as part of an outdoor art exhibit produced in two locations: Bath and Neenah. When creating a Pareto chart, MINITAB allows data to be entered into its Worksheet one of two different ways, as shown in Figure P-1: • Detailed information—The raw data is entered by column, and each column is labeled by the category name (that is, type of defect). • Summary information—A column lists the category names (that is, type of defect) and the adjacent column of summary frequency (or count) data alongside the appropriate category. Figure P-1 displays a partial snapshot of the same set of data, formatted in the two different MINITAB Worksheet structures. Notice the asterisk (*) in the Frequency (Freq) column of the Summary Structure image of the MINITAB Worksheet. That asterisk indicates the absence of data; hence, no counts of chipping existed in the sample set.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure P-1: Example of MINITAB Worksheet Data Structure for Pareto Chart

448

Encyclopedia

Use the following procedure to develop a Pareto chart using MINITAB: Given that numeric data has been entered into the MINITAB Worksheet, select the following commands from its drop-down menu: Stat > Quality Tools > Pareto Chart…. A B C D E F G H I

(Given that the Pareto is recognized as an early quality tool, it is found under MINITAB’s Quality Tool category of its main menu, rather than under Graphing.) If the data were entered into the Worksheet using the Summary Structure, the procedure to construct a Pareto chart would involve the lower half of the Pareto Chart main screen. Select the Chart defects table and enter the appropriate column of data in the two dialog boxes—”Labels in” (indicating the column containing the category names), and “Frequencies in” (referring to the column containing the summary counts by category), as shown in Figure P-2. Keep the MINITAB default “Combine remaining defects into one category after this percent—95” as is. Click OK to generate the Pareto chart, as displayed in Figure P-3.

J K L M N O P Q R S T U V W X Y Z

Figure P-2: Example of MINITAB Pareto Chart Main Screen Using Data in a Summary Structure

Pareto Chart—7QC Tool

449

A B C D E F G H I J K Figure P-3: Example of MINITAB Pareto Chart

Combining both the locations, Bath and Neenah, shows that about 75% of the defects come from paint that is either Missing (Thin) or Peeling. Hence, if a process improvement team were evaluating this data, those would be the two defects they would address first. Look at the data by location to see if the defect prioritization is the same in the two locations. Create a stratified Pareto chart by using the data found in the Detailed Structure and select the top portion of the Pareto chart screen labeled “Chart defects data in,” as displayed in Figure P-4. Select the column containing the defects listed as attribute data and enter it in the first dialog box, which is placed along this screen portion’s title. Enter the optional variable data column containing the segmented information in the By variable in dialog box, (in this case, the column labeled “Location”). Keep the MINITAB defaults—both the highlighted for the Default (all on one graph, same ordering of bars) and the combine remaining defects into one category after this percent—95. Click OK to produce the final segmented Pareto chart, shown in Figure P-5.

L M N O P Q R S T U V W X Y Z

450

Encyclopedia

A B C D E F G H I J

Figure P-4: Example of MINITAB Pareto Chart Main Screen Using Data in a Detailed Structure

K L M N O P Q R S T U V W X Y Z Figure P-5: Example of MINITAB segmented Pareto Chart by Location

Pareto Chart—7QC Tool

451

For this example scenario, when stratified by location, the Missing (Thin) defect is the largest defect category in both locations. However, Peeling appears to be of interest at the Bath location as the second largest contributor, but not in Neenah. Neenah’s second biggest defect is Interfacial Cracking. A

Supporting or Linked Tools Supporting tools that might provide input when developing a Pareto chart include

B C D

• Data Gathering (metrics) (See Also “Data Collection Matrix,” p. 248)

E

• Performance Charts and Dashboards

G

A completed Pareto chart provides input to tools such as

F H

• Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173)

I

• FMEA (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173)

K

• QFD (See Also “Quality Function Deployment,” p. 543) • Statistical Analysis Tools (See Also “Hypothesis Testing” and “Regression Analysis,” p. 335 and p. 571, respectively.) Figure P-6 illustrates the link between a Pareto chart and its related tools and techniques. Cause-Effect Diagram

J L M N O P Q R S T

Data Gathering (metrics)

FMEA

V

Pareto Chart Performance Charts and Dashboards

U

QFD

W X Y

Statistical Analysis Tools

Figure P-6: Pareto Chart Tool Linkage

Z

452

A B C D E F G H I J K L M N

Encyclopedia

Variations Weighted Pareto Analysis A weighted Pareto is used when a Pareto chart is needed, but the categories represent unequal magnitude of importance. Hence, a conventional Pareto with each frequency weighted equally might misrepresent the prioritization. A category weighting system often reflects either unequal issues of importance (that is, health and safety) or cost concerns. Constructing a weighted Pareto chart requires that a weight of importance be assigned to each category, typically 1 to 100. It gets applied to all the counts in that category. Multiply the frequency counts by its given category weights and record the weighted-count in the data collection sheet or MINITAB Worksheet. To construct a weighted Pareto chart, use the same procedure as constructing a Pareto chart, but reference the new weighted-count data. Figure P-7 shows the preceding example of painted metal components from both Bath and Neenah, with the defect categories weighted based on repair and maintenance costs. The resulting weighted Pareto chart emphasizes a different set of vital few defects compared with the conventional non-weighted Pareto. Notice that Peeling has increased in overall magnitude of impact in the weighted Pareto, versus Missing (Thin) in the conventional approach. Also notice in Figure P-7 that the vertical axis scale of the weighted Pareto chart reflects the new weighted-counts.

O P Q R S T U V W X Y Z

Figure P-7: Example of Weighted Pareto Chart Compared with Non-weighted

PERT (Program Evaluation and Review Te chnique) Chart

453

PERT (Program Evaluation and Review Technique) Chart What Question(s) Does the Tool or Technique Answer? What is the most efficient way to complete this process or project? A PERT chart helps you to • Plan and manage the timing of activities and overall completion of a project or process. • Graphically organize process steps into the most efficient sequence. • Show the most critical path and any parallel paths with time to completion estimates.

A B C D E F G

• Evaluate and reorganize the step sequence. It identifies any simultaneous tasks and tasks that will take the longest to complete.

H

• Identify any slack time—that is the amount of time a non-critical path task can be delayed without delaying the project.

J

• Manage resources and understand upstream and downstream dependencies within the process.

I K L M N

Alternative Names and Variations • Variations on the tool include Activity Network Diagram (AND), Arrow diagram, or critical path (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127)

O P Q R

When Best to Use the Tool or Technique The Program Evaluation and Review Technique (PERT) is a project planning and decision-making tool to evaluate project timelines.

S T U V

Brief Description The Program Evaluation and Review Technique, a variation on the Activity Network Diagram (AND) technique, accounts for uncertainty in projected estimates of activities in the time management planning of programs and projects. PERT follows the same approach as the AND technique, up until the very last step in the procedure, wherein the three estimates are calculated (minimum, maximum, and most likely), to bound the range of timing possibilities. (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127)

W X Y Z

454

Encyclopedia

To construct a PERT chart, abide by five general network planning tool guidelines, which include • Activities should be numbered, and each must have a unique number. A

• The network diagram starts only with a single activity and ends with only a single activity.

B

• Before an activity begins, the preceding activity is completed.

C

• Arrows represent the sequencing of activities; however, the length of the arrow between different activities does not represent the amount of time between each, as found when one arrow is longer than another.

D E F G H I

• Two events may be connected by only one activity. Moreover, the PERT chart distinguishes itself as a network-planning tool by also following three additional requirements needed to complete it:

J

• Include all project individual activities.

K L

• Calculate three time estimates for each activity—minimum (or pessimistic), maximum (or optimistic), and most likely.

M

• Determine the critical path and slack times.

N O P Q R S T U V W X Y Z

The critical path is defined as the sequence of activities that requires the longest amount of expected time; therein determining the shortest amount of time to complete the project. Slack time (denoted as S) is the amount of time a task can be delayed without delaying a project. It is the latest date that an activity can be completed without extending the project. Hence, events on the critical path have zero slack time. In contrast, parallel paths contain slack time or a shorter duration than the critical path. The PERT method may complicate the project planning process because it requires more data to calculate the activity network. However, this extra work pays off given that the PERT chart works well for unique, non-repetitive projects where their historical data can provide (at best) estimates as to the project schedule. The PERT method helps to define interrelationships between and among activities, which can identify problem areas. The probability of achieving the project deadline may be better estimated, and the work spawns alternative planning to improve the likelihood of on-time completion. Moreover, the impact of any project changes can be quickly evaluated. And finally, it is a good communication tool to illustrate the critical path and make decisions about streamlining it or how best to execute parallel paths while conserving resources (considering the PERT chart organizes a large amount of data into a diagram format). If the project complexity and uncertainty impact the risk of

PERT (Program Evaluation and Review Te chnique) Chart

455

a significant project, consider using Monte Carlo simulation to better estimate the probabilities to complete activities and the overall project. (See Also “Monte Carlo Simulation,” p. 431)

How to Use the Tool or Technique The procedure to develop a PERT chart follows the similar steps as an Activity Network Diagram (AND). (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127) The procedural steps are as follows: Step 1.

Identify the topic, and then collect any data about the task. Examples include, Brainstorming, Process maps, Procedure manuals

Step 2.

Step 3.

List the necessary tasks.

B C D E F G H I

One technique used is to create a master list and then put one activity on either an index card or sticky note to make it easy to sequence the tasks. (Only one activity step per index card or sticky note.)

K

Sequence the tasks into a logical order.

M

Identify which tasks must precede another task and arrange them accordingly. If some tasks can occur simultaneously, then put them on the same vertical plane (parallel to one another). If the flow involves loop-back tasks, identify the determining (or evaluation) task and place in the sequence, similar to a process map’s decision diamond that poses a binary question (that is, yes/no; pass/fail). (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522) If using sticky notes, follow the same procedure and place the activities on a flip-chart sheet or section of the wall. Continue this step until all the individual steps are placed in a logical flow. Be sure to leave space in-between each step to allow additional notes to be documented. Step 4.

A

Calculate and document the three time estimates a. Pessimistic (P)—Represents the timing if everything were a disaster; all that could go wrong did (minimum time, denoted as P).

J L N O P Q R S T U V W X Y Z

456

Encyclopedia

b. Most Likely (ML)—Represents the average or most expected timeframe, which may or may not be halfway between the optimistic and pessimistic timing (denoted as ML). c. Optimistic (O)—Represents the timing if everything happened in the ideal state—smoothly, without any delays (maximum time, denoted as O).

A B

d. Document all three times on the network diagram in sequence from lowest to highest time, separated by dashes: Pessimistic-Most Likely-Optimistic (or P-ML-O).

C D E

e. Calculate the Expected time and Variance for each activity using the following formulas:

F G

i. Expected Time (TE):

H

TE = (P + 4ML + O)/6

I J

ii. Variance (V or σ2):

K L M N O P Q R S T U V W X Y Z

V = ((O–P)/6)2 Step 5.

Determine the critical path, using the Estimated Times (TE ). a. Identify all the possible path combinations from start to finish, using the estimated times (TE ) for each task or activity. b. Sum the estimated times (TE ) of each activity in each of the possible path combinations and record the path and its respective duration. c. Identify the longest path and highlight by bolding (highlighting or darkening) the path arrows—this is the critical path. d. Calculate the Earliest Start (ES) times and Earliest Finish (EF) times based on how long the preceding task(s) take. Hence, the first activity has zero ES time, and its EF time equals the total duration for the task. Preceding tasks have a compounding (or cumulative) effect on subsequent tasks’ ES and EF times. i. ES = the latest EF from the preceding task. ii. EF = the ES time–the actual task time.

PERT (Program Evaluation and Review Te chnique) Chart

457

e. Calculate the Latest Start (LS) times and Latest Finish (LF) times by starting at the end of the network (project or process completion) and working backwards toward the start activity. i. LF = the shortest LS of the subsequent task. ii. LS = the LF time–the actual task time. f. Calculate the slack times for each task and for the project. Total slack is the estimated time (TE ) an activity can be delayed without affecting the project schedule. Free slack is the time an activity can be delayed without affecting the early start of any subsequent task. i. Total slack = LS–ES = LF–EF ii. Free slack = the earliest ES of all subsequent tasks— EF Step 6.

Determine the probability (P) that the project will be finished by the deadline (TD). a. Calculate the Total Time Expected time and Total Variance for the project using the following formulas: i. Total Time Expected (TTE): TTE = Sum of the expected times on the critical path or

A B C D E F G H I J K L M N O P Q

TTE = Σ (all the TE on the Critical Path) ii. Total Variance (TV or σT ): 2

TV or σT2 =Sum of the variances (for example, σ2 = ((O—P)/6)2 ) on the critical path or TV = Σ (all the V or σ2on the Critical Path) iii. Z value for the Deadline less the Total Time Expected, divided by the standard deviation: Z = ((TD–TTE )/(Square Root of σT2 ))

R S T U V W X Y Z

458

Encyclopedia

iv. Look up the Z value on the Standard Normal Table to find the area under the curve. The table value represents the probability that the project is completed by the deadline. A B C D E F G H I

Note If the Standard Normal Table displays the area for the tail (as in the Appendix B table), and the area for the main body of the curve is what is needed, then subtract the tail area from 1. (See Also “Statistical Tools,” for a discussion on Standard Normal distribution, p. 684; and “Standard Normal Distribution,” in Appendix B, p. 975)

Step 7.

Examine and adjust the diagram as needed.

J K L M N O P Q R S T U V W X Y Z

Examples The following example depicts a PERT Chart for six activities using the Finish-Start dependency. The list of activities includes: • Activity A @ 3-day most likely duration; with 2 for pessimistic, and 5 for optimistic • Activity B @ 5-day most likely duration; with 2 for pessimistic, and 8 for optimistic • Activity C @ 7-day most likely duration; with 5 for pessimistic, and 10 for optimistic • Activity D @ 4-day most likely duration; with 3 for pessimistic, and 7 for optimistic • Activity E @ 9-day most likely duration; with 7 for pessimistic, and 12 for optimistic • Activity F @ 3-day most likely duration; with 1 for pessimistic, and 4 for optimistic The project’s deadline is 25 days. Table P-1 provides the necessary calculations of the time estimates to determine the critical path, Total Time Estimates (TTE), Total Variance (TV), Z value and the Probability.

PERT (Program Evaluation and Review Te chnique) Chart

459

Table P-1: Calculations for Six-step PERT Example Activity A

P-MLO

TE

V

TE = (P + 4ML + O) / 6

V = ((O – P) / 6)2

2-3-5

3.17

0.25

B

2-5-8

5

1

C

5-7-10

7.17

0.69

A

D

3-4-7

4.33

0.44

B

E

7-9-12

9.17

0.69

F

1-3-4

2.83

0.25

C

Path Options A-B-D-F A-B-C-E-F A-B-C-F

(3.17 + 5 + 4.33 + 2.83) =

(0.25+ 1+ 0.69 + 0.69 + 0.25) =

27.33

2.88

Total for the Critical Path For a Time Deadline @ 25 (a given)

E

15.33

(3.17 + 5 + 7.17 + 9.17 + 2.83) =

(3.17 + 5 + 7.17 + 2.83) =

18.17

TTE = 27.33

TV (σT2) = 2.88

Z = (TD-TTE) / (Sq.Rt. TV) (25 – 27.33) / (Sq. Rt. 2.88) =

Look up -1.37 in Z table; Find Z value @ 0.0853; Probability = 1

(-2.33) / 1.62 =

-1,373

D

- 0.0853) = 0.9147 or about 92% probability

F G H I J K L

Notice that Table P-1 shows a 92% probability of finishing on time for this example project to complete within the 25-day deadline. Hence, the project planning probably requires very few adjustments, if any. Table P-1 shows the example’s path calculations as • A-B-D-F = 15.33 days • A-B-C-F = 18.17 days • A-B-C-E-F = 27.33 days The longest path is A-B-C-E-F, at 27.33 days; therefore, it is the critical path. The formulas for the Earliest and Latest times are as follows, and this example’s calculated results can be found in Figure P-8: • ES = the latest EF from the preceding task. • EF = the ES time–the actual task time. • LF = the shortest LS of the subsequent task. • LS = the LF time–the actual task time.

M N O P Q R S T U V W X Y Z

460

Encyclopedia

The formulas for the slack times are as follows, and this example’s calculated results can be found in Figure P-8: • Total slack = LS–ES = LF–EF; • Free slack = the earliest ES of all subsequent tasks—EF. A

Notice, there is zero slack time along the critical path.

B

The PERT chart showing the example’s six-step scenario, its critical path and slack time calculations displayed on a network diagram can be found in Figure P-8:

C D E F G H I

TE = 3.17; Slack: 0

TE = 5; Slack: 0

TE = 4.33; Slack: 12.01

TE = 2.83; Slack: 12.01

ES: 0

EF 3.17

ES: 3.17

EF: 8.17

ES: 8.17

EF: 12.50

ES: 12.50

EF: 15.33

LS: 0

LF: 3.17

LS: 3.17

LF: 8.17

LS:20.18

LF: 24.51

LS: 24.51

LF: 27.34

Activity A

Activity B

Activity D

Activity F

2-3-5 days

2-5-8 days

3-4-7 days

1-3-4 days

J

TE = 7.17; Slack: 0

K L

TE = 9.17; Slack: 0

ES: 8.17

EF:15.34

ES:15.34

EF: 24.51

LS: 8.17

LF: 15.34

LS:15.34

LF: 24.51

M

Activity C

N

Activity E

5-7-10 days

7-9-12 days

O

Figure P-8: Six-step PERT Chart Example

P Q R S T U V W X Y Z

Hints and Tips • Building a PERT chart by hand can be tedious. For complex projects, use computer software packages to calculate the critical path and probabilities. • If the project complexity and uncertainty impact the risk of a significant project, consider using Monte Carlo simulation to better estimate the probabilities to complete activities and the overall project. (See Also “Monte Carlo Simulation,” p. 431)

PERT (Program Evaluation and Review Te chnique) Chart

461

Supporting or Linked Tools Supporting tools that might provide input when developing a PERT chart include • Activity Lists • Brainstorming sessions (See Also “Brainstorming Technique,” p. 168)

A

• Process map (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

B

A completed PERT chart provides input to tools such as • Brainstorming sessions (See Also “Brainstorming Technique,” p. 168)

C D E F

• Process map (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

G

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287)

H

• Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173)

J

• Value Stream Activity Matrix (See Also “Value Stream Analysis,” p. 727) Figure P-9 illustrates the link between a PERT chart and its related tools and techniques.

I K L M N O P

Brainstorm

Q R

Activity list

Process Map

S T

Process Map

PERT Chart

FMEA

U V W

Brainstorming

Cause-Effect Diagram

X Y

Value Stream Activity Matrix

Figure P-9: PERT Chart Tool Linkage

Z

462

Encyclopedia

Poka-Yoke What Question(s) Does the Tool or Technique Answer? How best to prevent or correct in-process errors (often human mistakes)? A B C D

Poka-yoke helps you to • Detect errors before they occur, and prevent them • Respond and adjust to in-process errors in real time

E F G

Alternative Names and Variations This tool is also known as

H

• Mistake-proofing

I

• Fail-safing

J K L M N O P

When Best to Use the Tool or Technique Ideally, the best time to implement Poka-yoke strategies is at the onset of planning and designing a new process, system, product, or service to prevent errors from occurring. However, Murphy’s Law is ever present; errors will occur at some point within a process involving humans. Hence, any time is an appropriate time to deploy Poka-yoke.

Q R S T U V W X Y Z

Brief Description Poka-yoke [pronounced “POKA-yolk-ey”] is a mistake-proofing method or device to prevent or detect an in-process error. The approach was developed by the Japanese and popularized by Shigeo Shingo in the midto late-1980s, to avoid inadvertent errors from turning into defects or defective products. A successful mistake-proofing approach is relatively inexpensive to implement and is often designed or developed by the process players. The objective strives for zero defects using one of many in-process techniques to prevent (mostly human) errors from occurring; without sampling. The technique tries to design the single method—the right way—to do a job, and make it impossible for mistakes (or deviations on the method) to occur. Poka-yoke also aims to detect errors before they become defects by using shutdown, control, or warning techniques. For example, it prevents incorrect parts from being made or assembled or easily identifies a flaw or error by providing a visual or other signal to

Poka-Yoke

463

indicate a characteristic state. Another example is in a control system, where operations halt when a defect is detected; immediately root cause analysis begins, and the problem is fixed before resuming operations. This avoids scrap from increasing. Shingo understood that human errors take on many forms and can occur at any time in a process that involves people. People include not only the process players, but also the partners and customers. The concept of concentrating on the task 100% of the time still fails to guarantee zero defects. Thus, if an error occurs, Poka-yoke minimizes the time between the error occurring and the detection warning. Poka-yoke purports five basic principles with varying degrees of effectiveness. The intent is to utilize the best most effective approach possible in a given situation. The following lists the principles in descending order of effectiveness: • Elimination—The preferred and most effective method, which removes the possibility of an error. Example: redesign the process (or product) so that the task is no longer necessary; fewer component parts • Replacement—A better method than the subsequent approaches; it substitutes a procedure with a more reliable one. Example: use robotics or automation; bar-coding prevents data entry errors • Facilitation—A slightly less effective method than those previously mentioned, but still good since it makes the work easier to perform. Example: color-coding; combining steps; checklist; and so on • Detection—A less desirable method because it failed to prevent the error from occurring, but it quickly identifies the mistake before further processing. Example: develop computer software that notifies a worker when a wrong input is made (an error message); fail-safe cut-off mechanism that shuts-down the process; flashing light or buzzer • Mitigation—The least effective, but good strategy that minimizes the effect of the error. Example: utilize electrical fuses for overloaded circuits; a computer back-up system (redundancy) Poka-yoke devices can be coupled with other inspection systems to enhance their effectiveness. However, the best mistake-proofing is

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

464

Encyclopedia

designing the quality into the process or product from the beginning, as in Design for Six Sigma (DFSS).

Porter’s 5 Forces A B C D E

What Question(s) Does the Tool or Technique Answer? How are we performing relative to the competitive business threats? Porter’s 5 Forces analysis helps you to

F

• Develop a competitive business strategy to gain or maintain an advantageous market position

G

• Communicate a high-level summary of the competitive landscape

H I J K L M

When Best to Use the Tool or Technique A planning tool used to analyze the competitive marketplace at an industry level to identify capabilities needed to enhance competitive advantage and to provide useful information to a SWOT Analysis. (See Also “SWOT (Strengths-Weaknesses-Opportunities-Threats),” p. 699)

N O P Q R S T U V W X Y Z

Brief Description The Porter’s 5 Forces analysis is a method to determine your firm’s position relative to the competitive forces active in the markets you are currently serving or intend to serve. Michael Porter, a Harvard Business School professor, developed this technique of analyzing industries and competitors for organizations to use an outside-in perspective as input to their competitive business strategy. In his book, Competitive Strategy: Techniques for Analyzing Industries and Competitors, Porter introduces this structural analysis of industries with a five competitive forces model. Figure P-10 illustrates the model and its five components. This model analyzes the attractiveness of an industry opportunity by examining its five competitive forces in aggregate, which are • Intensity of Competitive Rivalry—Determine the rivalry among the current market players. Is the competition strong among the existing organizations? Are they relatively equally balanced in size, strength, offerings, market share, or is there a dominant company? The intensity of rivalry depends on several factors including: size of market (regional, domestic, global); degree of offering differentiation (commodity versus specialty); leadership costs and market maturity.

Porter’s 5 Force s

465

POTENTIAL ENTRANTS

Threat of new entrants

A Bargaining power of suppliers

B

COMPETITIVE RIVALRY

SUPPLIERS

Intensity among existing firms

BUYERS Bargaining power of buyers

Substitute Not-In-Kind products or services

SUBSTITUTES

C D E F G H I J

Figure P-10: Porter’s 5 Forces Model

K L

• New Entrant—Ascertain the likelihood of a new entrant to start competing in the market. Do barriers exist to block entrance, or is it easy for a new entrant to offer a new, unique, or difficult (NUD) offering to meet current customer demands? (For example, the entrance of the Japanese firms, such as Toyota and Honda, into the U.S. auto market.) (See Also the “Kano Model: NUD versus ECO” section within the “KJ Analysis,” p. 376) Threats to new entrants depend on several factors including: economies of scale; capital investment requirements; technological capabilities and reliability; brand loyalty; customer switching costs (such as with information technology (IT) platform vendors); competitive response from current market players; and any government assistance and/or regulations. • Substitute—Evaluate how easily a substitute offering (product or service) can be made; often times it is an offering intended for one purpose but expanded to address a new application, and sometimes will be faster, cheaper, or better. (For example, some cell phones replace a PDA (personal digital assistant), such as the Palm Pilot, or the Hummer substituting for SUV (Sports Utility Vehicles), versus its original military purpose.)

M N O P Q R S T U V W X Y Z

466

Encyclopedia

Threats to substitutes depend on several factors including: quality (for example, reliability, durability); feature/functionality; buyer’s willingness to switch; relative value of new versus current offering (price to feature, functionality and performance, and total cost of ownership) and switching costs. A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

• Power of Suppliers—Assess the bargaining power of the suppliers relative to the sellers. Are there several suppliers to choose from or only a few? Is their offering a commodity, or do they have a unique, specialized skill and/or offering? The bargaining power of suppliers depends on several factors including: their profitability; their offering as a commodity versus specialization; number of suppliers; brand loyalty; quality and switching costs. In addition, as companies right-size, mergers and acquisitions take place or capability expansion or retrenching occurs, these trends introduce threats—for example, suppliers expanding to establish their own retail outlets and sell directly to consumers, rather than business-to-business. Conversely, buyers could threaten to integrate backward into the supply chain. • Power of Buyers—Evaluate the bargaining power of the buyers. Can they form alliances (that is, associations, such as AARP) and in aggregate exert power to make demands, including ordering volumes, availability, pricing discounts, quality, and regulations? The bargaining power of buyers depends on several factors including: the concentration of buyers; disposable income and the relative need versus want and switching costs. The model sometimes adds a sixth competitive force if appropriate—the government, or a regulatory agency. This dimension is industry-specific and may or may not be applicable on a steady-state basis.

How to Use the Tool or Technique The following procedure provides a guideline to applying the Porter’s 5 Forces model to an organizations business strategy. Step 1.

Rate each of the five competitive forces. a. Determine the list of appropriate assumptions behind each of the five competitive forces. b. By dimension, rate each assumption. Determine the rating scale to score the assumptions. Options to score the impact or magnitude include

Porter’s 5 Force s

467

i. High or Low ii. High, Medium, or Low iii. 9 (High), 3 (Medium), 1 (Low) Figure P-11 illustrates an example of rating of the forces’ assumptions.

A B

Threat of a New Entrant The Threat if a New Entrant is High When:

High

Economies of scale

Capital requirements Switching costs

Incumbent’s access to raw materials

X X X X

Incumbent’s access to gov’t subsidies

X

Incumbent’s control of distr. channels Incumbent’s proprietary knowledge

D E F

Intensity of Competitive Rivalry

Power of Suppliers High

Concentration relative to the buyer industry

C

X X X

Product differentiation

The Power of Suppliers is High When:

Low

Low

X X

Importance of customer to the supplier

The Threat if a New Entrant is High When: Number of competitors

X

Availability of substitute products

Power of Buyers High

Low

Concentration of buyers relative to suppliers

X

Industry growth rate

The Power of Buyers is High When:

X

Volume of purchase

Fixed costs

X

Product differentiation of suppliers

Storage costs

X

Threat of backward migration by buyers

High

Low

X X X

Differentiation of the supplier’s products and services

X

Switching costs of the buyer

X

Switching costs

Threat of forward integration by the supplier

X

Exit barriers

X

Cost savings from the supplier’s product

X X

Strategic stakes

X

Importance of the supplier’s input to quality of buyer’s final product

X

Product differentiation

X X

High

Rate of improvement in the priceperformance relationship of a substitute product

X

Profitability of the industry producing a substitute

X

Switching costs for the buyer of a product

Extent of buyer’s profit

Percentage of total buyer’s cost spent on the supplier’s input

Threat of a Substitute The Threat of a Substitute Product is High When:

Buyer’s knowledge about supplier’s cost structure

X X

X

Low

G H I J K L M

X

Figure P-11: Example of Rating Porter’s 5 Forces’ Assumptions

N O P

c. Evaluate the rating for the overall competitive force, using the same scale selected in Step 1.b. Step 2.

Summarize the overall competitive force ratings in a spider chart. a. Using Microsoft Excel’s spider chart format, enter the summary ratings for each of the five competitive forces. Figure P-12 illustrates an example of the overall rating of the five forces’ and a spider chart template.

Q R S T U V W X Y Z

468

Encyclopedia Overall Assessment Competitive Forces

Name of Market or Segment Intensity of Competive Rivalry

Industry

5

Intensity of Competitive Rivalry

Low, Medium, or High

Threat of a New Entrant

Low, Medium, or High

Threat of Substitute Products

3 Power of Suppliers

Low, Medium, or High

2

Threat of a New Entrant

1 0

A

Power of Buyers

Low, Medium, or High

B

Power of Suppliers

Low, Medium, or High

C

Overall Competitive Assessment

Low, Medium, or High

D

4

Power of Buyers

Threat of a Substitute Product

Figure P-12: A Second Example of Rating Porter’s 5 Forces’ Assumptions

E F G H I J K L

How to Analyze and Apply the Tool’s Output The stronger the competitive forces, the more evidence there is to support one or more of the following marketplace characteristics: • Strong rivalry among the sellers • Low barriers to entry

M

• Strong competition by substitutes

N

• Power held by the suppliers

O

• Power held by the buyers

P Q R S T U V W X Y Z

As a result, the stronger the competitive forces, the higher the score. The large scores indicate less profitable market opportunities and purportedly a less attractive industry. Conversely, the smaller the score, the less competitive forces exist. Figure P-13 shows a completed Porter’s 5 Forces diagram as a spider chart for the cell phone industry, indicating a strong threat with the entrance of MP3 players from companies such as Apple’s iPhone and the Asia-Pacific manufacturers. The MP3 devices may serve as one of the cell phone’s threats as a substituted product.

Porter’s 5 Force s

469

Cell Phones Intensity of Competive Rivalry

5 4 3 Power of Suppliers

2

Threat of a New Entrant

1

A B

0

C D Power of Buyers

Threat of a Substitute Product

Key: Smaller value is better

Figure P-13: Example Porter’s 5 Forces Illustrated as a Spider Diagram

E F G H I

Supporting or Linked Tools Supporting tools that might provide input when developing a Porter’s 5 Forces Analysis include • Market and Segment characteristic data • Buyer, supplier, and competitive intelligence data A completed Porter’s 5 Forces Analysis provides input to tools such as

J K L M N O P Q

• Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173)

R

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287)

T

• Prioritization Matrix (See Also “Prioritization Matrices—7M Tool,” p. 470) • SWOT Analysis (See Also “SWOT (Strengths-WeaknessesOpportunities-Threats),” p. 699) Figure P-14 illustrates the link between a Porter’s 5 Forces Analysis and its related tools and techniques.

S U V W X Y Z

470

Encyclopedia Cause-Effect Diagram Market and segment characteristics data

A B C

FMEA Porter’s 5 Forces

Competitive intelligence

Prioritization matrix

D SWOT

E F G

Figure P-14: Porter’s 5 Forces Analysis Tool Linkage

H I

Prioritization Matrices—7M Tool

J K L M N O P

What Question(s) Does the Tool or Technique Answer? What is the best option among the several possibilities for a crucial (often mission-critical) decision that carries risk of significant consequences if wrong? • Prioritization Matrices help you to Narrow down options for a weighty decision to select the best one using several subjective criteria

Q R S T U V W X Y Z

Alternative Names and Variations This tool is also known as • Analytical Criteria method or Full Analytical Criteria method • Consensus Criteria method • Combination I.D./Matrix method, where I.D. refers to Interrelationship Diagraph (See Also “Interrelationship Diagram—7M Tool,” p. 369) Variations on the tool include • Decision matrix (See Also Matrix Diagrams—7M Tool,” p. 399) • Pugh Concept Evaluation (See Also “Pugh Concept Evaluation,” p. 534) • Solution Selection Matrix (See Also “Solution Selection Matrix,” p. 672)

Prioritization Matrice s—7M Tool

When Best to Use the Tool or Technique When the decision is of high importance to the organization, the stakes are high, and grave consequences threaten the business if wrong. The set of Prioritization Matrices are best used to narrow a list of available options, by comparing the options using three or more essential (often subjective) criteria to select the best one.

471

A B

Brief Description Prioritization Matrices represent a set of three different matrices. These set of tools were introduced early in the history of quality and Six Sigma to balance the heavy quantitative emphasis of the matrix data analysis introduced by the Japanese with some qualitative-oriented tools. This set of tools is designed for crucial organizational decisions that carry significant consequences if wrong. Depending on the size, scale, and scope of the organization, such weighty decisions may include selecting an acquisition candidate, making key personnel decisions, or choosing a new enterprise-wide IT platform. These matrices can highlight any disagreement among a project team and help to focus them on the best alternative by building consensus around the unbiased decision-criteria. Typically, these matrices use an L-shaped structure to compare a set of options with a set of decision-criteria. The Prioritization Matrices entail a systematic approach to select criteria, determine criteria weights, and compare choices. Of all the matrix diagram tools, the Prioritization Matrices are considered the most rigorous and time-consuming to use. (See Also “Matrix Diagrams—7M Tool,” p. 399, for a discussion on other matrices and their shapes.) Although the names are similar, the Prioritization Matrices distinguish themselves from two others—the Decision Matrix and Decision Authority Matrix. A Decision Matrix may refer to a specific tool or a category of matrices used to evaluate and prioritize a list of options, often called by a specialized name—Pugh Matrix and Solution Selection Matrix. A Decision Authority Matrix identifies the roles or people accountable for making decisions in a process or organization. Prioritization Matrices are a member of the 7M Tools, attributed in part to Dr. Shewhart, as seven “management” tools, or sometimes referred to as the 7MP or seven management and planning tools. These 7M Tools make up the set of traditional quality tools used to analyze quantitative data. The 7M Toolset includes: 1) Activity network diagrams or Arrow diagrams; 2) Affinity diagrams; 3) Interrelationship digraphs or Relations diagrams; 4) Matrix diagrams; 5) Prioritization Matrices, often replacing the more complex matrix data analysis; 6) Process decision program charts (PDPC); and 7) Tree diagrams. The book, The Quality Toolbox by Nancy Tague, presents the 7M Tools ranked from those used for

C D E F G H I J K L M N O P Q R S T U V W X Y Z

472

Encyclopedia

abstract analysis to detailed planning: Affinity diagram, Relations diagram, Tree diagram, Matrix diagram, Matrix data analysis (commonly replaced by a more simple Prioritization Matrix), Arrow diagram, and Process Decision Program Chart (PDPC). The three types of Prioritization Matrices include A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

• Full Analytical Criteria method—The most complex of the three. • Consensus Criteria method—Moderate complexity. • Combination I.D./Matrix method—Simplest of the three.

Full Analytical Criteria Method The Prioritization Matrix produced from the Full Analytical Criteria method should be reserved for those most significant or critical issues. It requires multiple matrices to conduct a pair-wise comparison before completing the final Prioritization Matrix. Though it is the most complex of the three Prioritization Matrices, it is a simplified derivative of the Analytical Hierarchy Process (AHP). The method’s guidelines recommend that a small team of three to eight people share in building this matrix. Otherwise, working through the process and accounting for a large number of perspectives and input can become unwieldy. When the team remains small, reaching a complete consensus for each of the multiple matrices becomes more facile. A Full Analytical Criteria method prioritizes the decision criteria, weights them, and applies numerical values to the options to indicate the best alternative. A limited number of options and criteria are recommended (about ten or less of each) since the approach is time-consuming and the calculations can become cumbersome.

Consensus Criteria Method The Consensus Criteria method builds a Prioritization Matrix for scenarios where the options appear virtually equal. This Prioritization Matrix essentially is a simplified version of the Full Analytical Criteria method, given that it skips the multiple pair-wise comparisons. Therefore, it takes less time to create. Similar to the previously mentioned method, it requires multiple matrices to complete the final matrix. The Consensus Criteria method uses weighted voting and ranking and applies numerical values to the options to indicate the best alternative. A limited number of options and criteria are recommended (about ten or less of each) to minimize the time to produce the final output. Combination I.D./Matrix Method The Combination ID/Matrix method creates a Prioritization Matrix for a complex issue that involves a cause-and-effect relationship to determine where to start tackling the problem. Thus, the method focuses on a

Prioritization Matrice s—7M Tool

473

causal-based approach rather than criteria-based. It prioritizes and selects the root cause to address first. The team assigned to resolve this issue should have a first-hand working knowledge of the process. This combination method employs a hybrid structure made up of a matrix chart and a Tree diagram. It uses the cause-and-effect relationship information from an Interrelationship diagram (ID) or Tree, rather than decision-criteria. The Tree diagram parses the ideas into further detail and then combines the detail with the criteria in an L-shaped matrix. Within an L-shaped matrix, the root cause options are compared to each other to determine their relative strength. It is considered the simplest of the three Prioritization Matrices because it uses non-numerical values (or symbols and arrows) to identify the best option. (See Also “Interrelationship Diagram—7M Tool,” p. 369)

A B C D E F G

How to Use the Tool or Technique The imperative decisions facing an organization are best tackled by a team of the key individuals. Often such a crucial decision elicits varying opinions, all worth consideration. Usually the decision-criteria include qualitative ones that represent different impacts for the different business perspectives. Building a Prioritization Matrix brings this team of key individuals together to gain consensus on the selection criteria and the optimal alternative.

H I J K L M N

The procedure to develop each of the three Prioritization Matrices is as follows.

O

Full Analytical Criteria Method The procedure using the Full Analytical Criteria method involves a small team of the key individuals (ideally three to eight people) to perform the following tasks:

Q

Step 1.

Agree on the goal statement.

Step 2.

Gain consensus on the final list of decision-criteria.

Step 3.

Create the decision-criteria matrix to weigh each of them against one another. a. Using an L-shaped matrix structure, starting in the second row of the first column, record the final list of decision-criteria by placing one criterion per cell, to serve as the row heading. b. Starting in the first row, second column, and using the same sequence of decision-criteria as in Step 3.a., record

P R S T U V W X Y Z

474

Encyclopedia

each of them from left to right, placing one topic per cell, to create the column headings. c. Starting in the upper-left corner of the matrix, shade each cell along the diagonal to the lower-right corner because these cells compare the same decision-criteria to itself, which is unnecessary.

A B

d. Add two additional columns on the far right and title them with the following headings—”Row Total” in the second column from the right and “Relative Value Rating” in the far right column.

C D E

e. Add one additional row at the bottom of the matrix and title it “Grand Total.”

F G

Figure P-15 provides an illustration of a completed sample decision-criteria matrix.

H I J K L

Step 4.

Rate the decision-criteria against one another, using the standard rating scale found in Table P-2:

Table P-2: Decision-Criteria Rating Scale for Full Analytical Criteria Method Prioritization Matrix

M

10

Row Criteria is MUCH MORE IMPORTANT than the Column Criteria

N

5

Row Criteria is MORE IMPORTANT than the Column Criteria

1

Row Criteria is EQUALLY IMPORTANT to the Column Criteria

1/5

Row Criteria is LESS IMPORTANT than the Column Criteria

1 / 10

Row Criteria is MUCH LESS IMPORTANT than the Column Criteria

O P Q R S T U V W X Y Z

a. Always starting with the decision-criteria listed along the vertical axis of the matrix—the Row Criteria—rate the row criterion versus the column criterion within each unshaded cell, using the rating scale found in Table P-2.

Prioritization Matrice s—7M Tool

475

b. Continue until each un-shaded cell contains a rating. Step 5.

Complete the decision-criteria matrix. a. Sum the row ratings to compute the Row Total for each row criterion and record the score in the second from the right column, Row Total. Sum the row totals of this column to compute the Grand Row Total. Record that summed ratings score in the last row’s cell of that column, of the row, Grand Total. b. Compute the Relative Value for each Row Criteria by dividing each Row Total score by the Grand Row Total. This calculation determines what percentage a given Row Criteria is of the Grand Total. Each computation should be carried to two decimal positions. Record the calculated relative weighing of each Row Criteria in the adjacent cell of the far right column, Relative Value Rating. Figure P-15 displays a completed decision-criteria matrix for an acquisition example. c. If a decision-criteria’s Relative Value is extremely small compared with the others on the list, the team may decide to remove it from the list. Now the decision-criteria list with its Relative Value scores is ready to use to evaluate the options under consideration.

A B C D E F G H I J K L M N O P Q R

Criteria Customer base

Criteria Customer base Increased Capacity Technological Advantage Grand Total

Increased Capacity

5 1/5

Technological Advantage

10 5

(= 0.2)

S

Row Total

Relative Value Rating

15.00

0.73

( = 5 + 10)

(= 15 / 20.50)

U

T V

5.20

0.25

( = 5 + 0.2)

(= 5.2 / 20.50)

W

0.02

X

1/10

1/5

0.30

(= 0.1)

(= 0.2)

( = 0.5 + 0.1)

20.50 (= 15+5.2+0.3)

(= 0.3 / 20.50)

100%

Figure P-15: Sample Decision-criteria Matrix for the Full Analytical Criteria Method Prioritization Matrix

Y Z

476

Encyclopedia

Note The shaded cells along the upper-left-to-lower-right diagonal create a mirror-image of cells on either side. Hence, if the rating for Row CriA B C D

terion 1 versus Column Criterion 2 were 5 and placed above the shaded-diagonal; then the mirror-image cell found below the shaded-diagonal formed by Row Criterion 2 versus Column Criterion 1 would have an inverse rating of 1/5.

E F G H I J K L M N O P Q R S T U V W X Y Z

Step 6.

Create the set of options rating matrices to weigh each of them against one another using one decision-criterion and its relative values. a. Create additional L-shaped matrices that total the number of final decision-criteria. Hence, if there are three final decision-criteria, create four additional matrices. Assign each decision-criterion to a matrix and title it as such, one for each decision-criterion. Place each title in the upper-left cell of each matrix. b. Build each matrix structure similar to the procedure described in Step 3. Start in the second row of the first column and record the list of options by placing one option per cell, to serve as the row heading. c. Starting in the first row, second column, and using the same sequence of options as in Step 6.b., record each of them from left to right, placing one topic per cell, to create the column headings. d. Starting in the upper-left corner of the matrix, shade each cell along the diagonal to the lower-right corner because comparing the same option to itself is unnecessary. e. Add two additional columns on the far right and title them “Row Total” in the second column from the right and “Relative Value Rating” in the far right column.

Prioritization Matrice s—7M Tool

477

f. Add one additional row at the bottom of the matrix and title the row “Grand Total.” Step 7.

Complete each option rating matrix, by rating the options within each option-rating matrix, using the standard rating scale found in Table P-2. a. Always start with the options listed along the vertical axis of the matrix—the Row Options. Using the pair-wise approach, rate the row option versus the column option within each unshaded cell, using only one criterion and the standard rating scale found in Table P-2.

A B C D E F

Note

G

The shaded cells along the upper-left-to-lower-right diagonal create

H

a mirror-image of cells on either side, as explained in Step 4.a.

I J

b. Continue until each unshaded cell contains a rating. c. Sum the row ratings to compute the Row Total for each row option and record the score in the second from the right column, Row Total. Sum the row totals of this column to compute the Grand Row Total. Record that summed ratings score in the last row’s cell of that column, of the row, Grand Total. d. Compute the Relative Value for each Row Option by dividing each Row Total score by the Grand Row Total. This calculation determines what percentage a given Row Option is of the Grand Total. Each computation should be carried to two decimal positions. Record the calculated relative weighing of each Row Option in the adjacent cell of the far right column, Relative Value Rating. e. Continue this sequence of tasks until all the criterionspecific options rating matrices are completed. Hence, the number of completed option-rating matrices equals the number of decision-criteria, as shown in Figure P-16, which continues with the company acquisition scenario illustrated in the previous figure.

K L M N O P Q R S T U V W X Y Z

478

Encyclopedia Customerbase Criteria Company A Company B

1/5

Company C

1

B

Increased Capacity Criteria

C

Company A

D E

Row Total

Relative Value Rating

5

1

6.00

0.48

5

5.20

0.42

1/5

Company B

1/10

Company C

1

Technological Advantage Criteria

G

Company A

H I J

Company B

10

Company C

5

Grand Total

0.10

12.40

100%

Company C

Row Total

Relative Value Rating

10

1

11.00

0.63

1/5

0.30

0.02

5

6.00

0.35

17.30

100%

Company B

Company C

Row Total

Relative Value Rating

1/10

1/5

0.30

0.01

10

20.0

0.79

Options Company A

1.20

Company B

Options Company A

Grand Total

F

L

Company C

Grand Total

A

K

Company B

Options Company A

1/10

5.10

0.20

25.40

100%

Figure P-16: Sample Set of Option Rating Matrices for the Full Analytical Criteria Method Prioritization Matrix

M N O

Note

P

At this point only the Summary Prioritization Matrix will be

Q

incomplete.

R S T U V W X Y Z

Step 8.

Complete the Summary Prioritization Matrix that compares the weighted options to the weighted decision-criteria to identify the optimal alternative. a. Create the last L-shaped matrix and title it “Summary Prioritization Matrix.” b. Add two additional columns on the far right and title them “Row Total” in the second column from the right, and “Relative Value Rating” in the far right column. c. Add one additional row at the bottom of the matrix and title it “Grand Total.” d. In the first row, second column, record each decision-criterion from left to right, placing one topic per cell, to create

Prioritization Matrice s—7M Tool

479

the column headings. In parentheses under the column heading, record each decision-criterion relative value within each appropriate cell. (Recall these values come from the decision-criteria matrix.) e. Create the row headings by starting in the second row of the first column; record the list of options by placing one option per cell, to serve as the row heading. f. Gather the option rating matrices. Using one matrix at a time, record the appropriate option’s relative value in a given column that corresponds with the decision-criteria used to create that specific option-rating matrix. Continue until all the columns have been filled with the appropriate option’s relative value. g. By cell, calculate and record the priority rating by multiplying the appropriate decision-criteria’s relative value times the option’s relative value.

A B C D E F G H I J

You may want to circle or bold this number to distinguish it from the option’s relative value in the cell.

K

Continue until each cell contains a priority rating.

M

h. Sum the row priority ratings to compute the Row Total for each row prioritized option and record the score in the second from the right column, Row Total.

L N O P

Sum the row totals of this column to compute the Grand Row Total. Record that summed ratings score in the last row’s cell of that column, of the Grand Total row.

Q

i. Compute the Relative Value for each Row Prioritized Option by dividing each Row Total score by the Grand Row Total. This calculation determines what percentage a given Row Prioritized Option is of the Grand Total. Each computation should be carried to two decimal positions. Record the calculated relative weighing of each Row Prioritized Option in the adjacent cell of the far right column, Relative Value Rating.

T

The largest Prioritized Option Relative Value indicates the best alternative, as shown in Figure P-17, which continues with the company acquisition scenario illustrated in the previous two figures.

R S U V W X Y Z

480

Encyclopedia

If there are two or more of the top prioritized options that have ratings relatively close together, conduct further exploration.

A

Figure P-17 displays the final Prioritization Matrix using the Full Analytical Criteria method. For this scenario, Company A appears to be the best option of the three, with a total score of 0.51.

B C

Summary Prioritization Customer base (CB) Matrix

D

(0.73)

E F G

0.63 x IC = 0.16

0.01 x TA = 0.0002

0.51

0.51

Company B

0.42x CB = 0.31

0.02 x IC = 0.01

0.79 x TA = 0.02

0.34

0.34

Company C

0.10x CB = 0.07

0.35 x IC = 0.08

0.20 x TA = 0.004

0.15

0.15

0.02

1.00

1.0

Grand Total

K M

Row Total

Relative Value Rating

0.48 x CB = 0.35

J L

Technological Advantage (TA) (0.02)

Company A

H I

Criteria Increased Capacity (IC) (0.25)

0.73

0.25

Figure P-17: Sample Option Rating Matrices for the Full Analytical Criteria Method Prioritization Matrix

N O Q

Consensus Criteria Method The procedure using the Consensus Criteria method involves the team of key individuals to perform the following tasks:

R

Step 1.

Agree on the goal statement.

Step 2.

Gain consensus on the list of decision-criteria.

Step 3.

Create the decision-criteria weightings.

P

S T U V

a. Stack rank or sequence each of the criteria from most import to least important.

W X

b. Given that the total weightings must equal 100%, assign a percentage weight to each of the decision-criteria.

Y Z

Step 4.

Create the Consensus Criteria matrix to weigh each option against the weighted decision-criteria. a. Using an L-shaped matrix structure, in the first row, second column, record each of the prioritized decision-criteria from left to right, starting with the most important, and placing one criterion per cell, to create the column headings.

Prioritization Matrice s—7M Tool

481

Underneath each decision-criteria column heading, record the appropriate weighting (as a decimal) for each. Recall, these weightings should sum to 100%. b. In the second row of the first column, record the options by placing one topic per cell, to serve as the row heading. c. Add one additional column on the far right and name the column “Row Total.” Step 5.

Complete the Consensus Criteria matrix as the final Prioritization Matrix. a. Count the number of options under consideration; that number becomes the highest possible rank for the options. Create a ranking scale: i. 1 = worst (or lowest rank), to ii. “X” number of options under consideration = best

(or highest rank). b. Within a column, stack rank and record each option against that column’s decision-criteria, from best to worst. Continue this task until all the decision-criteria columns have been filled with the appropriate stack rankings. c. By cell, calculate and record the prioritized rankings by multiplying the appropriate decision-criteria’s weighting times the option’s stacked rank value.

A B C D E F G H I J K L M N O P Q

You may want to circle or bold this number to distinguish it from the option’s stacked rank value in the cell.

R

Continue until each cell contains a priority rating.

T

d. Sum the row prioritized rankings to compute the Row Total for each row option and record the score in the far right column, Row Total. Each computation should be carried to two decimal positions. e. The largest Prioritized Option score indicates the best alternative, as shown in Figure P-18. If there are two or more of the top prioritized options have ratings relatively close together, conduct further exploration.

Figure P-18 shows the final Prioritization Matrix using the Consensus Criteria method. For this scenario, Company B appears to be the best

S U V W X Y Z

482

Encyclopedia

option of the three, with a total score of 2.6; however, that score is close to that of Company C. Thus, further exploration should be conducted to ensure which option appears to be the clear winner. Consensus Criteria Prioritization Matrix

A B C

Company A

D Company B

E F

Company C

G H

Criteria Customer base (CB) (0.7)

1 x CB =

0.7 3 x CB =

2.1 2 x CB =

1.4

Increased Capacity (IC) (0.2)

Technological Advantage (TA) (0.1)

Row Total

1 x IC = 0.2

2 x TA = 0.2

1.1

2 x IC = 0.4

1 x TA = 0.1

2.6

3 x IC = 0.6

3 x TA = 0.3

2.3

Figure P-18: Sample Prioritization Matrix Using the Consensus Criteria Method

I J K L M N O

Combination I.D./Matrix Method The Prioritization Matrix developed from using the Combination I.D./Matrix method involves the team of key process players to perform the following tasks: Step 1.

Agree on the problem statement.

Step 2.

Generate list of root cause options.

P

a. Gather information and data about the issues its causeand-effect relationships. Potential input sources include Interrelationship diagram, Fishbone (or Cause-and-Effect diagram), Brainstorming using the 5Ms and P technique and/or 5-Whys, graphical methods, Control charts, and a Process map. (See Also “Interrelationship Diagram—7M Tool,” p. 305; “Cause-and-Effect Diagram—7QC Tool,” p. 173; “Brainstorming Technique,” p. 168; “5-Whys,” p. 305, “Graphical Methods,” p. 323, “Control Charts— 7QC Tool,” p. 217; “Process Map (or Flowchart)—7QC Tool,” p. 522)

Q R S T U V W X

b. Summarize and document the cause-and-effect relationships in a Tree diagram.

Y Z

c. Generate a list of the root causes options from the most detailed level of the Tree diagram. Step 3.

Create the Combination ID/Matrix to compare each root cause option against one another, using the cause-and-effect relationship as the evaluation standard.

Prioritization Matrice s—7M Tool

483

a. Using an L-shaped matrix structure, in the first row, second column, record each of the options from left to right, placing one per cell, to create the column headings. b. In the second row of the first column, again record the options going down the column, in the same sequence used for the column headings, by placing one topic per cell, to create the row heading. c. Starting in the upper-left corner of the matrix, shade each cell along the diagonal to the lower-right corner because comparing the same option to itself is unnecessary. d. Add four additional columns on the far right and name the columns in sequence from left to right as follows:

Step 4.

A B C D E F G

i. Total In—in the column fourth from the right

H

ii. Total Out—in the column third from the right

I J

iii. Total In and Out—in the column second from the right

K

iv. Strength—in the far right column

L

Conduct vertical versus horizontal pair-wise comparisons on the root cause options to prioritize and select which to address first. a. Start with the first option listed on the vertical axis (the Yaxis) in row 3, compare it to the second option listed on the horizontal axis (the X-axis), and find where they intersect in a cell—in the third row, second column. b. Evaluate the strength of the cause-and-effect relationship between the two options by asking, “Does the row option effect or influence the column option? If so, what is the strength of the relationship—strong, moderate, or weak-to-none?” Select one of the top three geometric symbols (solid circle, box, or triangle) in following standard rating scale found in Table P-3 to indicate the response. If no relationship exists, leave the cell blank. Record the geometric symbol in the appropriate cell of the Combination ID/Matrix. Continue this task until all the paired option cells have been filled with the appropriate rating symbols.

M N O P Q R S T U V W X Y Z

484

Encyclopedia

Relationship Strength

Table P-3: Cause-and-Effect Rating Scale of Symbols for a Prioritization Matrix Using the Combination ID/Matrix Method

A B

Cause Source

C D E

STRONG cause-effect relationship (9 points) MODERATE cause-effect relationship (3 points) WEAK cause-effect relationship (1 point) Up (OUT) Arrow indicates Driving Cause of the other option Side (IN) Arrow indicates Effect

F G H

Note

I

Recall: Blank cells indicate no relationship. Continue this task until all the

J

paired option cells have been filled iwith the appropriate rating symbols.

K L

c. Indicate which option in a pair-wise comparison drives the cause, by using the arrow symbols (up or side) from the standard rating scale found in Table P-3 to indicate the response. Record the arrow symbol next to the geometric symbol in the appropriate cell of the Combination ID/Matrix.

M N O P Q R S T U V W X Y Z

Step 5.

Complete the Combination ID/Matrix as the final Prioritization Matrix. a. Identify the root cause option associated as driving the most causal relationships—the biggest Cause Source. i. By row, count the number of side (IN) arrows pointing left in that row, and record the total in that row’s cell within the Total In column. ii. By row, count the number of up (OUT) arrows in that row and record the total in that row’s cell within the Total Out column. iii. By row, add the “Total In” and “Total Out” numbers together and record the sum in the row’s cell within the Total In and Out column. iv. Continue until each row is completed. b. Identify the root cause option associated with the strongest cause-and-effect relationship.

Prioritization Matrice s—7M Tool

485

i. Refer to the point values associated with each causeand-effect relationship strength rating found in Table P-3. ii. By row, add up the point values associated with each geometric strength relationship symbol found in that row and record the total in that row’s cell within the Strength column. c. The top root cause option candidates are identified as those with the highest strength score and highest number of total in and out arrows, as show in example Prioritization Matrix of Figure P-19. i. Highest Strength indicates a strong relationship with the other options. ii. Highest Total In and Out indicates many connections, with the Highest Out characterizing a strong root cause relationship and should be addressed first. The Highest In indicates that these options may serve as good leading indicators. If there are two or more of the top prioritized root cause options that have ratings relatively close together, conduct further exploration. Figure P-19 shows the final Prioritization Matrix using the Combination ID/Matrix method. For this scenario, an organization wanted to improve employee morale but did not know where to start. The resulting Prioritization Matrix using this combined approach indicates that apparent lack of interest in exercising empowerment and workload issues probably are the two best places to start to address the employee morale problem. Perhaps delving deeper into both of the top candidate topics may uncover a more clear starting point. Further investigation techniques may include using the 5-Whys or interviewing the employees and conducting a KJ analysis on the results. (See Also “5-Whys,” p. 305 and “KJ Analysis,” p. 375)

A B C D E F G H I J K L M N O P Q R S T U V W

Root Cause Options

X

Total IN

Total OUT

Total IN & OUT

Strength

High absenteeism

1

1

2

6

Unexercised empowerment

1

1

2

12

Overworked from down-sizing

0

2

2

12

Combination ID/ Matrix method

High absenteeism

Unexercised empowerment

Overworked

Figure P-19: Sample Prioritization Matrix Using the Combination ID/Matrix Method

Y Z

486

Encyclopedia

Process Capability Analysis What Question(s) Does the Tool or Technique Answer? Is the process able to meet customer requirements? A

Process Capability analysis helps you to

B C D E F

• Understand if an existing, steady state, normal process, is capable of producing output within specified customer specification limits • Understand whether the actual process results are acceptable with respect to the customer specifications

G

• Evaluate the performance of a process

H

• Compare the performances of two different processes; to select one

I J K L M N O

Alternative Names and Variations This tool is also known as • Capability Analysis or Capability Study • Process Capabilility Study

P Q R S T U V W X Y Z

When Best to Use the Tool or Technique Process Capability analysis can be preformed only on processes known to be in “statistical control,” meaning that the exhibited process variation is random and steady over time. The term random addresses that something occurs as a common event nature, and not a purposeful, special cause event. The term steady over time refers to the ability to predict the event within the process’ inherent fluctuation. The process data also must be normally distributed. If a process proves incapable of producing customer requirements, then improvements may be in order, if the business can afford the investment and time. Consider the alternatives—redesigning the current process or stopping to use the process all together and designing a new “clean sheet” process. If the process is capable, however, the number of defects or amount of variation miss the target, then process improvements may be in order and probably requires less investment than a process that is incapable. (See Also “Control Charts—7QC Tool,” p. 217)

Proce ss Capability Analysis

Brief Description The concept of process capability describes a predictable pattern of statistically stable behavior wherein the process variation occurring by chance alone is compared to a set of specifications. Ideally, those specifications (upper and lower specification limits) are set by the customer, but often times they are set by the technical engineering design. A capable process is one whose distribution is narrow enough to fit within the specification range. Process Capability analysis judges whether the process variation fits within the upper and lower specification limits. If the analysis determines the process incapable, the process requires improvements to reduce the variation (as in a Lean Six Sigma project)—variation is the enemy. If after improvement initiatives, the process remains incapable, consider either a more aggressive process redesign or clean-sheet design approach (as in DFSS). As a last resort, consider examining the process specifications settings. Perhaps they are out of alignment with both the customer and process requirements, therefore requiring an adjustment.

487

A B C D E F G H I

There are various statistical metrics used to describe process capability, including “Number of Defects, Process Capability Indexes (CPIs), and Yield.

K

Number of Defects

L

• Defects per Unit (DPU), which is the total number of defects divided by total number of units. Average number of defects found in a sample.

J

M N O P Q

Note

R

Defects are defined as a number of things wrong within a product,

S

though, the product is still considered good or acceptable.

T U

Defective is defined as a bad part or product, one that is not fit for use and must be either scrapped or reworked. • There are two ways to make a “Bad Part”: • Process has shifted off target. The center of the process can be measured three ways: mean, median, and mode. • Process variation is too wide and extends beyond the customer specification limits. The process spread is measured by the Range, Variance, and Standard Deviation.

V W X Y Z

488

Encyclopedia

• Defects Per Million Opportunity (DPMO): • A common simple process capability metric. • The calculation can be used with either attribute or continuous data and is described as the “Average number of defects per unit divided a million opportunities within which to make a defect.” DPMO = DPU x 1,000,000.

A B

• It can also be expressed as DPMO = [(# of Defects) divided by [(# of Units) times (# of Opportunities)] times 1,000,000].

C D E F G

Table P-4 summarizes the relationship among DPMO, Yield, and the Sigma-level. Table P-4: Translating DPMO into Yield and Sigma Level

H

DPMO

%Yield

Sigma Level

I

308,538

69.1

2

66,807

93.32

3

L

6,210

99.379

4

M

233

99.9767

5

3.4

99.99966

6

J K

N O P Q R S T U V W X Y Z

Process Capability Indexes (CPIs) Process Capability Indexes (CPIs) statistically quantify the ability of a process to produce output that meets customer requirements by calculating variability differently, depending on the duration of the data (shortor long-term). While there are several CPIs available, the most commonly used indexes are Cp and Cpk. Both of these indexes view process capability through a “short-term” lens. Short-term analysis requires less data to gather by estimating the standard deviation and mean and predicting it onto the population. Over the long-term, a process becomes “less capable” because of changes in its mean over time. Motorola discovered and established that with time, a process tends to drift about 1.5 sigma, as shown in Figure P-20. However, some current industry experts argue against the 1.5 sigma shift as a cover-up and have written several articles on the topic. The other indexes are defined as follows, but the focus of this section is on the calculation and interpretation of Cp and Cpk. The condition where both Cp > 2.0 and Cpk > 1.5 exist defines a Six Sigma process.

Proce ss Capability Analysis

489

Time Series Plot

13

12

Long-term variability

Short-term variability

A

11

B

10

C D

9

E

8

F G

7 1

10

20

30

40

50

60

70

80

90

100

Time

Figure P-20: Short-term Versus Long-term Variability

H I J K L

• Cp: Specification Spread (USL—LSL) divided by Process Spread (6s); or [(USL—LSL)/(6s)]; known as “short-term” process capability, based on the dispersion of the sample data.

M

• Examines whether the process spread fits within the customer specification limits. It examines the process variation to the specification limits. This index does not account for increased defect rates due to shifts in the mean.

P

• USL = Upper Specification Limit. • LSL = Lower Specification Limit. • s = Standard Deviation for a sample, or an estimated standard deviation.

N O Q R S T U V W

• x-bar = the sample mean (or average); often denoted as an “X” with a line or a bar drawn above it.

X

• Larger the Cp the better.

Z

Rule of Thumb Cp > 1.5 to be capable, ideally. Some experts accept with a Cp > 1.0; however, that is the bare minimum.

Y

490

Encyclopedia

• Figure P-21 illustrates the comparison of the process spread versus the specification spread, wherein LSL represents the Lower Specification Limit and USL is the Upper Specification Limit.

A

Lower Spec Limit

Upper Spec Specifications are about 250% Limit the width of the process spread. Therefore, cp~2.5.

B C D E F G

1/2 Process Spread

1 Process Spread

1 Process Spread

Figure P-21: Generic Process Spread Versus Specification Spread (Cp)

H I J K L M N O P Q R S T U V W X Y Z

• Cpk—The difference between the Closest Specification Limit (closest SL) to the process mean (x-bar) divided by half of the Process Spread (3s); or [(USL–x-bar)/(3s)] OR [(x-bar–LSL)/(3s)]; whichever is smaller, using the sample mean (centeredness). • The most popular calculation of Process Capability because it considers both variability and mean of process output relative to customer requirements. • Examines the proximity of the current process mean to the closest customer specification limit (either the Upper or Lower, whichever is closer). This also is referred to as the process centeredness. • Therefore, Cpk = 1.0 when the mean of the current process is exactly 1/2 of the curve away from the nearest specification limit. One Cpk measures 1/2 of the process curve. • Cp is bigger than Cpk, given that Cp accounts for the full process spread, and Cpk accounts for half of the spread; Cp>Cpk. • Cp can be considered the “upper bounds” of Cpk when the process mean is on target. The difference between Cp and Cpk represents the potential gain in process capability if the process were to be centered. • Cpk is largest when the mean is located at the midpoint of the specification spread.

Proce ss Capability Analysis

491

Caution Always use the smaller Cpk number to reflect the closest specification limit distance to the process mean so as not to overstate the process capability.

A B C

Rule of Thumb

D

Cpk > 1.33, where 1.0 is barely capable, and 1.33 or larger is preferred.

E

For a Six Sigma (6σ) process, Cpk = 2.0.

F G

• Figure P-22 illustrates Cpk relative to the Lower Specification Limit (LSL); representing the following formula [(xbar–LSL)/(3s)].

H I J K

X Lower Spec Limit

Closest Specification Limit (LSL) appears to be of the process spread, or one complete -Bell. Therefore, cpk~ 1.0.

Upper Spec Limit

L M N O P Q R

1/2-Process Spread

Figure P-22: Generic Process Capability (Cpk)

• If the furthest specification limit is used, the process capability is overstated, as shown in Figure P-23.

Upper Miscalculated Cpk using the Spec Furthest Specification Limit (e.g. USL) appears to be about Limit

X

T U V W X

X Lower Spec Limit

S

3 bells of process spread. Therefore, cpk~ 3.0.

1/2-Process Spread

Figure P-23: Overstated Generic Process Capability (Cpk)

Y Z

492

Encyclopedia

Hint If Cpk is negative, the process mean is outside the closest specification limit as shown in Figure P-24. A B C

X

D E F G

Closest Specification Limit (LSL) appears to be Upper Lower inside of the Spec process mean Spec (x-bar) by one Limit Limit complete -Bell . ~ Therefore, cpk (1.0); or -1.0.

H I J K L

1/2-Process Spread

Figure P-24: Generic Negative Process Capability (-Cpk)

M N

Hint

O

Count “1/2-bells” or 1/2 of the process capability curve to equal one Cpk unit.

P Q R S T U V W X Y Z

• Cpm—Number of standard deviations the process mean is from a target value. • Examines the process centeredness relative to a target. • If the process were perfectly centered on the target, and if the process were capable, the Cp would equal the Cpm. That is, Cp = Cpm when the process is perfectly centered on its target. • Larger Cpm is better. • Cm—Capability of a machine; often used by DFSS (Design for Six Sigma). • Engineering Tolerance divided by 8 standard deviations. • [(USL–LSL)/(8s)]

Rule of Thumb Cm > 1.33 to be acceptable.

Proce ss Capability Analysis

493

• Capability Ratio (CR): Process Spread (6s) divided by Specification Spread (USL–LSL); or [(6s)/(USL–LSL)]. • Sometimes referred to as the natural tolerance. • CR is the numerical inverse of Cp. A

• Often expressed as a percentage (%), thereby multiplying the ratio by 100.

B

• Smaller the CR the better.

C D

Rule of Thumb CR < 0.5 to be “good” (or 50%).

E F G H

• Pp: Specification Spread (USL—LSL) divided by Process Spread (6 σ); or [(USL—LSL)/(6 σ)]; known as “long-term” process capability based on population data. • Pp and Ppk are referred to as performance metrics to distinguish from the short-term capability metrics (that is, Cp and Cpk). • Defines the long-term process capability, which reflects more inherent variation over time, relative to customer requirements. This index does not account for increased defect rates due to shifts in the mean. • Similar to Cp, but using the population data, rather than estimating with sample data. • USL = Upper Specification Limit. • LSL = Lower Specification Limit. • σ = Standard Deviation for a population. • The larger the Pp the better. • Ppk—The difference between the Closest Specification Limit (closest SL) to the population mean µ) divided by half of the Process Spread (3 σ); or [(USL–µ)/(3 σ)] OR [(µ–LSL)/(3 σ)]; whichever is smaller, using the population mean (centeredness). • Pp and Ppk are referred to as a performance metrics to distinguish from the short-term capability metrics (that is, Cp and Cpk). • Defines the long-term process capability, which reflects more inherent variation over time and considers both variability and mean of process output relative to customer requirements.

I J K L M N O P Q R S T U V W X Y Z

494

Encyclopedia

• Similar to Cpk, but using the population data, rather than estimating with sample data.

A B C D E F G H I J K L

• Pp is bigger than Ppk because Pp accounts for the full process spread of the population, and Ppk accounts for half of the spread; Pp>Ppk. Pp can be considered the “upper bounds” of Ppk when the process mean is on target. • ZL—Upper Specification Limit (USL) minus the mean (x-bar), divided by the standard deviation (s). • Measures the percentage area either inside or out of the requirement process by examining the process’s location (centeredness of the mean) relative to both the standard deviation and the lower specification limit. • Uses the Z statistic and references a Standard Normal Table to determine the area under the process curve. • ZL = [(x-bar–LSL) divided by 1s] • The bigger ZL the better.

M N

Rule of Thumb

O

ZL > +3 to be acceptable, and produce less than 0.01% defects. For a Six Sigma (6 σ) process, ZL = +6.

P Q R S T U V W X Y Z

• ZU—Lower Specification Limit (LSL) minus the mean (x-bar), divided by the standard deviation (s). • Measures the percentage area either inside or out of the requirement process by examining the process’s location (centeredness of the mean) relative to both the standard deviation and the upper specification limit. • Uses the Z statistic and references a Standard Normal Table to determine the area under the process curve. • ZU = [(USL–x-bar) divided by 1s] • Similar to ZL, the bigger ZU the better.

Proce ss Capability Analysis

495

Rule of Thumb ZU > +3 to be acceptable and produce less than 0.01% defects. For a Six Sigma (6 σ) process, ZU = +6. A

• ZMIN—Simply the smaller of the two ZL and ZU calculations, determining which is closest to the nearest specification limit to accurately state the process capability. Otherwise, the process capability would be overstated.

B C D E

• Used to help compute Cpk, since Cpk = ZMIN divided by 3.

F

• Similar to ZL and ZU, the bigger ZMIN the better.

G H

Rule of Thumb ZMIN > +3 to be acceptable, and produce less than 0.01% defects. For a Six Sigma (6 σ) process, ZMIN = +6.

I J K L M N

Yield • First Pass Yield (FPY) —Simply divides the number of good units produced by the number of units entering the process. • Also known as First Time Yield. • FPY ignores, or does not account for, the “hidden factory” which consumes resources to do “rework” and scrap products with defects, since the calculation occurs after any “inspection” is conducted to determine if a unit is good or not. • FPY = [Total Good units produced/Number of Units Entering] = (1-p)n = qn , where p = the probability of the number of defects, and q = probability that the defect will not occur. • Rolled Throughput Yield (RTY)—Measures the total DPMO for an entire process or product by multiplying the DPMO of each process step. • Represents the probability that a single unit can pass through the process free of any defects, thereby decrementing the capability calculation for any rework (and scrap) involved in making a unit good. Hence, RTY is the preferred yield calculation over FPY.

O P Q R S T U V W X Y Z

496

Encyclopedia

• RTY = [(1-(DPMO1/1,000,000)) x (1-(DPMO2/1,000,000)) x … (1-(DPMOn/1,000,000))], for “n” number of process steps. • For example, if the process had four steps, and each step produced a 90% yield, the Rolled Throughput Yield would equal: A

RTY = 90% x 90% x 90% x 90% = 66% (not 90%).

B

The math calculations breakdown to:

C

0.9 x 0.9 = 0.81 (for steps 1 and 2); then 0.81 x 0.9 = 0.73 for steps 1, 2 and 3; then 0.73 x 0.9 = 0.66 for all 4 steps, as shown in Figure P-25.

D E F G

Step 1

Step 2

Step 3

Step 4

Final Output

@ 90% Yield

@ 90% Yield

@ 90% Yield

@ 90% Yield

@ 66% Yield

H

Rolled Yield

81%

73%

66%

= 0.81 x 0.9

= 0.73 x 0.9

I J

= 0.9 x 0.9

Figure P-25: Rolled Throughput Yield Example

K L M N O P Q R S T U V W X Y Z

• The RTY is smaller than the lowest yield of any single process step. The RTY becomes exponentially smaller as the number of steps in the process grows.

How to Use the Tool or Technique Prerequisites to Examining Process Capability Prior to conducting a Process Capability study, the process in question must pass two tests. The first test—is process stability. This means that the process results remain in control, showing random, and steady state pattern over time. Process stability allows relatively accurate predictions about future yields or outputs, as well as the defect rate. Without process stability, the capability measures would be meaningless. Process stability can be determined by examining the process variation and whether it is randomly distributed around the mean and within the control limits. Typically three standard deviations on either side of the process mean define the control limits. A steady-state process is absent of any patterns such as • A trend (or drift) up or down • An oscillating pattern around the mean • A shift (or jump) in the data over time

Proce ss Capability Analysis

Any process spread is explained by common cause variation, rather than special cause variation. Common cause variation is the inherent process fluctuations present in all processes. An in-control process with only common cause variation is stable and predictable. In contrast, special cause variation causes the process to be unstable and unpredictable. A special cause occurrence is not inherent to, or beyond the natural variation of, the process (or system). Special cause variation is “assignable” to an unusual event, whereby once identified, it should be eliminated. This removal of special cause events from the data should occur prior to conducting a Process Capability analysis.

497

A B C D E

Note

F

If common cause is treated as special cause, resulting in process

G

adjustments (or tampering), then more variation is introduced into

H

the process. If special cause is treated as common cause, then the

I

special event is neglected, and the delayed reaction may increase the

J

defect rate.

K L

Tools and techniques determine if a process is in control, include a suite of Statistical Process Control (SPC) charts, or a Time Series plot or Run chart, and show the process data over time. Figure P-26 illustrates a process in control (delineated by a type of Statistical Process Control chart known as an x-bar-R chart). In contrast, Figure P-27 depicts four examples of processes out-of-control. (See Also “Control Charts—7QC Tool,” p. 217 for more detail.)

Normal Probability Plot The second test is the process normally distributed. When graphically represented, the frequency distribution follows a bell-shaped curve (that is, symmetrical around the process mean).

M N O P Q R S T U V W

Note

X

Normal distribution is a bell-shaped symmetrical distribution created

Y

by data said to be normally distributed.

Z

498

Encyclopedia Xbar-R Chart

A

Sample Mean

1.28

B

1.26

LCL=1.22556 1.22 1

D

G H

3

5

7

9

13

15

17

19

UCL=0.0944

0.075 0.050 R=0.0447 0.025 0.000

I

LCL=0 1

3

5

7

9

J K

11

0.110

Sample Range

F

= X=1.25132

1.24

C E

UCL=1.27708

11

13

15

17

19

Sample

Figure P-26: Generic Process in Control

L M

2 Special Cause Events

Trend Downward

N O

R S

Yield_Trend

Q

Yield_Extreme

P

90

90

80

70

60

70

50

1/1

1/12

T

Z

2/5

2/17

1/1

1/12

1/24

Date

2/5

2/17

Oscilating Pattern

90

Yield_Cycle

Yield_Shift

Y

Date

90

V X

1/24

Process Shift

U W

80

80

70

80

70

60

1/1

1/12

1/24

2/5

2/17

1/1

1/12

1/24

Date

2/5

2/17

Date

Figure P-27: Examples Failing Steady-State, In-Control Process Test

This is commonly displayed as a histogram, as shown in Figure P-28. Another means to test if the data are normal involves using the normal

Proce ss Capability Analysis

probability plot that examines the data around confidence intervals (usually 95%), as displayed in Figure P-29. If the data are normal, 95% of the data points will fall within the confidence interval. Notice how tight the data points are around the mean and within the confidence intervals displayed in Figure P-29. Moreover, the tool provides a statistic that represents the probability that the data falls within that confidence band. Normal data will yield a probability statistic (a.k.a. the P-value) greater than 0.05, or P-value > 0.05, again indicating that 95% confidence that the data could be well represented using a normal distribution.

499

A B C D E

Histogram of Sample 1 Normal

Normal

90

90

80

80

70

70

Frequency

Frequency

F

Histogram of Sample 2

60 50 40

H

60

I

50

J

40

30

30

20

20

10

10

0

G

K L

0 -3

-2

-1

0

1

2

3

-3

-2

-1

0

Mean 0.01499 StDev 1.007 N 1000

1

2

3 Mean -0.05705 StDev 1.021 N 1000

M N O P

Figure P-28: Examples of Two Normal Histograms

Q R

Probability Plot of Sample 2

S

Normal - 95% CI 99.99

Mean StDev N AD P-Value

99

-0.05705 1.021 1000 0.222 0.828

Percent

95

T U V W

80 P-value > 0.5; hence data are normally distributed, within the 95% Convidence Intervals.

50 20

X Y Z

5 1

0.01 -5

-4

-3

-2

-1

0

1

Figure P-29: Example of Normal Probability Plot

2

3

4

500

A B C D E F G H I J K L M N

Encyclopedia

Transforming Non-Normal Data If the data are non-normal, then the estimated defect rates and yields will be incorrect. (Note: non-normal data, being the opposite of normally distributed data, is defined as being distributed in any shaped-curve other than a symmetrical bell-shaped curve.) The calculated defect rates would be either too high or too low depending on the type of non-normality. Non-normal data can be remedied by transforming it. Using the principle of multiplying by a constant (that is, lambda), data transformation enables tools such as Capability Studies that require normal data to be used. These tools apply a non-linear, shape-changing model. A common transformation tool for non-normal data is called the BoxCox Transformation. Such a data conversion simply applies a new scale, similar to exchanging one monetary currency for another (that is, U.S. dollars to the Japanese Yen or the European Euro), converting the temperature scale from Celsius to Fahrenheit, measuring distance in meters rather than inches, or converting liquid measurement from gallons to liters. Transformation algorithms are available with most statistical business software packages, such as MINITAB. (See Also “Histogram—7QC Tool,” p. 330 for more detail.)

Required Information When calculating the more common Process Capability Indices, only four pieces of information are required:

O

1. the Upper Specification Limit (USL)

P

2. the Lower Specification Limit (LSL)

Q R S T U V W X Y Z

3. the standard deviation (s); or the “fatness” or dispersion of the process 4. the process mean (x-bar); or the center of the process

Capable Versus Acceptable In addition to answering affirmatively, the two prerequisite questions are 1) Is the process stable and in control? and 2) Is the process normally distributed? two other key questions will be asked when interpreting the results of the capability study. These questions are 1. Is the process capable? 2. Are the process results acceptable? The current process spread (of a steady-state process) answers the question of capability. If the customer specification limits fall 1.5 times outside the natural process variation (that is, Cp > 1.5), then the process is capable of producing the outcomes specified by the customer. However, a capable process’s results may fall shy of actually producing the desired deliverables and hence may not be acceptable.

Proce ss Capability Analysis

501

Producing “acceptable” results requires the process mean to be centered on or extremely close to the target value. This question focuses on the centeredness of the process—how “accurate” is the process in meeting customer requirements?

Adjustments When using Process Capability Indices to determine what action to take to improve the process, know that it is easier to center the process (move Cpk onto the target), rather than reduce the process spread (a.k.a. dispersion, width, or variation). (Note: Process capability means that the process spread is capable of fitting within the customer specification limits; it does not necessarily need to be doing so. For example, if you were about to purchase a new Hummer vehicle and you live in a home built before the 1950s, you might question whether the car could fit within the width of your garage door (that is how to consider process capability). When studying process capability, what could be impacting the variation (or standard deviation)? By understanding the cause behind the variation, one can begin to develop and focus the improvement strategies. The variance (or spread) of a steady state, in control process is explained by the common cause. Commonly used classifications of potential root causes include: Machine, Materials, Methods, Metrics, Mother Nature (or Environment), and People—also often referred to as 5Ms and P. (See Also “Cause-and-Effect Diagram,” in the “Hints and Tips” section, p. 182, for more detail on 5Ms and P.)

Manual Calculations of Process Capability (Cp and Cpk) Example 1 The following sample data is provided:

A B C D E F G H I J K L M N O P Q R

• The process is stable, in control, and normally distributed.

S

• Mean (x-bar) = 40

T

• Standard Deviation (s) = 2.6

U

• Lower Specification Limit (LSL) = 25 • Upper Specification Limit (USL) = 65

V W X

The manual calculations to determine Cp and Cpk are as follows:

Y

Step 1.

Z

Draw the bell-shaped curve and place the given values on the curve, as illustrated on Figure P-30, to serve as a double-check and minimize any inadvertent human errors in the manual calculations. a. Presume the curve’s spread is normal (that is, 3 standard deviations on either side of the mean, or 3 x 2.6 = 7.8). Therefore, the upper tail is at 40 (mean) + 7.8 (1/2-spread) = 47.8; and the lower tail is at 40 (mean)–7.8 (1/2-spread) = 32.2.

502

Encyclopedia X LSL

USL

A B C

32.2 25

D E F G

Step 2.

I K L N

b. Estimate the Cpk. How big does it look? Which specification limit is the mean closest to, and how many 1/2-bell shapes are between it and the mean?

O P Q

ANSWER: “Closest spec. limit is the LSL, and it looks to be about two 1/2-bells away (or 2 Cpk).”

R

V W

Ask the key capability questions before calculating the actual indexes and develop a “ballpark” estimate as a double-check in anticipating the results.

ANSWER: “Yes, the process spread looks to fit inside the specification spread.”

M

U

Step 3.

What is the Cp? Calculate Cp using the following formula: [(USL-LSL)/6s]. a. ANSWER: Cp = [(65-25)/(6 x 2.6)] = (40/15.6) = 2.56; Cp @ 2.56 > the target Cp of 1.5. Therefore, the specification spread is 256% (or about 2.5 times) wider than the process spread. “Yes, this is a very capable process.”

X Y Z

65

a. Estimate whether the process appears capable. Does the process spread appear to be more narrow than the specification spread?

J

T

47.8

Figure P-30: Example 1—Process Capability Curve

H

S

40

Rule of Thumb Cp should be > 1.5.

Proce ss Capability Analysis

Step 4.

503

How do you calculate Cpk? First, recall that the process mean is closest to the Lower Specification Limit (LSL). Thus, the formula to use is [(x-bar–LSL)/3s]. ANSWER: Cpk = [(40-25)/(3 x 2.6)] = (15/7.8) = 1.92. Cp @ 1.92 > the target Cpk of 1.33. Therefore, the process is “acceptable.” (This matches the “ballpark” estimate of about 2 determined in Step 2.)

A B C

Rule of Thumb

D

Cpk should be >1.33.

E F

Example 2 The following sample data is provided: • The process is stable, in control, and normally distributed. • Mean (x-bar) = 40

G H I J K L

• Standard Deviation (s) = 2.6

M

• Lower Specification Limit (LSL) = 35

N

• Upper Specification Limit (USL) = 65

O

The manual calculations to determine Cp and Cpk are as follows: Step 1.

Step 2.

Draw the bell-shaped curve and place the given values on the curve, as illustrated on Figure P-31, to serve as a double-check and minimize any inadvertent human errors in the manual calculations.

P Q R S T U

a. Presume the curve’s spread is normal (that is, 3 standard deviations on either side of the mean, or 3 x 2.6 = 7.8). Therefore, the upper tail is at 40 (mean) + 7.8 (1/2-spread) = 47.8; and the lower tail is at 40 (mean)–7.8 (1/2-spread) = 32.2.

W

Ask the key capability questions before calculating the actual indexes and develop a “ballpark” estimate as a double-check in anticipating the results.

Z

a. Estimate whether the process appears capable? Does the process spread appear to be more narrow than the specification spread?

V X Y

504

Encyclopedia X LSL

USL

A B C

32.2 35

D E

40

47.8 65

Figure P-31: Example 2—Process Capability Curve

F

ANSWER: Yes; the process spread looks to be narrow enough to fit within the specification spread.

G H

b. Estimate the Cpk; how big does it look? Which specification limit is the mean closest to, and how many 1/2-bell shapes are between it and the mean?

I J K

ANSWER: Closest spec. limit is the LSL, and it looks to be slightly less than one 1/2-bells away (about one-third less, or Cpk~0.6).

L M N O

Step 3.

P

ANSWER: Cp = [(65-35)/(6 x 2.6)] = (30/15.6) = 1.92. Cp @ 1.92 > the Cp Rule of Thumb of 1.5. Therefore, the specification spread is 192% wider than the process spread. “Yes, this process is capable.”

Q R S T U V W X Y Z

How do you calculate Cp using the following formula: [(USL-LSL)/6s]?

Step 4.

How do you calculate Cpk? First, recall that the process mean is closest to the Lower Specification Limit (LSL). Thus, the formula to use is [(x-bar—LSL)/3s]. ANSWER: Cpk = [(40-35)/(3 x 2.6)] = (5/7.8) = 0.64. Cp @ 0.64 < the Cpk Rule of Thumb of at least 1.0. Therefore, the process is “unacceptable.” (This matches the “ballpark” estimate of less than a 1/2-bell determined in Step 2.) Recall that the difference between Cp and Cpk represents the potential gain in process capability if the process were to be centered. In this case, the difference between Cp and Cpk (1.92–0.64) is 1.28. Thus, exploring possible adjustments to improve the centeredness of the process probably would be worthwhile.

Proce ss Capability Analysis

505

Automated Calculations of Process Capability (Cp and Cpk) Using MINITAB Example 3 The following sample data is provided: • The data set found in Table P-5, (where n = 50 and no sub-groups):

A B

Table P-5: Example 3 Data Set

C

26.5

25.9

23.9

22.1

23.8

23.6

30.1

20.7

25.3

27.6

25.2

25.4

23.3

22.4

26.6

31.3

26.3

24.4

26.2

22.6

E

30.2

22.9

21.6

31.2

25.6

23.2

26.6

23.1

26.5

22.3

F

25.2

24.1

25.4

26.5

29.3

26.5

25.4

31.5

28.7

28.2

26.5

21.2

22.1

22.7

24.9

28.8

19.3

23.0

22.8

28.6

D

G H I J

• Upper Specification Limit = 50

K

• Lower Specification Limit = 20

L

• The process is stable as illustrated in Figure P-32. In MINITAB, select Graph > Time Series Plot > Simple. (See Also “Graphical Methods,” p. 323)

M N O P Q

Time Series Plot of Example 3 Data 32

R S

30

T

C9

28

U V

26

X 24

W X Y

22

Z

20

1

5

10

15

20

25

30

Time

Figure P-32: Example 3—Time Series Plot, Using MINITAB

35

40

45

50

506

Encyclopedia

• The process is normally distributed, as shown in Figure P-33. [In MINITAB, for the histogram, select Graph > Histogram > With Fit; for the Normal Probability plot, select Graph > Probability Plot > Single. (See Also “Graphical Methods,” p. 323) Histogram of Example 3 Data 12

B

10

E F

I J K L

6

5

1 0 21

U V

24

27

30

15

20

Data Fits Bell-Shaped Curve

25

30

35

Normal Distribution; P-value > 0.05.

Figure P-33: Example 3—Histogram and Normal Probability Chart Using MINITAB

The procedure to calculate Cp and Cpk using MINITAB are as follows: Step 1.

Arrange the data in a MINITAB Worksheet, preferably in a single column. (Note that the data may be in a group of columns, with subgroups across rows; however, this book’s examples follow the single column approach.)

Step 2.

In MINITAB, select Stat > Quality Tools >Capability Analysis > Normal… from the toolbar at the top of the screen, and a “dialog box” will appear.

Step 3.

Within the MINITAB Capability Analysis (Normal Distribution) dialog box, found in Figure P-34, select the appropriate column of data in the worksheet. Enter the subgroup size for the data set. (If no subgroups, enter 1 in the appropriate field.) Enter the values for the upper and lower specification limits. Select OK to generate the graph and output found in Figure P-35.

Q

T

50 40 30

10 2

P

S

70 60

20

4

N

R

25.34 2.904 50 0.477 0.228

80

M O

Mean StDev N AD P-Value

95 90

G H

25.34 2.904 50

Frequency

D

Normal - 95% CI 99

Mean StDev N

8

Frequency

C

Probability Plot of Example 3 Data

Normal

A

W X

Note

Y

Notice that the MINITAB dialog box contains a button in the upper-

Z

right corner, called Box-Cox, which transforms non-normal data if needed. In Example 3, the data are normal, hence by-pass the extra step of selecting the Box-Cox Transformation. Ask the key capability questions, and answer them based on the above results in the MINITAB output, found in Figure P-35.

Proce ss Capability Analysis

507

A B C D E F G H I

Figure P-34: MINITAB Dialog Box for Example 3

J K L

Process Capability for Example 3 LSL LSL Target USL Sample Mean Sample N StDev(Within) StDev(Overall)

M

USL

20 * 50 25.342 50 2.70842 2.91862

Within Overall

N O

Potential (Within) Capability Cp

1.85

CPL

0.66

CPU

3.03

Cpk

0.66

Overall Capability

18.0

Observed Performance PPM < LSL PPM > USL PPM Total

20000.00 0.00 20000.00

22.5

27.0

31.5

36.0

40.5

45.0

Exp. Within Performance

Exp. Overall Performance

PPM < LSL PPM > USL PPM Total

PPM < LSL PPM > USL PPM Total

24283.98 0.00 24283.98

49.5

33601.31 0.00 33601.31

Pp

1.71

PPL

0..61

PPU

2.82

Ppk

0.61

Cpm

*

P Q R S T U V W X Y Z

Figure P-35: Example 3 MINITAB Process Capability Output

Step 4.

The first questions are a. Does the process appear to be capable?

508

Encyclopedia

b. Does the process spread appear to be more narrow than the specification spread? ANSWERS: Yes; the process spread looks to be narrow enough to fit within the specification spread. And MINITAB calculated Cp @ 1.85 is greater than the Cp Rule of Thumb of 1.5. Therefore, the specification spread is 185% wider than the process spread. Yes, this process is capable.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Step 5.

The second questions are, Is the process acceptable? Is the process centered? a. Determine which specification limit the mean is closest to,

and estimate how many 1/2-bell shapes are between it and the mean? ANSWER: Closest spec. limit is the LSL, and it looks to be slightly less than one 1/2-bell away (about one-third less, or Cpk~0.6). b. ANSWER: MINITAB calculated Cpk @ 0.66, which is less than the Cpk Rule of Thumb of at least 1.0. Therefore, the process is “unacceptable.” (This matches the ballpark estimate of less than a 1/2-bell.) c. Recall that the difference between Cp and Cpk represents the potential gain in process capability if the process were to be centered. In this case, the difference between Cp and Cpk (1.85–0.66) is 1.19. Thus, exploring possible adjustments to improve the centeredness of the process probably would be worthwhile. Another powerful display of the same data uses the MINITAB Six Pack option that provides six process capability output graphs, as illustrated in Figure P-36. From the drop-down menu, select Stat > Quality Tools >Capability Sixpack > Normal… and follow the same procedure outlined in Steps 3 through 5. Figure P-36 uses the same data as displayed in Figure P-35. The Six Pack includes a Control chart [I Chart (Individuals Control chart), or an x-bar chart if the data contains subgroups] of the individual observations, a Moving Range Chart (MR chart), and a run chart of the last 25 observations (or subgroups) displayed down the left side of the output. The top-right corner Capability Histogram of the data is the same graph used in the Capability Analysis option (Figure P-35). Also on the right side, the output includes a Normal Probability Plot and a Process Capability Plot that provides the within and overall capability statistics Cp,

Proce ss Capability Analysis

509

Cpk, Cpm (if a target is specified), Pp and Ppk. (See Also “Control Charts—7QC Tool,” p. 217 for a discussion on I charts, x-bar charts, MR charts and Normal Probability plots; “Run Chart—7QC Tool,” p. 610; and “Histogram—7QC Tool,” p. 330) A B C D E F G H I J K L Figure P-36: Example 3 MINITAB Process Capability in a Six Pack

M N O

How to Analyze and Apply the Tool’s Output Additional Information from MINITAB Output MINITAB provides output in addition to Cp and Cpk. Referring to Figure P-36, MINITAB publishes a Process Data table profiling the data set that is • Lower Specification Limit (LSL). • Target value (if known; if unknown MINITAB places an asterisks (*) in place of a value). • Upper Specification Limit (USL).

P Q R S T U V W X

• Sample mean (x-bar).

Y

• Sample size (n).

Z

• Standard deviation (StDev (Within))—looking at variation within the sample subgroups; also indicated on the graphical image by the

510

Encyclopedia

solid line. In case of Example 3, there were no sub-groups, but this estimates the “short-term” standard deviation (s). • Standard deviation (StDev (Overall)) across all measures, estimating long-term or the population ( σ); also indicated on the graphical image by the dashed-line. A B C D E F

• If the Within and Overall curves display a significant difference, it may indicate a non-random pattern of variability found when a process is out of control due to the presence of a special cause event. The MINITAB Potential Capability in the upper-right corner of the output contains the following information (as shown in Figure P-35):

G

• Cp (Process spread)

H

• CPL (Calculates the Cpk for the Lower Specification Limit (LSL). Recall that this Capability Index that accounts for both the process mean and the LSL, and the one-sided process spread relative to the LSL, is represented by three times the Within standard deviation.

I J K L M N

• CPU calculates the Cpk for the Upper Specification Limit (USL). Recall that this Capability Index that accounts for both the process mean and the USL, and the one-sided process spread relative to the USL, is represented by three times the Within standard deviation.

O

• Cpk (the smaller of the CPU and CPL).

P

• Pp (process spread for the overall capability).

Q R S T U V W X Y Z

• PPL calculates the Ppk for the Lower Specification Limit (LSL). Recall that this Capability Index accounts for both the process mean and the one-sided process spread relative to the LSL. This is represented by three times the Overall standard deviation. • PPU calculates the Ppk for the Upper Specification Limit (USL). Recall that this Capability Index accounts for both the process mean and the one-sided process spread relative to the USL. This is represented by three times the Overall standard deviation. • Ppk (the smaller of the PPU and PPL). • Cpm indicates the number of standard deviations the process mean is from a target value. This number is provided only if a target is specified, otherwise an asterisk (*) will represent an empty value.

Proce ss Capability Analysis

511

The MINITAB Process Performance results can be found in the lower bottom section of the output and provides the following information (as shown in Figure P-35): • Observed Performance—Actual number of parts per million (PPM) (similar to DPMO) observed beyond the specification limits.

A

• PPM < LSL—Number of PPM found less than the LSL.

B

• PPM > USL—Number of PPM found above than the LSL.

C

• PPM Total—The total number of PPM found outside of both specification limits.

D E F

• Exp. Within Performance—The expected values actual number of parts per million (PPM) observed beyond the specification limits and the Within standard deviation.

H

• PPM < LSL—Number of PPM found less than the LSL.

I

• PPM > USL—Number of PPM found above than the LSL.

J

• PPM Total—The total number of PPM found outside of both specification limits. • Exp. Overall Performance—The expected values actual number of parts per million (PPM) observed beyond the specification limits and the Overall standard deviation. • PPM < LSL—Number of PPM found less than the LSL. • PPM > USL—Number of PPM found above than the LSL. • PPM Total—The total number of PPM found outside of both specification limits. For additional detail on the MINITAB output, explore its Help menu and its “StatGuide” found on the toolbar. MINITAB offers concise and complete explanations of terms and the statistics used. In addition, MINITAB includes a thorough Help feature within the dialog boxes, shown in Figure P-34.

Additional Process Capability Analysis and Application The “How to Use the Tool or Technique” section of this “Process Capability Analysis” entry embeds relevant analysis and hints and tips as it introduces a concept. However, Figure P-37 summarizes the process capability concept with four diagrams illustrating capable versus incapable processes and acceptable versus unacceptable processes.

G

K L M N O P Q R S T U V W X Y Z

512

Encyclopedia • Mean centered on target • Spread beyond Specs; Poor Cp and Pp, Process Incapable • Less than ½-bell between Mean and closest spec; Poor Cpk and Ppk Unacceptable

• Mean centered on target • Spread beyond Specs; Good Cp and Pp, Process Capable • At least one ½-bell between Mean and closest spec; Good Cpk and Ppk Acceptable

Incapable; Unacceptable

A

LSL

B

Target

Capable; Acceptable

USL

LSL

Target

USL

C D E F G

7

8

9

LSL

10

11

Target

12

13

14

7

USL

8

9

LSL

10

11

Target

12

13

14

13

14

USL

H I J K L M N O P Q R S

7

8

9

10

11

12

13

14

7

Incapable; Unacceptable • Mean Off target • Spread beyond Specs; Poor Cp and Pp, Process Incapable • Less than ½-bell between Mean and closest spec; Poor Cpk and Ppk; Unacceptable

8

9

10

11

12

Capable; Acceptable • Mean Off target • Spread within Specs; Good Cp and Pp, Process Capable • Less than ½-bell between Mean and closest spec; Poor Cpk and Ppk; Unacceptable

Figure P-37: Summary Process Capability Diagram

T U V W X Y Z

Hints and Tips • Conduct Process Capability studies when the performance of a steady-state process needs to be monitored. If the focus is to reduce any out-of-spec products, then the following Process Capability Indices are preferred: Cpk and Cp (or Ppk and Pp). • If the goal of monitoring the process performance is to minimize variation of the process mean from a target, then the preferred Process Capability Index is Cpm. • Prior to beginning a Process Capability study, ensure that both the process exhibits random variation over time (using either a Control

Proce ss Capability Analysis

513

chart or run chart) and the data are distributed normally (using either a Normal Probability chart or histogram). Otherwise, the process capability calculations will be meaningless. • When conducting a manual Process Capability analysis, remember to: • Diagram the process bell-shaped curve, • Label the curve with the known data which should include the mean (x-bar), the standard deviation (s), and the upper and lower specification limits (USL and LSL), and • Calculate the process spread (+ 3-sigma, equating to 99.73% of the data) by multiplying the standard deviation by three, then both add it and subtract it from the mean, and label the curve with those values to define where the tails narrow. • Good process performance—capable and acceptable: • Capability determined by variation (or spread): Cp > 1.5; some experts accept 1.0, but with the long-term tendency to drift +1.5 sigma, a Cp > 1.5 is preferred (or Pp > 1.5). Then the process spread fits within the specification spread. • Capability and acceptability calculated by Cpk because it considers both the process variability and the proximity of the mean to the nearest specification limit. Cpk > 1.33; some experts accept 1.0, depending on the circumstances (or Ppk > 1.33). • Acceptable: Cpm: centered on the target. • If process perfectly centered, then Cp = Cpk. • If the process is perfectly centered and perfectly on target, then Cpm = Cp = Cpk. • Recall that the tails of a normal distribution extend beyond + 3-sigma, or 99.73% of the data. Hence, the probability of “poor” product from a capable and acceptable process can happen (probably 0.27% of the time, represented by either tail of the distribution). • Motorola’s criteria for a Six Sigma process: • A Six Sigma capable process refers to tight process variation, wherein a 12 σ spread easily fits within the specification limits

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

514

Encyclopedia

and still has room for a 1.5 σ drift to account for long-term variation. Hence, with the process mean centered, the process spread of 6 σ on either side of the mean would fit well within the A B C D E F G H I J K L M

specification limits, and accommodate for a mean drift of + 1.5. • Process adjustments: • It is easier to adjust centeredness than it is to address variation (spread). • Recall that the difference between Cp and Cpk represents the potential gain in process capability if the process were to be centered. • The amount of out-of-spec product can be calculated by using DPMO and Yield calculations. It can also be calculated by using both Cp and Cpk, but not either index alone. • Two processes with identical Cpk-values may have different amounts of product out-of-spec and may need different actions to improve each of them.

N O P

Caution

Q

As with all calculated statistics, be aware that they contain inherent

R

tendency for error—human error. Potential error topics include sam-

S

pling, sample size, measurement error, non-transformed non-normal

T

data, and lack of confidence interval communication. Be aware of

U

these potential “gotchas” and plan to prevent them.

V W X Y Z

Supporting or Linked Tools Supporting tools that might provide input when developing a Process Capability Analysis include • Histogram (See Also “Histogram—7QC Tool,” p. 330) • Normal Probability Plot, (including the Normality Test) (See Also “Control Charts—7QC Tool,” p. 217) • Control charts (including Time Series plots or run charts) (See Also “Control Charts—7QC Tool,” and “Run Chart—7QC Tool,” p. 217 and 610, respectively.)

Proce ss De cision Program Charts (PDPC)—7M Tool

515

A completed Process Capability Analysis provides input to tools such as • QFD (See Also “Quality Function Deployment (QFD),” p. 543) • Root Cause analysis tools and techniques • Concept Generation methods (See Also “Failure Modes and Effects Analysis (FMEA),” in the “Hints and Tips” section, p. 295) • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) Figure P-38 illustrates the link between the Process Capability analysis and its related tools and techniques. 5-Whys Technique

Root Cause Analysis Techniques Concept Generation Methods

Control Charts

Run Charts (or Time Series plots)

C D E G

Histogram

Process Capability Analysis

B

F

5M and P Technique and its variants QFD

Normal Probability Plot

A

H I J K L M N O

Brainstorming Technique

FMEA

Figure P-38: Process Capability Analysis Tool Linkage

P Q R S T U

Process Decision Program Charts (PDPC)—7M Tool

V W

What Question(s) Does the Tool or Technique Answer? What might go wrong during the planning of this complex, significant project? PDPCs help you to • Plan and prevent problems from occurring in a high-stakes project • Graphically plot course of action when many events or milestones are uncertain

X Y Z

516

Encyclopedia

Alternative Names and Variations • Variations on the tool include FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) A B C

When Best to Use the Tool or Technique A Process Decision Program Chart (PDPC) is a planning and decisionmaking tool used when developing a large, complex project plan.

D E F G H I J K L M N O P Q R S T U V W X Y Z

Brief Description The PDPCs use a structured technique to identify what could go wrong in a complex, high-stakes project. Projects with a number of uncertainties and with critical deadlines are good candidates for the PDPC tool, and it works well on one-time unique projects. The technique maps the various event sequences, typically into a Tree diagram structure, and develops contingencies in parallel. Because often the appropriate countermeasures are unknown, the PDPC technique accounts for easy plan adjustments in response to a problem. PDPC applies best to projects addressing new, unique, or complex problems, requiring perhaps difficult and challenging activities, such as the first-time planning and implementing a mission critical project—for example, the first-time planning, designing and implementing of 100% digital-voting system for the upcoming national political election for President. Also the PDPC technique is dynamic, so it deals well with the unknown, where problems may be anticipated, but the appropriate contingency remains uncertain. Examples of such projects include accident prevention and a new system (or policy) being implemented. Given that the PDPC is a problem-prevention, risk planning management tool, it is comparable to the FMEA. Both tools explore what the potential failures could be and document countermeasures against the potential causes. Both tools are designed to be used dynamically and updated, refreshed and revised as part of an evergreen process. The PDPC evolved from a need in project planning to address risk in high-stakes, often unique projects with lots of uncertainty and critical deadlines to meet. Hence, the PDPC follows and supports the project planning process. The FMEA tends to enjoy a broader appeal because it applies simple risk scenarios in addition to the complex, and it also applies to ongoing operations, not just onetime projects. The FMEA distinguishes itself from a PDPC primarily because of its more rigorous prioritization component, but also its more detailed structure to dissect a potential root cause’s impact (or effect) and to document a more complete action plan. If the scenario warrants it, both the PDPC and FMEA techniques could apply, and complement one another in the problem-prevention, risk management planning effort. (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287)

Proce ss De cision Program Charts (PDPC)—7M Tool

The PDPC is a member of the 7M Tools, attributed in part to Dr. Shewhart, as seven “management” tools, or sometimes referred to as the 7MP or seven management and planning tools. These 7M Tools make up the set of traditional quality tools used to analyze quantitative data. The 7M Toolset includes: 1) Activity network diagrams or Arrow diagrams; 2) Affinity diagrams; 3) Interrelationship digraphs or Relations diagrams; 4) Matrix diagrams; 5) Prioritization matrices, often replacing the more complex Matrix data analysis; 6) Process decision program charts (PDPC); and 7) Tree diagrams. The Quality Toolbox, by Nancy Tague, presents the 7M Tools ranked from those used for abstract analysis to detailed planning: Affinity diagram, Relations diagram, Tree diagram, Matrix diagram, Matrix data analysis (commonly replaced by a more simple Prioritization Matrix), Arrow diagram, and Process Decision Program Chart (PDPC).

Three Different PDPC Structures A PDPC can be constructed one of three ways: 1) a forward linear (horizontal) Tree diagram, 2), a reversed linear (vertical) Tree diagram, and 3) an outline format. The first Tree-like method depicts the sequence of prescribed steps from beginning to end. Structurally, the diagram often flows from left to right. This horizontal structure typically includes five levels of branches or detail. As the project progresses, if an event (or step) encounters difficulties, then contingency steps to resolve the issue are developed realtime and continue the progression toward the final goal. Figure P-39 illustrates a horizontally structured PDPC.

517

A B C D E F G H I J K L M N O P

Low Participation

Subject Matter Experts Identified

Curriculum Developed

Use Discoverbased learning and Integrate with job profile

Q R

Programs Deployed

S

Learning Program Deployed

Deploy Six Sigma throughout organization

Program Office Formed

Key Stakeholders Identified

Requirements Defined

Projects Approved

Mentors Assigned

Board of Directors Identified

Project Portfolio Selected

Governance Policies Defined

Program Office Headcount Established

Staff Hired and Trained

Certification Requirements Defined

U Six Sigma evident throughout operations

V W

Community of Practice Infrastructure Defined

X

Communication Plan Developed

Y Z

Roles Defined

Reverted to old roles

T

Knowledge and Skills Identified

Analyze root cause; take action

Figure P-39: Illustration of a Horizontal PDPC

Reverted to old roles

Integrate into Presidental operational reviews

518

A B C

Encyclopedia

The second Tree diagramming method starts with the end (or goal) of the project and walks backward. Hence, the mapping of events involves the various options or considerations at critical junctures. Structural layout typically flows from the top, with the higher level details, and flows downward to the bottom of the page, with the lower level details. Typically, the structure contains five levels of detail as rows and resembles a classic Work Breakdown Structure (WBS). At the lowest level, a set of conditional what-if questions lead to corresponding countermeasures. Figure P-40 shows a schematic of a vertical-structured PDPC.

D

Deploy Six Sigma throughout organization

E F G

Learning Program Deployed

Key Stakeholders Identified

Program Office Formed

H I

Subject Matter Experts Identified

Board of Directors Identified

Requirements Defined

Program Office Headcount Established

Communication Plan Developed

Roles Defined

J K

Curriculum Developed

Programs Deployed

Projects Approved

Mentors Assigned

Certification Requirements Defined

Knowledge and Skills Identified

Project Portfolio Selected

Governance Policies Defined

Staff Hired and Trained

Curriculum Unfinished

Low Participation

No Sponsors

Mentors Inexperience

Projects Failed

More than Six Sigma required

Projects Failed

Not adhered to

Overworked

Review Business Benefits

Use Consultants temporarily

Review business need and sponsor-level

Explore operational integration

L M

Community of Practice Infrastructure Defined

Too busy

Roadblocks Put Up

Revert to old roles

N O P Q R S T U V

Hire Consultants

Use Discoverbased learning and Integrate with job profile

Use Consultants temporarily

Look at project team dynamics and Discoverybased learning

Review priorities

Review WIIFMs

Analyze root causes

Analyze root causes

Figure P-40: Illustration of a Vertical PDPC

The third structure uses an outline format with the major categories representing the milestones and subheadings as the lower level of detail. This method resembles an outline of a paper, wherein the parsing of information into more granular detail is indicated by increasing the indentation of the subtopics.

W X Y Z

How to Use the Tool or Technique The procedure to develop a PDPC, using the vertical-Tree diagramming approach, illustrated in Figure P-40, is as follows: Step 1.

Confirm the project objective and create a high-level Process map of three to eight milestones or key activity categories. (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

Step 2.

Record the milestones and activities in the PDPC for the top three levels of detail.

Proce ss De cision Program Charts (PDPC)—7M Tool

519

a. Select one of the three PDPC structures. b. Document the high-level process contents in the PDPC— level one. c. Brainstorm the next two levels of activities detail and document them on the appropriate hierarchical relationship of the PDPC—levels two and three. (See Also “Brainstorming Technique,” p. 168) Step 3.

Develop the PDPC contingencies. a. Referencing the lowest level of detail documented on the PDPC, from Step 2.c., brainstorm what could go wrong. Discuss what-if scenarios to uncover potential failures. b. Discuss and evaluate the potential problems. Identify those that are improbable and/or have an insignificant impact and eliminate them. c. Document the remaining problems on the PDPC—level four. d. Brainstorm potential countermeasures for each PDPC problem. These contingencies could include modifications to the planning process or developing a respondent action plan if the problem occurs. The action plan could aim to prevent, minimize, or transfer the risk to another party. (See Also “Poka-Yoke,” p. 462, for ideas on potential countermeasures.) e. Document the countermeasures on the PDPC—level five. This level is often denoted by dashed Tree branch lines to the countermeasure, or drawing a cloud-like structure around it.

Step 4.

Finalize the PDPC and communicate it to the appropriate stakeholders to incorporate the tool into ongoing planning and review sessions for the project. a. Review, refresh, revise, and communicate the PDPC throughout the duration of the project.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

520

Encyclopedia

Hints and Tips • If using multiple contingency planning tools, start with the PDPC to organize and diagram potential problems. The Tree-structure A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

serves as an easy communication vehicle to focus high-level discussions about uncertainties and risk. Use the PDPC as input to other contingency tools, such as FMEA. (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) • Thought-starter questions used to identify potential problems in the brainstorming session include • What if…? • What inputs are required? What are their respective operational definitions of “good,” and what could go wrong…? • Who is supplying the inputs? Is there a mutual understanding around supplier specifications and requirements? What are the suppliers’ CTQs? Are they documented and agreed to, and what could go wrong…? (See Also “Measurement Systems Analysis (MSA),” for a discussion on operational definition, p. 412, and “Critical to Quality (CTQ),” p. 242) • What are the dependencies, minimum conditions, and assumptions for each of the inputs, processes, and outputs, and what could go wrong…? • What is out of scope, and what could go wrong…? • What are the potential impacts from the 5Ms and P (Machine (Technology), Method, Materials, Measurement, Mother Nature (environment, including cultural, political, regulatory), and People), and what could go wrong…? (See Also “Causeand-Effect Diagram—7QC Tool,” p. 173, for further discussion on triggers to uncover potential root cause.) • What has been the past experience, and what went wrong that could be replicated in this situation…? What is different from before…? • How could we sabotage this project’s success if we wanted to?

Proce ss De cision Program Charts (PDPC)—7M Tool

521

Supporting or Linked Tools Supporting tools that might provide input when developing a PDPC include • Project charter (See Also “SMART Problem and Goal Statements for a Project Charter,” p. 665)

A

• Process map (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

B

• VOC and VOB data gathering (See Also “Voice of Customer Gathering Techniques,” p. 737)

D

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

F

C E G

A completed PDPC provides input to tools such as

H

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

I

• Control plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

J

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) • Transition and/or Implementation plan (See Also “Matrix Diagrams— 7M Tool,” p. 399) Figure P-41 illustrates the link between a PDPC and its related tools and techniques.

K L M N O P Q

Project Charter

Control Plan

R S

Process Map

PDPC

FMEA

T U

Transition and/or Implementation plan

VOC and VOB Data Brainstorming Technique

Figure P-41: PDPC Tool Linkage

Variations FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287)

V W X Y Z

522

Encyclopedia

Process Map (or Flowchart)—7QC Tool What Question(s) Does the Tool or Technique Answer? What are the components of the process; what is involved? A

Process maps help you to

B C D E F G H I J K L M N

• Understand the process and where the opportunities for improvement exist. • Gain a common understanding of a process and identify waste and areas where the process is poorly defined. • Display and examine the current process and compare it with what could be. • Communicate what is entailed in the process—its sequence of activities, inputs, outputs, and who is involved. Graphically outline a procedure. • Plan (or simulate) the future or improved process. • Categorize current process into its key components: activities, inputs, and outputs.

O P Q R

Alternative Names and Variations This tool is also known as

S

• Flowchart; flow diagram

T

• Activity flowchart or Activity Process map

U V

• Deployment flowchart or process map, cross-functional flowchart, or swim-lane Process map

W

• Detailed Process map (or flowchart)

X Y Z

• High-level Process map (or flowchart) • Workflow diagram Variations on the tool include • Arrow diagram or Activity Network Diagram (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127)

Proce ss Map (or Flowchart)—7QC Tool

523

• SIPOC (See Also “SIPOC (Supplier-Input-Process-Output-Customer,” p. 663) • Value Stream map or Process map (See Also “Value Stream Analysis,” p. 727) A

When Best to Use the Tool or Technique A Process map is a communication tool that diagrams what activities occur today in a process or what could or should happen. It is a good planning and analytical tool to examine the components of a process (inputs, activities, outputs, metrics, and process players) to identify potential areas for improvement.

B C D E F G

Brief Description A Process map (or flowchart) schematically portrays a set of interrelated resources and activities to transform inputs into outputs. It identifies the steps in sequential order and its associated inputs and outputs. It may vary in degree of detail from a high-level, depicting only the critical few process categories or milestones, to the minutest detail, describing the lowest level tasks. Fundamentally, the flowchart centers on the sequenced activities but also may include information about its inputs, outputs, metrics and the process players. A process can be defined as a group of logically related activities and tasks involving people, procedures, and equipment needed to change information and/or materials into a specified product or service. It can include key decisions, handoffs, and conditional situations. Some may ask why a process perspective is important, particularly if they do not believe that their work is part of a process. However, if work produces a result or outcome, that output is attributable to a process and its inputs, regardless of the size, scale, or scope of the work. Furthermore, a common business goal is to be successful. Success translates into driving some sort of result or output. Nevertheless, current or past results alone cannot predict future outcomes or success. The process must be understood to predict future outcomes. A business concerns itself with establishing a plan and targets to drive results, using the Process map to monitor results, to determine progress, and predict its ability to obtain those goals by serving as a checklist or notepad to record data. The results-orientation comprehends that the outputs are a function of the process and its inputs used to produce them. This translates into the mathematical equation Y = f(X), where Y represents the output, and X the key process and its input variables. The Process map graphically displays this Y = f(X) equation. Thus, a Process map often is the starting point for a process improvement initiative. (See Also “Y = f (X),” p. 758)

H I J K L M N O P Q R S T U V W X Y Z

524

A B C D E F G H I J K L M

Encyclopedia

The Process map is a powerful and flexible tool because it applies to a variety of situations from manufacturing, to services; from transactional or administrative to continuous; from a program to a project; from the simple to the complex; and from the known to the planned. Some experts have modified the seven quality control tools (7QC Tools) to include the Process map. The 7QC Tools are viewed as the core quality improvement tools. The original list, attributed to Dr. Kaoru Ishikawa, included: Causeand-Effect diagram; Check sheet (or checklist); Control charts; Histogram; Pareto chart; Scatter diagram; and Stratification. However, more recently, the 7QC Toolset has been revised by substituting the Stratification technique with either a flowchart (or Process map) or a run chart (or Time Series plot). Given the flexible nature of the Process map, it can reflect four different perspectives: 1) What you think the process is, 2) What the process actually is, 3) What the process could be, and 4) What the process should be. Moreover, a complete Process map portrays two dimensions—first, the process parameters that describe the consumables required to complete the step and second, what the process players do. It also includes product parameters, which are the characteristics the tangible item must have to be acceptable—whether it is an actual product or documented information. In conclusion, there are three Process map applications:

N

• High-level Process map

O

• Detailed Process map representing two states:

P Q R S T U V W X Y Z

• As is state • To be state (also known as should be, improved, or desired state)

High-level Process Map A high-level Process map consists of broad categories that describe a set of activities or milestones. If a macro view captures categories, then often only nouns are used as titles. Characteristically, the high-level map contains three to eight items that represent the end-to-end scope. In addition, a high-level Process map captures such an elevated perspective such that it generally remains unchanged, even after a process improvement has occurred. This type of Process map serves as a good communication tool to establish project boundaries and focuses a project improvement team on the scope of interest. Thus, the start and stop points should clearly mark its boundaries to define the project scope. Figure P-42 shows a high-level Process map of completing a successful sale.

Proce ss Map (or Flowchart)—7QC Tool

525

Stop Boundary

Start Boundary

Identify a Prospect

Negotiate Terms

Present Value Proposition

Gain Signature and Close Sale

Successful Sale

A B C D E

Figure P-42: High-level Process Map Example

F

Detailed Process Map This flowchart contains the greatest amount of information. It is used to identify opportunities for improvement and therefore is as complete a picture as possible including the activities (process and sub-process steps), inputs, outputs, metrics, and often roles and responsibilities of the process players. A detailed Process map should only contain as much information as needed. Hence, the appropriate amount of detail depends on its use. If a cross-functional, multi-disciplined management team uses the chart, it may contain fewer details than if it were being used by a function’s process players. Figures P-43 and P-44 exhibit two different types of detailed Process maps for a selling process. Figure P-43 highlights one area within the high-level map and drills down into that step as the area of focus; sometimes called a Top-down flowchart. Figure P-44 provides a complete end-to-end picture of a selling process.

G H I J K L M N O P Q R S

Identify a Prospect

Present Value Proposition

Negotiate Terms

Gain Signature and Close Sale

T U V

Draft Counter Measure

Get Pricing and Terms

Gain Approvals

Gain Customer’s Signatures and Close Sale

W X Y Z

Gain District Manager’s Approvals

Figure P-43: Top-down Process Map Example

Gain HQ Finance’s Approval

526

Encyclopedia Contact Prospect to Schedule Time

Identify a Prospect

Meet Prospect?

Discuss Current Issues or Pain Points

Yes

DecisionMaker?

Identify Potential Needs to Satisfy

Yes

No No Confirm Specific Requirements

A

Qualify Customer Interest

Present Value Proposition

C

Translate Requirements into Supplier Specs

D

H I J K L M N O P Q R S T U V W X Y Z

No

Presentation Approved?

Yes

Present Recommendation to Customer

Negotiation Required?

No

Yes

E G

Draft Proposal with Pricing and Terms

Develop Value Propostition and Presentation

Presentation Approved?

No

B

F

Yes

Contact New Prospect

Manage Ongoing Customer Business

Document Lessons Learned and Recycle

Evaluate Execution and Customer Satifsfaction

Close Order and Gain Customer Signature

Figure P-44: Detailed Process Map Example

As Is State The As Is flowchart documents actual operations (the good, the bad, and the ugly), rather than how it is supposed to operate—the as is; not the to be. Walk the process to ensure the diagram depicts what actually occurs, rather than the procedural documentation. Walking the process can catch hidden factors of rework and do-loops, workarounds, and undocumented contingencies. Thus, the content should be periodically reviewed to reflect any changes. Process changes need to be understood but may not necessarily be bad. They may be adjustments in response to a change and call for a more permanent change in the process or perhaps it could be a kaizen improvement, or positive deviance, that should be communicated broadly and become part of the standard operating procedure. To Be State The To Be Process map depicts the improved state after the process has been leaned of its wastes. Improvements may include elimination of recycle loops (that is, rework and some inspections or approvals, delays and waiting, and excessive handoffs; modification or elimination of activities that produce defects or non-value adds; and modifications to increase flow (that is, streamlining, creating parallel paths and re-sequencing). It clarifies the improvements made to not only the process steps and their sequence (sequential versus parallel), but also potentially the inputs, outputs, metrics, and process players. A powerful communication tool depicts those eliminated activities as crossed-out steps on the current As Is map, as an intermediate detailed Process map. (See Also “Lean and Lean Six Sigma,” in Part I, p. 29, for detail on the different types of waste.)

Proce ss Map (or Flowchart)—7QC Tool

Since the To Be Process map documents a goal or direction for the process players, it becomes a training document and part of the standard operating procedure used in the transition phase to the improved state. Careful monitoring of the process players’ progress and adaptation to the change ensures that the To Be map reflects the new reality, versus merely a “should be” procedure. Upon implementation of a process change, the To Be Process map quickly becomes an As Is map. Hence, the flow should be periodically reviewed to validate adherence and accuracy.

Graphical Layout The flowchart tool, albeit flexible, has conventions to follow with respect to the symbols and layout design. Process maps may be created by hand or by using a variety of software application packages including Microsoft Word, PowerPoint, and Visio.

527

A B C D E F G

Common Symbols Process map conventions for symbols are extensive, as many as 185 shapes. The most commonly used symbols are shown in Figure P-45.

H I J

Process Step or Activity

Initiation or Termination

Decision Diamond

Sometimes Output

Document

Data

Directional Flow of sequence

Input or Output

Process Delay

A

Manual Input

Manual Operation

Predefined Process

Data Store

K L

Merge Preparation Extract

On-Page Reference Connector Links one area to another

Figure P-45: Common Process Map Symbols

Common Layout Configurations Its layout can flow either horizontally or vertically. The individual steps or activities always are labeled with a title or description distinguishing them from the others, typically in a verb-noun combination. Sometimes they can be numbered. The boundary conditions should be clearly identified using the Initiation or Termination symbol identified in Figure P-45. The sequence or direction of the flow must be clearly marked with directional arrows. In addition, traditionally decision diamonds ask binary questions that call for two arrows exiting it to indicate both the affirmative and negative paths (that is, yes/no or good/bad, for example). The layout displayed in Figure P-44 is a classic activity diagram, wherein the focus is primarily activity-centric and often captures decision points, rework loops, and complexity. A deployment diagram or swim-lane Process map is another layout configuration that includes bands depicting functional groups involved in the process. Activities contained within a swim-lane are performed by that specific functional group. As the activity flow crosses these swim-lanes, it connotes a handoff from one party to

M N O P Q R S T U V W X Y Z

528

Encyclopedia

another. This type of diagram is particularly useful for processes involving information flow between people or functions, as it highlights handoff areas, which represent an opportunity for consolidation to eliminate potential waste and improve cycle time. Figure P-46 provides a vertically displayed deployment flowchart example. A

Sales Proces

B

Sales District Manager

Sales Rep

C D E F G

Identify a Prospect

Meet with Prospect

DecisionMaker?

Discuss Current Issues or Plain Points

No Identify Potential Needs to Satisfy

I J

Develop Value Proposition and Presentation

Yes

M

Draft Proposal with Pricing and Terms

N

V W

Z

Proposal Approval? Yes Yes

Yes Negotiation Required?

Yes

Provide Pricing and Terms

Draft Counter Offer

No

Yes

Close Order and Gain Customer Signature

Evalute Execution and Customer Satisfaction

Counter Approved?

Counter Approved?

No

X Y

Draft Approved?

Present Recommendation to Customer

No

Q

U

Provide Pricing and Terms No

O

T

Yes No

L

S

Presentation Approved?

Present Value Proposition

Prospect Interested?

K

R

HQ Finance

Yes

H

P

HQ Pricing

No

Document Lessons Learned and Recycle

Manage Ongoing Customer Business Contact New Prospect

Figure P-46: Deployment Process Map (or Swim-Lane) Example

Proce ss Map (or Flowchart)—7QC Tool

529

How to Use the Tool or Technique Given the flexible nature of a Process map, the intended use should direct how the diagram is built. The following procedures will serve as a guideline when developing a detailed Process map with a process improvement project team. Step 1.

Assemble a representative cross-section of the multiple perspectives of process players and explain the purpose of meeting. a. Meeting preparation: Secure sticky note pads and markers to hand out to each person. The larger sticky notes often are better. You may want to consider segmenting the cross-functional team by color of sticky pad and/or marker. b. Room preparation: Hang blank flip chart sheets around the perimeter of the room to serve as working space for the team to post the individual activities. c. If the high-level Process map is known, label at least one flip chart sheet per each category.

Step 2.

Step 3.

Step 4.

Individually, brainstorm the various activities, documenting one idea per sticky note. Ensure the person places his/her initials in the lower corner to help identify author during the clarification step. (Allow 10 to 20 minutes of this silent activity.) Upon completion of Step 2, invite each person to post the sticky notes on the flip charts around the perimeter of the room.

A B C D E F G H I J K L M N O P Q

a. If high-level Process map is posted, they align the stickies with the appropriate category.

R

b. If flip chart sheets are blank, the team needs to group activities into affinities, and eventually identify and label the high-level categories.

T U

Clarify stickies by reading each one, verifying if clarification is needed, and modify sticky as appropriate.

W

a. If duplications exist, keep only one. b. If gaps exist, create additional stickies to complete the representation. c. If stickies belong in multiple places, create duplicates and put in all the appropriate places. Further clarification and process may determine differences and then document distinguishing characteristics.

S

V X Y Z

530

A

Encyclopedia

Step 5.

Arrange the activities in the proper sequence. Arrange each activity sticky to reflect the current flow until all are accounted for and everyone agrees with the sequence.

Step 6.

Draw flow arrows to indicate direction.

Step 7.

Document any supporting detail necessary and number the activities and person accountable for activity. Identify inputs, outputs metrics, and critical decision points.

B C

a. Inputs and outputs encompass not only those at the beginning and end of the process, but also those within the process (or in-process).

D E F

b. Inputs can be further classified into what is controllable (C), part of standard operating procedure (S), or noise (N)—that which is uncontrollable.

G H

c. Decision points represent questions that lead to multiple paths.

I J K

Step 8.

Review flowchart to ensure completeness and accuracy. Dismiss the meeting.

Step 9.

Document the flowchart electronically. Distribute to the meeting participants and ask them to review and edit the document over a period of time while working in (or walking) the process, and submit any edits back to the document creator for changes.

L M N O P Q R S T U V W X Y Z

a. You may assign “observers” to walk the process and make notations on the Process map draft to look for “subconscious” activities, exceptions and “hidden factories” of rework. b. Sometimes it is insightful to walk the process backward, from the end to the beginning. c. Record or double-check supporting information—person accountable for activity, inputs, outputs, and metrics. d. Upon completion of the Process map, use it as a tool to analyze improvement opportunities. e. After improvements are made, revise and update the Process map to reflect the improved state.

Proce ss Map (or Flowchart)—7QC Tool

531

Hints and Tips Creation process • Develop a Process map with a team of people who work in the process, but designate one person responsible for managing the overall process of mapping the total flow. • Build a process from end-to-end, starting from either the beginning or end and avoiding simply pasting together individual process areas that management built. • Allow time for reflection and edit of a Process map draft to capture potentially forgotten or infrequent activities. • Capture the perspectives from as many different types of process players as possible—the process workers, process suppliers, process customers, process owner(s), and supervisors. • High-level Process maps depict a macro perspective of the overall process categories that remain constant even after an improvement or leaning initiative. Conversely, a detailed Process map’s activities, and sometimes flow, reflect a change after an improvement initiative. • Add as much supporting detail to a detailed flowchart as relevant to the process analysis—cycle time (average and range) and metrics (that is, cost, quality and performance measurements). • Sometimes a “new” set of eyes (objective observer) helps “see” the process and identify the rote activities that are performed unconsciously and often go undocumented. These often may be workarounds in response to a situation to make the process work. Layout • Capture the content and sequence of activities, rather than worry about what is the “right way” to draw it. • Draw the Process map manually before capturing it electronically. • Diagram the sequence of activities that accomplish work to depict their flow, rather than mapping the organization. • Map the appropriate level of detail for the given business question or problem trying to be addressed, rather than every aspect of a process.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

532

Encyclopedia

Analysis • Identify value adding, value enabling, and non-value adding activities throughout the process. A B C D E F G H I J K L M N O

• Use a detailed flowchart as a checklist to ensure consideration of process components end-to-end. • Mark-up a detailed Process map to highlight eliminated steps. • Pose questions (similar to the following suggestions) and document the answers on a detailed Process map: • Who? (supplier, accountable process player, decision-maker, inspector, customer) • What? (item or information, value-add versus non-value-add) • When? (what preceded it, what follows) • Where? (produced, comes from, it goes) • How? (does it flow—how is it being transported, handed off, and how it is being transformed (value add), used by downstream customer) • Why? are things done this way? It serves as a follow-on question for subsequent exploration.

P Q R S T U V W X Y Z

Supporting or Linked Tools Supporting tools that might provide input when developing a Process map include • Brainstorming (See Also “Brainstorming Technique,” p. 168) • Standard operating procedure (SOP) A completed Process map provides input to tools such as • Activity Network Diagram (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127) • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173) • Control plan (See Also “Matrix Diagrams—7M Tool,” p. 399) • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287)

Proce ss Map (or Flowchart)—7QC Tool

533

• SIPOC (See Also “SIPOC (Supplier-Input-Process-OutputCustomer),” p. 663) • Standard operating procedure (SOP) • Training plan • Value Stream Analysis (See Also “Value Stream Analysis,” p. 727) Figure P-47 illustrates the link between a Process map and its related tools and techniques.

A B C D

AND

E F

SOP

Brainstorming Technique

Process Map

Cause and Effect

G H

Control Plan

I

FMEA

K

J L

SIPOC

M N

Training Plan

O P

Value Stream Analysis

Figure P-47: Process Map Tool Linkage

Q R S T

Variations

U V

• Arrow diagram or Activity Network diagram (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127)

W

• SIPOC (See Also “SIPOC (Supplier-Imput-Process-Output-Customer,” p. 663)

Y

• Value Stream map or Process map (See Also “Value Stream Analysis,” p. 727)

X Z

534

Encyclopedia

Pugh Concept Evaluation What Question(s) Does the Tool or Technique Answer? Which design or potential solution option is best? A

The Pugh Concept Evaluation helps you to

B C D E F G H

• Articulate fundamental principles important to the integrity and feasibility of a concept • Gain better insight into the concept requirements, design problems, and potential solutions • Refine design options and select the optimal one based on a common set of criteria from benchmark data

I J K L

Alternative Names and Variations This tool is also known as

M

• Pugh Concept Evaluation and Selection Process

N

• Pugh Concept Selection

O

• Pugh Concept

P Q R S T

• Pugh process Variations on the tool include • Solution Selection matrix (See Also “Solution Selection Matrix,” p. 672)

U V W X Y

When Best to Use the Tool or Technique When there are a set of alternative improvement designs (or potential solutions) to select from, use the Pugh Concept to improve and evaluate the options and ultimately select the winning concept.

Z

Brief Description The Pugh Concept Evaluation process is a structured technique to refine concept alternatives and select the best one using a benchmark. The benchmark represents a best-in-class design, and is referred to as the datum. The datum is defined as the evaluation criteria—the concept that

Pugh Concept Evaluation

the solution alternatives must beat. The datum is the best-in-class, industry showcase for a solution or solution component that is identified through benchmarking, from customers, or from industry analysts. The use of such a benchmark approach simplifies the evaluation comparison process, as opposed to comparing all concepts to all other concepts simultaneously. The approach recommends a cross-functional team to further develop and improve the initial set of concepts. As a result, the team often generates additional new concepts. The team uses a matrix-based process in several two to three hour-long team meetings over a few weeks to converge on a superior design. The process aims to ensure that the final concept best meets the customers’ new, unique, and difficult (NUD) requirements, such that often the final selection is a modification of an originally considered design. NUD requirements translate into • New—A requirement the customer never asked to be fulfilled before; it is completely a fresh requirement. • Unique—A requirement being fulfilled by a competitor or a substitution product or service your organization currently is not providing. • Difficult—A requirement that is very difficult to fulfill (not necessarily new or unique but often is). (See Also the “NUD versus ECO” section within “Kano Model: KJ Analysis,” p. 376) This approach is named after its developer, Stuart Pugh, a professor of Engineering Design at the University of Strathclyde in Glasgow, Scotland. Professor Pugh died in 1993 but was still viewed as a leader in product development and total design methodology. The Total Design concept envelopes seven sources of information: Market Knowledge; Requirement Knowledge; Concept Knowledge; System Knowledge; Detail Design Knowledge; Manufacturing Knowledge; and Sales/Service Knowledge. Pugh’s concepts are taught internationally in many Design for Six Sigma (DFSS) programs. Don Clausing, professor at MIT and one of Pugh’s biggest advocates, continues to popularize Pugh’s concepts through his teachings and writings. The Pugh process, within the Total Design methodology, is a procedural tool for controlled divergence and convergence of the best possible solution to a given set of design requirements. It involves three main phases: 1) the initial phase of Concept Evaluation and Hybridization, 2) the second phase of Concept Evaluation (often final), and 3) the additional phase of Concept Evaluation, as necessary. It requires several inputs that fall into five main categories:

535

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

536

Encyclopedia

• Product Requirements—defined and structured. • Ranking—from Customer Requirements Survey. • Quantitative Technical Requirement Prioritization. A B C

• Competitive Benchmark Data—engineering data on competitive product performance. • Concepts generated to equal levels of expression: • Math models—numeric (and sometimes verbal) description of the concept.

D E

• Graphical models—pictorial (and sometimes verbal) depiction of the concept.

F G

• Physical models—visual display of the concept:

H

• Conduct initial studies in Design For Manufacturing and Assembly (DFMA) and Design Failure Mode and Effect Analysis (DFMEA).

I J K L M N O P Q R S T

• Funding and Resources Development teams using this process often achieve greater insight and awareness of possible solutions by building on and integrating each other’s ideas. This method stimulates the generation of new and better hybrid-design concepts to better meet customers’ NUD requirements. Teams other than product development teams find they benefit from this approach in enhancing and evaluating potential solutions. Regardless of the type of improvement team, its make-up should comprise diverse talent and perspectives working on a common goal to stimulate the creative process and ensure a robust design. The Pugh process draws out and highlights clear distinctions among solution alternatives.

U V W

How to Use the Tool or Technique The Pugh Concept Evaluation technique employs the following procedure:

X

Step 1.

Y Z

Select the decision-criteria used to evaluate the various concept alternatives. a. A good source includes a completed Quality Function

Deployment (QFD) matrix, using the CTQ column (the vertical axis on the far left). (See Also “Quality Function Deployment (QFD),” p. 543)

Pugh Concept Evaluation

Step 2.

Discuss and clarify the different concepts. Try to discuss each concept to a similar level of detail to avoid bias and misunderstanding. Document the concepts with appropriate visuals and verbal descriptors.

Step 3.

Select the benchmark or datum. Consider the technical requirements when selecting the benchmark. Select the datum concept that is among the best concepts available for the baseline. A “datum” concept is selected as the initial comparison point for all the other concepts. a. In the absence of a strong competitive design, select the

best internal design based on the consensus of the team (often this will be the current “best” design). b. A good source includes a completed Quality Function

Deployment (QFD) matrix, referencing the competitive annex. Step 4.

Construct an L-shaped matrix with the vertical axis (left side) for the criteria and the horizontal axis (top row) for the different concepts. (See Also “Matrix Diagrams—7M Tool,” for a discussion on matrix types, p. 399) a. Ensure that all the team members can view the matrix. Construct the matrix one of the following ways: electronically (using a spreadsheet, such as Excel) and project it or on a flip chart paper posted on the walls or whiteboard. b. The first column on the far left should be entitled “Criteria.”

537

A B C D E F G H I J K L M N O P Q

i. In each cell under the title of the first column, docu-

R

ment each decision-criterion, one per cell working downward, to label each row in the matrix.

S

ii. After all the criteria are listed, add three more rows

and label them as follows:

T U V

• Total Pluses

W

• Total Minuses

X

• Total Same c. The second column contains the benchmark information and is labeled “Datum.” d. The remaining columns represent the various alternative concepts under evaluation and should be labeled accordingly. The matrix should reference the appropriate visuals and verbal descriptors for each concept.

Y Z

538

Encyclopedia

Step 5.

Compare each concept relative to the datum using non-numeric rating scale: “+” (a plus-sign) for better than the datum; “–” (a minus-sign) for worse than the datum; an “S” for the same as the datum. This begins the Concept Evaluation and Hybridization phase.

A

a. Rate each concept using the non-numeric rating scale: +, -, and S versus the datum.

B C D

Step 6.

a. Total the score for each concept—the total number of plus, minus, and same scores and record in the appropriate cell.

E F

b. Identify any pattern across the different alternatives. Certain concepts will exhibit distinct strengths; others will display clear patterns of weakness. As the concepts are evaluated, the number may get reduced. Ensure that all concepts are compared on the same basis and at the same generic level.

G H I J K

c. See what the positives will contribute to one’s insight in the design.

L M N O P Q R S T

Evaluate the ratings.

Step 7.

Optimize the design concepts. Attack the negatives and enhance the positives. a. Discuss the most promising concepts. i. Isolate and document weaknesses of the strong concepts. ii. Identify what could be changed to eliminate the negatives.

U

iii. Check that this does not reverse a positive.

V

iv. Identify the positive interactions between concepts, which create new hybrid concepts. Define a new concept from this process and add it to the matrix (divergence).

W X Y Z

v. Leave the existing concept as is; it may be used later. vi. The matrix has now expanded. b. Assess weak concepts. Attack negatives to see if they can be improved relative to the datum.

Pugh Concept Evaluation

539

i. If they can be significantly improved, add this new concept to the matrix (leave the weak concept as it is). ii. Synthesize the strengths of weak concepts that otherwise could not be helped. Add the new synthesized concept to the matrix. iii. Kill or modify the negative ones. iv. The matrix will shrink. c. If a number of strong concepts do not emerge, assess any apparent uniformity of strengths and weaknesses. Step 8.

Select a new datum and “rerun” the matrix. When any one concept persists as uniquely strong, rerun the matrix with that strong concept as the datum. A new hybrid can be entered into the matrix for consideration. a. See if the strength still persists. If it does, this is a confirmation of the initial conclusion. b. If it is beaten, then repeat Step 6 until distinctly strong concepts persist. You are DONE at this point if the decision is made to further develop the strongest concepts from phase 1. i. The strong concepts are put through more develop-

ment and testing. Ensure each concept receives equally balanced attention so as to not bias one over another. ii. A deeper understanding and maturing may require

the development requirements or CTQs to be expanded or refined. Step 9.

Optional: Plan for further work. At the end of the first working session, the team may need to collect more information, perform experiments, and evaluate technical capabilities. This begins the second phase—Concept Evaluation.

Step 10.

Optional: Reconvene the team to refine the concepts and rerun the matrix for further analysis based on the additional information collected. Continue this iterative process until a winning concept beats the datum. a. With an expanded matrix (criteria and concepts), repeat the evaluation steps starting from Step 4 through 7.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

540

Encyclopedia

b. Question the results in detail. Look at new patterns of strengths and weaknesses, especially if criteria changed significantly. The re-affirmation of the trend established in the first phase tends to give confidence in the emerging design Figure P-48 illustrates a generic Pugh Concept Evaluation matrix. A B C D E F G H I

Product Concepts Criteria

A

B

C

D

Evaluation Criteria #1

Benchmark

s

-

+

s

Evaluation Criteria #2

+

-

-

s

Evaluation Criteria #3

D

+

+

+

s

Evaluation Criteria #4

A

-

-

s

-

Evaluation Criteria #5

T

-

-

s

-

Evaluation Criteria #6

U

+

-

+

+

Evaluation Criteria #7

M

+

-

+

+

Total Pluses

4

2

4

2

Total Minuses

2

6

1

2

Total Same

1

0

2

3

J K L M N

Figure P-48: Basic Pugh Concept Evaluation Matrix

O P

Hints and Tips

Q

• Typically, a superior concept is attained by the Convergence Phase

R S T

of the process. • The more the different points of view and perspectives explored the better.

U

• The concept of safety and wisdom coming from the variety of

V

cross-function team members provides additional confidence

W

that vulnerability has been minimized.

X Y Z

• All this work is done on the large matrix in front of the entire team; all actions are clearly evident to each member. • Everyone must scan and understand all concepts to actively participate in the evaluation. Otherwise, some opportunity will be missed. • Try to describe and model all concepts to same level of detail; consider mathematical, graphical, and physical models if appropriate. • Beware of poorly defined criteria, primarily due to the lack of properly understood requirements and poorly developed CTQs.

Pugh Concept Evaluation

541

Ambiguous criteria lead to different interpretations by team members. (See Also “Critical to Quality (CTQ),” p. 242) • Purify, redefine, or eliminate the poor criterion. • Sometimes they are eliminated during the process (called a run). • Persistence of uniformity of strengths across concept alternatives usually means that one or more concepts are subsets of the others (essentially the same). • Take time to understand the feasibility of each concept alternative, as well as their associated risks and consequences. • Be careful that the brainstorming portion of the process is taken seriously and with responsibility to avoid perception of simply a freewheeling process. • Introduce a facilitator to help balance the input from team members, if needed. Strong-willed team members (usually with a lot of experience whose design has fallen) may begin to become leery of the emerging dominant concept. They may suggest that early concepts were better than this new emerging dominant concept. Their defense may be based on emotion, experience, bluster, and outright denial of facts. However, if they can be won over using the process, the final design concept may be stronger.

A B C D E F G H I J K L M N O P Q R

Supporting or Linked Tools Supporting tools that might provide input when developing a Pugh Concept Evaluation matrix include

S T U V

• Benchmarking (See Also “Benchmarking,” p. 160 and “Benchmarking— Avoid Arrogance and Lethargy,” in Part III, p. 789)

W

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

X

• Real Win Worth (RWW) (“See Also “Real-Win-Worth (RWW) Analysis,” p. 560) • QFD (See Also “Quality Function Deployment (QFD),” p. 543) • SWOT Analysis (See Also “SWOT (Strengths-WeaknessesOpportunities-Threats),” p. 699) • VOC gathering techniques (See Also “Voice of Customer Gathering Techniques,” p. 737)

Y Z

542

Encyclopedia

A completed Pugh Concept Evaluation matrix diagram provides input to tools such as • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) A B C D E

• Pilots • Transition and Control plans (See Also “Matrix Diagrams—7M Tool,” p. 399) Figure P-49 illustrates the link between a Pugh Concept Evaluation matrix and its related tools and techniques.

F G

Benchmarking

H I

Brainstorming

FMEA

J K L

RWW

M N

Pilots Pugh Concept Evaluation

QFD

Transition Plan

SWOT

Control Plan

O P Q R S T U

VOC

Figure P-49: Pugh Concept Evaluation Tool Linkage

V W X

Additional Resources or References

Y

• Pugh, Stuart. Total Design; Massachusetts: Addison Wesley (1990).

Z

• Creating Innovative Products Using Total Design; edited by Don Clausing & Ron Andrade; Massachusetts: Addison Wesley (1996) ISBN: 0-201-63485-6.

Variations Solution Selection matrix (See Also “Solution Selection Matrix,” p. 672)

Quality Function Deployment (QFD)

543

Q Quality Function Deployment (QFD) A

What Question(s) Does the Tool or Technique Answer? Which technical specifications for a product or services offering best meet a specific set of customer requirements? QFD helps you to • Translate customer requirements into specific offering specifications. • Prioritize possible offering specifications and make trade-off

decisions based on weighted customer requirements and ranked competitive assessment.

B C D E F G H I J K

Alternative Names and Variations This tool is also known as • House of Quality (HOQ) • QFD matrix • Requirement matrix

L M N O P Q R

When Best to Use the Tool or Technique Use the QFD when you are designing a product or services offering and need to select among several alternative features and functionality.

S T U V

Brief Description The Quality Function Deployment is a powerful prioritization tool that combines several different types of matrices into one to form a house-like structure. Sometimes referred to as a House of Quality (HOQ), this tool captures the Voice of the Customer to identify the required quality, features, and functions needed to be deployed in a single offering. The QFD process uses a graphical format to document the information gathered and processed. In its most complete state, this information constructs the five “rooms” in the QFD house, as illustrated in Figure Q-1:

W X Y Z

544

Encyclopedia • The What or Wants—Customer requirements, needs, and priorities

that form the far left-wing of the house. • The Competitive Assessment—Compares customer priorities with

appropriate marketplace offerings, across key competitive deployments, which forms the right-wing annex of the house. A B

• The How—The offering’s technical design features, functionality, and

C

characteristics to meet the customer requirements. This forms the attic of the house.

D

• The Design Relationships—Describes the interrelationship between

E

the design features, which form the roof of the house.

F

• The Foundation—Uses benchmarked target values as objective meas-

G

urements to evaluate each characteristic, forming the basement of the house.

H I J

Design Features Interrelationships (Design Relationships)

K L

Design Features, Functionality and Characteristics (How)

M N O P Q

Competitive Comparison of Custom er Priorities (Competitive Assessment)

Custom er Requirem ents (What or Wants)

R Benchm arked Target Values (Foundation)

S T U

Figure Q-1: House Schematic of QFD

V W X Y Z

The QFD is a flexible tool, customizable for a given situation. The team using the technique can modify the guidelines that alter the scales used in the house. However, the basics of the QFD remain constant—the VOC selects the features and functionality of an offering. The tool was first used to design an oil tanker at the Kobe shipyards of Japan in 1972 by Yoji Akao and Shigeru Mizuno to design customer satisfaction into an offering before it is produced. Prior to this, quality control methods were primarily aimed at fixing a problem during or after production. In the mid-1980s, Don Clausing of MIT introduced this design tool to the United States. A classic product design application is in the automotive industry. In fact, Clausing tells of an engineer who initially wanted to place the

Quality Function Deployment (QFD)

545

emergency hand brake of a sports car between the seat and the door. However, VOC testing found that women drivers wearing skirts had difficulty with the new placement of the hand brake. The QFD highlighted potential dissatisfaction with the location of this feature, and the idea was scrapped. A prerequisite to building a QFD entails collecting customer requirements. Several VOC gathering techniques include surveys, focus groups, interviews, trade shows, and hot lines. Once the VOC data has been synthesized with the customers’ rankings, the QFD process is ready to begin. (See Also “Voice of Customer Gathering Techniques,” p. 737)

Progress Drilldown Approach The primary QFD component features the relationship matrix between the VOC and the potential design options. Once the design options are prioritized in a QFD, the design team can continue to drill down into more specific detail, starting with the major features to subcomponents, down to its component parts to create three separate QFD matrices. To do so, rotate the QFD 90-degrees to the left (counterclockwise), convert the prioritized attic features (the how) into the vertical axis (the what), and list the new more detailed technical specifications as the new attic items. This drill-down process continues until all the technical specification details are prioritized. Design engineers typically start this progressive drill-down approach with planning the development and go through four phases to reach a deeper understanding of the required process control and quality. This four-phase approach produces separate QFD matrices, as shown in Figure Q-2, and includes the following sequence: QFD 1.

QFD 2.

QFD 3.

QFD 4.

Product Planning QFD—Identifies the offering characteristics that best meets customer requirements, analyzes competitive opportunities, and establishes critical target values. Figure Q-9 illustrates this QFD matrix. Component Deployment QFD—Identifies the critical parts and assembly components using the prioritized offering characteristics in QFD 1 and establishes critical target values. Process Planning QFD—Determines critical process operational requirements and elements using the prioritized components in QFD 2 and establishes critical process parameters. Quality Control QFD—Prioritizes the process control methods and parameters and establishes production and inspection methods that best support the prioritized process elements of QFD 3.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

546

Encyclopedia

Trad eo ffs

Trad eo ffs

Co m p o n en t Ch arac teris tic s (Ho w )

Tec h n ic al Ch arac teris tic s (Ho w )

A B

Cu s to m er Req u irem en ts (Wan ts o r Wh at)

C

Co m p etitiv e Co m p aris o n o f Cu s to m er Prio rities

Tec h n ic al Ch arac teris tic s (Wh at)

Targ et Valu es (Fo u n d atio n )

B en c h m ark ed Targ et Valu es (Fo u n d atio n )

D E

QFD 1. Product Planning

QFD 2. Component Deployment

Pro c es s Co m p o n en ts (Ho w )

Qu ality Co n tro l (Ho w )

F G H I J

Co m p o n en t Ch arac teris tic s (Wh at)

Pro c es s Ch arac teris tic s (Wh at)

K L

Pro c es s Param eters (Fo u n d atio n )

M

QFD 3. Process Planning

N

QFD 4. Quality Control

Figure Q-2: Four-phase Progressive Drill-down QFD Methodology

O P Q R S T U V W X Y Z

How to Use the Tool or Technique The procedure to construct a single five-room QFD matrix is as follows. The guidelines describe how to develop each room, its purpose, and how the rooms relate to one another. Preparation: Assemble the VOC data and a crossfunctional team responsible for designing the offering. Create a blank QFD matrix template as show in Figure Q-1. (Typically use a spreadsheet software application such as Microsoft Excel.) Step 1.

Customer Requirements

Customer Requirements and Priorities (What and Wants): a. Document the customer requirements in the left-wing column of the house, which will serve as row headings as shown in Figure Q-3. b. Adjacent to the customer requirements column, insert a column to

Figure Q-3: QFD Construction—Step 1.a.: Customer Requirements

Quality Function Deployment (QFD)

547

document the customer rankings (or priorities) of each requirement relative to one another as shown in Figure Q-5. Use an ordinal weighting scale method to represent the customer’s importance ranking of their requirements. Ask the customers to stack rank the requirements during the VOC gathering initiative, avoiding “tie” rankings to better distinguish one need or want from another.

A B C

Note

D

The highest number reflects the most important requirement.

E F

Ensure that the customers provide the weightings; they’re not from internal judgments. c. Competitive Assessment. In the far right-wing annex of the house, document the customer’s perception of your company’s offering(s) and similar competitive offerings, which serves as part of your competitive assessment. A source for this data may include customer satisfaction surveys, but also industry analysts and industry associations—and some benchmark data may provide the needed information. Use a scale from 1 to 5, with 5 as the best performer and 1 as the worst. List the scale as column headings at the top of the right-wing annex. Determine a set of icons that represent your company and the competition. Figure Q-4 provides example icons.

G H I J K L M N O P

Competitive Assessment Icons = our company

Q R S

= Competitor 1

T

= Competitor 2

U

= Competitor 3

V

Figure Q-4: QFD Competitive Assessment Sample Icons

Place the appropriate company symbols aligned in the appropriate rating column to represent the customer’s perception of the offerings from your company and the competition.

W X Y Z

548

Encyclopedia

Figure Q-5 reflects the customer priorities for both their requirements (placed in the left-most gray column) and competitive assessment (in the right-most gray column) of the QFD. The black-shading indicates the QFD region already discussed.

A B C D

Step 2.

Technical Design Features, Functionality, and Characteristics (How):

E

a. Confirm the list of various design options (features, functionality, and characteristics) and refine it to ensure each directly affects a customer requirement and that each is described in terminology that is meaningful and actionable for your organization.

F G H I J K

b. In the attic, list the various design options across the top to serve as column headings for the matrix underneath in the main part of the house, as illustrated in Figure Q-6.

L M N O P

Customer Priorities

Figure Q-5: QFD Construction—Step 1.b.: Customer Priorities

Offering Technical Characteristics

Figure Q-6: QFD

Construction—Step 2.b. and c. Using the relationship scale of c.: Technical Characteristics 9, 3, 1, 0 (with 9 as the strongest relationship and 0 as no relationship), rate the strength of each technical design option relationship with each customer requirement and record the rating in the appropriate cell.

Q R S T U

d. Optional: Some QFD matrices insert a row just under the design characteristics to indicate the deployment direction of each item. Directionality defines how the requirement is best fulfilled, such that the options included hit the target requirement, or represent more or less than the targeted needs.

V W X Y Z

Hint Typically, arrows indicate the directional relationship to the target: up-arrow signifies bigger is better; down-arrow represents smaller is better; and a zero or empty cell symbolizes that nominal is best.

Quality Function Deployment (QFD)

Step 3.

549

Prioritized Design Characteristics a. Create a row just beneath the last customer requirement and label it “Total Ratings.” b. Working down one Design Feature column at a time, calculate its total prioritized score. To calculate this score, multiply each cell’s relationship rating times the customer requirement weighting. Sum the arithmetic product in each cell for that entire column and record the total in appropriate cell of the Total Ratings row. c. Repeat Step 3.b. until the Total Ratings are calculated for each of the Design Feature columns. d. Optional: Sometimes the relative weighting of the Total Ratings better communicates the relative impact of one Design Characteristic over another. To calculate the Relative Rating, sum the scores of all the Total Ratings and then divide each Design Feature’s Total Rating Score by the sum of all the Total Ratings. Record the appropriate proportion in each cell.

Step 4.

Design Relationships (between features) a. In the roof of the house, define the interrelationship or correlation between two different characteristics. This portion of the house sometimes is referred to as the Correlation matrix. If a pair of characteristics support each other, such that as one increases so does the second to meet customer requirements, then the pair has a positive relationship. If one increases and the other decreases to meet customer requirements, then they have a negative relationship. Negative relationships indicate requirement conflicts that are causes for concern. b. Use a “P” to indicate a positive correlation, and an “N” for a negative correlation. c. Rate each pair of technical features, functions, and characteristics and record the correlation rating in the appropriate cell in the roof. Take time to ensure that the proper two characteristics line up with the rating, given that the angularity of the roof may introduce alignment difficulty. Figure Q-7 illustrates this step. d. Optional: Total Correlation Ratings #1—If the design feature interrelationships have an impact on overall prioritization of the feature, the correlation rating should be

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

550

Encyclopedia

factored into the overall rating. i. Calculate the impact of the correlation. Translate these letters into a numeric score, with P equaling a +3 and N equaling a -3.

A B C D E F G

Design Relationships

P N

P = positive N = negativve

Figure Q-7: QFD Construction—Step 4: Design Relationships

ii. Create two rows just beneath the Total Ratings row and label the top row as the “Total Correlation Rating” and the last row as “Aggregate Rating.”

H I J

iii. Sum each of the correlations scores for each Design Characteristic and record the total in the appropriate cell in the Total Correlation Rating row.

K L

iv. Working by column, multiply the Total Correlation Rating times the Total Rating and record the arithmetic product in the appropriate cell of the Aggregate Rating row.

M N O P

Repeat this sub-step until the Total Correlation Rating and Aggregate Rating scores are calculated for each of the Design Feature columns.

Q R S

e. Optional: Total Correlation Ratings #2—If several pairs of characteristics have relationships with one another and more granularity is required, replace the P and N rating with the +3 value with a new scale.

T U V

Use the scale of + or - 9, 3, 1, 0 to indicate the strength of the correlation, with +9 indicating a strong positive or negative relationship and 0 signifying no relationship.

W X Y

If this option is selected, follow the procedure outlined in Step 4.d. but use the +9, 3, 1, 0 numeric scale rather than the +3.

Z

Step 5.

The Foundation (target values) a. The basement provides the foundation of the house of quality with objective measurements to evaluate each characteristic. Create a row at the bottom of the matrix (below the rating rows—Total Rating, Total Correlation

Quality Function Deployment (QFD)

551

Rating, and Aggregate Rating). Label this new row “Measurement Unit.” Quality Target b. Record the appropriate Values unit of measure for each Technical Design Characteristic. If the target Figure Q-8: QFD Construction—Step 5: value is known, record Target Values that number with the unit. Figure Q-8 illustrates Step 5.

Step 6.

A B C D E

Summary

F

Synthesize all the information in the QFD matrix to decide the appropriate strategy and prioritization of design characteristics. Evaluate the competitive landscape and determine how the relative market positioning in the right-wing annex impacts the overall strategy and selection of features. Assess the required investment and determine whether it aligns with the business goals of the organization.

G H I J K L M

Warning

N

The design features yielding the highest Total Rating may not neces-

O

sarily represent the best set of characteristics. Look at the QFD in

P

total and understand the interrelationships of each room in context

Q

with the overall organizational objective.

R S

Figure Q-9 shows a completed simple QFD matrix for a technology/professional consulting services business, comparing one company to three of its competitors.

T U V W

Hints and Tips

X

A QFD structure can be as simple or as complex as the situation

Y

requires. When assembling a team to build a QFD, ensure that the

Z

team is cross-functional, with multiple perspectives represented. Building a QFD rarely can be completed in one meeting. It takes several meetings, often over several weeks and even months to complete partially because of the input data required and partially due to the discussion needed to reach consensus on the relationship ratings.

552

Encyclopedia P N

H

Knowledgeable People 10 Deliver On Time 7 Deliver Within Budget 6 Provide good documentation 3

9

9 1

9 1

3 1

1 9

3 1

69

69

27

9

3 3 3

3 1

Worst Deployment

9

Average Deployment

Reps with Price Flexibility

G

Customer demands

Best Deployment

Sales Rep also Prjoject Manager

F

Project-based Pricing

E

Services Literature

D

Importance

C

Sales Rep also Cnosultant

B

Mixed Skills on Project Team

A

Semi-Annual Training

Time and Materials Pricing

N

Design requirements

3 1

I J Total Rating

K

N

39

51

111

= our company = Competitor 1

L M

123 111

Relative Rating

600 0.21 0.19 0.07 0.09 0.12 0.12 0.05 0.19

= Competitor 2 = Competitor 3

Figure Q-9: Simple Example of a QFD Matrix

O P Q R S

Supporting or Linked Tools Supporting tools that might provide input when developing a QFD matrix include

T

• Benchmarking (See Also “Benchmarking,” p. 160 and the article,

U

“Benchmarking—Avoid Arrogance and Lethargy,” in Part III, p. 789)

V W

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

X

• Data Collection Matrix (See Also “Data Collection Matrix,” p. 248)

Y

• Product Planning matrix (See Also “Matrix Diagrams—7M Tool,”

Z

and “Solution Selection Matrix,” p. 399 and p. 672, respectively.) • SWOT Analysis (See Also “SWOT (Strengths-Weaknesses-

Opportunities-Threats,” p. 699) • VOC data (See Also “Voice of Customer Gathering Techniques,”

p. 737)

Quality Function Deployment (QFD)

553

A completed QFD matrix provides input to tools such as • Control Plan (See Also “Matrix Diagrams—7M Tool,” p. 399) • DOE (See Also “Design of Experiment (DOE),” p. 250) • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),”

p. 287)

A B

• Project Planning tools for development (See Also “Activity Network

C

Diagram (AND)—7M Tool,” p. 127, and “PERT (Program Evaluation and Review Technique) Chart,” p. 453)

D

Figure Q-10 illustrates the link between a QFD matrix and its related tools and techniques.

F

E G H

Competitive Bencharking

I J Control Plan

Brainstorming

K L

Data Collection Matrix

DOE

N

QFD Matrix Solution Selection Matrix

M O

FMEA

P Q

SWOT

Project Planning Tools

VOC

R S T U

Figure Q-10: QFD Tool Linkage

V W X

Additional Resources or References Cohen, Lou. How to Make QFD Work for You. Massachusetts: Addison Wesley Longman, Inc., 1995. ISBN: 0-201-63330-2.

Y Z

554

Encyclopedia

R A

RACI Matrix (Responsible, Accountable, Consulted, Informed)

B C D E F G

What Question(s) Does the Tool or Technique Answer? Who is responsible for what? An RACI matrix helps you to • Identify who is involved in producing a task and/or deliverable

H

• Communicate responsibilities

I

• Identify any gaps or redundancies associated with people’s respon-

J

sibilities

K L M N O P

Alternative Names and Variations This tool is also known as • Responsibility chart • Roles and Responsibilities chart

Q R S T

When Best to Use the Tool or Technique To organize a project team or document the responsibilities of key process stakeholders with respect to deliverables or tasks. Use to document current state or improved state of a process.

U V W X Y Z

Brief Description Responsible, Accountable, Consulted, and Informed (RACI) is a straightforward matrix tool that identifies who needs to be involved in a task or activity and/or who is to produce a deliverable or output. The RACI tool uses a simple L-matrix structure and may contain either all tasks, all deliverables, or a blend of the two. (See Also “Matrix Diagrams—7M Tool,” p. 399 for detail on different matrix structures.) The RACI applies to two scenarios—a project team and the process players. When applied to a project team, the RACI designates a person by name. The RACI tool acknowledges the short-term nature of a project by using actual people’s names to make assignments. The RACI in this context serves as a good recruiting tool to request a particular individual

RACI Matrix (Re sponsible, Accountable, Consulted, Informed)

555

be assigned to the project team. By specifying a project-specific deliverable, selecting the optimal person with the appropriate expertise, knowledge, or functional background needed to produce that output becomes easier. When applied to a process, the RACI specifies a job title, role, or job function associated with the process. This generic work assignment allows for individuals to be interchangeable while keeping the job description’s responsibilities current. The RACI in this context defines the roles and responsibilities of the process players in each step of the workflow (task and/or deliverable). It describes either the current or the improved state of a process. The RACI serves as a good change management tool used to communicate and compare the pre- versus postimprovement conditions. The RACI uses common language but has a specific definition of its terms: • Responsible (R)—This person or role performs as the “worker bee.”

For any one given task or deliverable, it may have multiple people assigned to be responsible for it. These are the people required to do the work to produce the output or complete the task. • Accountable (A)—This represents the one person to whom the

responsible people are accountable. The accountable person is the single individual who must answer if the task is incomplete or fails to meet requirements. Any given task or deliverable may have only ONE accountable person assigned to it. The accountable person may also be a part of the other roles—responsible, consulted, or informed. • Consulted (C)—This person is known as the subject matter expert.

A B C D E F G H I J K L M N O P Q R

The consulted person possesses unique information or capabilities necessary to complete work. This person provides input—an information supplier. This role may or may not be required for any given task or deliverable.

U

• Informed (I)—This person needs to be notified of results and rarely

V

needs to be consulted. Often this person represents the customer of a task or output either directly or downstream in the process. The informed person also may be a key stakeholder who needs to be apprised of the task completion. This role may or may not be required for any given task or deliverable. The RACI tool is a powerful and flexible communication document. Project-based RACI sometimes expands the structure to include an additional column for due dates of the deliverables. Process-based RACI sometimes expands the structure to include an additional column for the deliverables’ metrics. Figure R-1 illustrates a generic template for both a project-based and process-based RACI.

S T

W X Y Z

556

Encyclopedia Project-based RACI Date:

D

Person H

C = Consult I = Inform

Person G

Due Date

Person E

C

Deliverable 1

Person D

Deliverable/Task Description

Actual Completed Date

Person C

B

(Not started, WIP/OK, WIP/Late, Completed)

Person B

Status

A

Person A

R = Responsible A = Accountable (only 1 entry)

Responsibility Matrix

Person F

Project Name:

By Individual Name

Deliverable 2 Deliverable 3

E F G

Process-based RACI

H

Project Name:

M

Deliverable 1

N

Deliverable 2

Role H

Role G

Metric

C = Consult I = Inform

Role F

Deliverable/Task Description

Role E

L

Role D

K

Role C

R = Responsible A = Accountable (only 1 entry)

Responsibility Matrix

Role B

J

Role A

I

Optional Region

Date:

By Process Player Role

Deliverable 3

O P

Figure R-1: Project Versus Process RACI Matrix Template

Q R S U

How to Use the Tool or Technique Regardless if the RACI applies to a project or process, build it using the following procedure:

V

Step 1.

Construct an L-shaped matrix. This can be done manually or with application software such as Microsoft Excel or the table functions in Word or PowerPoint.

Step 2.

Identify the tasks or deliverables required and list them down the far-left column, the vertical axis, to set them up as row headings.

T

W X Y Z

a. If constructing a task-based RACI, list the activities in the sequence that occur in the process. b. A Process map provides good input for this step.

RACI Matrix (Re sponsible, Accountable, Consulted, Informed)

Step 3.

557

Insert a blank row at the top of the matrix. Identify the job functions involved in the process and list them across the top row, the horizontal axis, as the column headings. a. If this is a project-based RACI, substitute the roles with the project team members’ names. b. A Process map provides good input for this step.

Step 4.

Working one row at a time, identify the various RACI roles for each task or deliverable and document the assignment in the corresponding cell. a. Repeat this step until all the rows have been addressed. b. Ensure that an accountability assignment appears in every row. c. Review the matrix for accuracy.

B C D E F G H I

Figure R-2 shows a project-based RACI, and Figure R-3 exhibits a process-based RACI.

J K L

Project: Sales Force Effectiveness

RACI Matrix

A

R = Responsible C = Consult A = Accountable (only 1 entry) I = Inform

M N

Sales pipline data

Figure R-2: Project-based RACI Matrix

Corbin

Garret

Monica

Lauren

Q

R

R

I

C

C

R

C

R

AR

Sales performance worldwide for 3 years Basic Sales Training Curriculum

Diane

A

Nancy

Field Sales Job Descriptions

P

Bill

Deliverable/Task Description

Mark

O

C AR R

R

I

C

A

R C

S R

R

R

T U V W X Y Z

558

Encyclopedia Project: Sales Force Effectiveness

D E F G

Identify a prospect

AR I/C

I

Meet prospect

AR I/C

I

Discuss current issues / pain points

AR I/C

Determine if Decision-maker

AR

Identify potential needs to satisfy

AR I/C

H

Develop value proposition and presentation AR I/C

I

Approve presentation

J

Customer Admin

I

Validate prospect’s interest

AR

I

R

I

I/C I/C R

A

C

I

I

I/C I/C R

A

C

I

L

Identify any special contract terms

AR

M

Draft proposal

AR I/C

I

I

I

Approve proposal

I

R

A

I

C

Inspect proposal

R

R

A

R

R I/C

A

R

R I/C C

I

I

I

O

Present recommendation to prospect

P

Negotiate any special conditions

R

R

Gain approval signatures

R

A

AR

I

I

I

AR

I

Close order

R

Evaluate customer satisfaction

S

Document lessons learned

T

Legal

AR

Determine pricing

Q

HQ Finance

R I/C

Present value proposition

K

N

HQ Pricing Analyst

Deliverable/Task Description

District Controller

C

District Manager

B

Sales Rep

A

Sales Manager

RACI Matrix

R = Responsible C = Consult A = Accountable (only 1 entry) I = Inform

AR I/C

R

I I

I

I

I

R AR

Figure R-3: Process-based RACI Matrix

U V W X Y Z

Hints and Tips • One accountable person (or role) must be assigned to every task or

deliverable. • Empty cells are acceptable for a given deliverable or task, as long

as the accountability is assigned. • Multiple responsible, consulted, and informed roles may be

assigned for any given task or deliverable.

RACI Matrix (Re sponsible, Accountable, Consulted, Informed)

559

• Multiple combinations may occur for a given person (or role) with an

assigned task or deliverable (for example Accountable/Responsible (A/R), Consulted/Informed (C/I), Accountable/Informed (A/I)). • A project-based RACI should identify individuals to the assign-

ments rather than job titles, roles, or functions. • A process-based RACI should identify job titles, roles, or functions

rather than individuals’ names to the assignments. • The list of assignments may be 100% tasks, 100% deliverables, or

a blend of tasks and deliverables. • Reference a Process map to build either a task-based, deliverables-

based, or hybrid RACI. • When analyzing an RACI for improvement opportunities, look for

any redundancies or gaps. Resolve the issue and revise the RACI and other supporting documentation. • Adapt the structure with additional columns to expand its purpose,

for example • Target completion date (for a given task and/or deliverable) • Progress to-date (for a given task and/or deliverable) • Completion date (for a given task and/or deliverable) • Metric (for a given output)

A B C D E F G H I J K L M N O P Q R

Supporting or Linked Tools Supporting tools that might provide input when developing an RACI matrix include • Brainstorming (See Also “Brainstorming Technique,” p. 168) • Process map (See Also “Process Map (or Flowchart)—7QC Tool,”

p. 522) A completed RACI matrix provides input to tools such as • Control plan (See Also “Matrix Diagrams—7M Tool,” p. 399) • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) • Process map (See Also “Process Map (or Flowchart)—7QC Tool,”

p. 522)

S T U V W X Y Z

560

Encyclopedia

Figure R-4 illustrates the link between an RACI matrix and its related tools and techniques.

Brainstorming

Control Plan

A B C D E F G

Process Map (for a processRACI)

RACI

Project Plan (for a project RACI)

FMEA

Process Map

Figure R-4: RACI Matrix Tool Linkage

H I J

Real-Win-Worth (RWW) Analysis

K

N

What Question(s) Does the Tool or Technique Answer? How would a potential offering be valued in the marketplace and positioned well against competition? Is it worth the investment to develop the idea? Would it be successful? Would the concept outpace competition?

O

An RWW analysis helps you to

L M

P Q R S T U

• Evaluate candidate offering elements by using a numerical ranking

and risk characterization technique • Determine if the potential offering can win in the competitive envi-

ronment and if it is worth the investment of time and resources to develop

V W X Y Z

When Best to Use the Tool or Technique The Real-Win-Worth technique is best used during both the portfolio refresh process and the offering development and commercialization preparation process. During the portfolio renewal process, RWW evaluates whether an offering concept should be funded and activated as a development project. It can examine several potential portfolio concepts and prioritize them in terms of risk to help determine the best candidate(s) for investment. The RWW tool carries over from the strategic portfolio renewal process into the tactical offering development process. The project team

Real-Win-Worth (RWW) Analysis

updates the RWW analysis to continue to evaluate marketplace changes and update with any new information. The technique helps to understand whether the marketplace would value the offering, how it would fare against competition, and what related potential risks might be. This evergreen process continues throughout the time the offering is being designed, developed, and prepared for launch. (See Also “Six Sigma for Marketing (SSFM),” in Part I for additional detail in the sections on the IDEA and UAPL methods, p. 73 and p. 81, respectively.)

561

A B C

Brief Description The RWW technique is a judgment enhancement technique that helps teams make choices. It identifies those markets and opportunity characteristics that make investing in new elements of an offering portfolio viable. It prioritizes and identifies the RWW elements to be deployed as part of the portfolio renewal process and defines the appropriate timing. The RWW technique also aids product development teams to understand how the marketplace values a specific offering. RWW examines the offering (or offering component) relative to competition. It weights the investment needs of time and resources against the opportunity. The technique comprises three components, each represented by its own matrix, to answer one of the following three RWW questions: 1. What are the Real market and opportunities? 2. Where can we Win? 3. Is it Worth our investment? The Real market is defined by data that characterizes its purchasing capacity (or size), the purchasing dynamics (or growth), and the purchasing capability (or demand). The Real opportunities disclose whether or not the team gathered data that clearly defines the market need, if the concept aligns with the market need, and if the idea aligns with the firm’s core competencies. Industry analysts and market segment associations can provide this data. The ability to Win is predicated on two dimensions—the competitive environment and the firm’s capabilities. RWW takes into account the timing, technologies and performance, pricing, brand identity, and market share to evaluate the competitive landscape. The firm’s business capacity and capability to compete in this marketplace involve examining its core competencies in technology, design, production, and services. In addition, the firm’s process and operational efficiencies are examined. Last, the firm’s resources are evaluated, which include money, space, facilities, and equipment.

D E F G H I J K L M N O P Q R S T U V W X Y Z

562

A B C D E

Encyclopedia

Whether the concept is Worth investment involves support from the firm’s financial community to determine the potential Net Present Value (NPV), its potential Internal Rate of Return (IRR), and proposed pricing and costs scenarios. Opportunity costs should be explored to understand the required versus the available budget. The team assesses if the potential offering financials align with the strategic financial goals of the firm (or division). And the team evaluates the opportunity’s risk relative to the other opportunities under consideration.(See Also “Cost/Benefit Analysis,” p. 238, for a discussion on NPV and IRR.) Before getting started, the project team needs to collect the following inputs:

F

1. Market, segment, and opportunity data

G H

2. Competitive benchmarking; market trend; and behavioral analysis data

I

3. Economic business case forecasts

J

4. Business strategy, financial goals, and core competencies

K L M N O P Q R S T U V W X Y Z

(See Also “Voice of Customer Gathering Techniques,” p. 737; “Benchmarking,” p. 160; the article, “Benchmarking—Avoid Arrogance and Lethargy,” in Part III, p. 789; “Market Perceived Quality Profile (MPQP),” p. 390; and “SWOT (Strengths-Weaknesses-Opportunities-Threats,” p. 699) The power of the RWW technique becomes evident as the project team updates the tool. As portfolio renewal deliverables are refreshed, identifying and answering key RWW questions get updated with increasing clarity of data. As the team moves from the portfolio renewal process into the product development process, the RWW technique encourages the team to customize and add to the depth and number of questions asked by the tool. A well-implemented RWW analysis is iterative as an offering moves from concept to launch-readiness. The RWW outputs include 1. Ranked values associated with each Real-Win-Worth criteria. 2. Ongoing development of data that builds insight on a phase-byphase basis from portfolio renewal to offering development and commercialization. 3. Prioritized weighting of a concept risk and the worthiness of its investment needs against a real opportunity.

Real-Win-Worth (RWW) Analysis

563

How to Use the Tool or Technique Presuming that the first application of the RWW technique is in the portfolio renewal process, the project team needs to gather the appropriate input data on the marketplace, competitive environment, and internal goals and capabilities, as just mentioned. For first-time users, a set of example RWW questions are provided in the following list to get started. However, the project team should customize the questions to better suit its unique business model and marketplace. Develop the RWW questions over time and across projects to translate the RWW questions for your own business and culture.

D

Step 1.

Create three “Scorecard” matrices to assess each RWW dimension.

E

Create a set of three matrices, one for each of the three RWW characteristics: Real market and opportunities, likelihood to Win, and if the concept is Worth the investment.

G

This might best be completed in Microsoft Excel, to make for easy computation and navigation among the three matrices. Establish each Excel tab as a unique matrix—Tab 1 for Real; Tab 2 for Win; Tab 3 for Worth. Construct each matrix with three columns. The far left column (column 1) lists the RWW questions; the middle column (column 2) contains the ranking values; and the far right column (column 3) is reserved for comments. Step 2.

A B C

F H I J K L M N O P

Record the questions for each RWW characteristic.

Q

Start with the baseline RWW questions and examine them by category—Real, Win, and Worth. Based on your business model, edit the baseline questions as appropriate. Add any additional questions to define the final set of RWW questions.

R S T U

Record each individual question on its respective RWW matrix, placing one question per matrix cell within the far left column (column 1). Each question serves as a row heading.

W

a. Matrix 1: The REAL market and opportunity questions:

X

i. Have we defined a legitimate market and segment

arena in which to participate? ii. Have we quantified the size of the market? iii. Have we quantified the market growth rate? iv. Have we quantified the purchasing capacity and

capability of the market players?

V

Y Z

564

Encyclopedia v. Have we defined clear market trends and customer

needs? vi. Do these market trends and needs align with our

ideas and capabilities (for example, technologies)? A B C D E F G H I J K L M N O

vii. Does the market have a strategic fit with our

business strategy for organic growth and core competencies? viii. Does the market have a strategic fit with our busi-

ness strategy for external partnerships, joint ventures, and acquisitions? b. Matrix 2: Can we WIN questions: i. Can we meet the launch cycle-time requirements to

be first or early in the market? ii. Do we have technology and performance superiority

relative to our competition? iii. Can we establish competitive pricing, and effectively

win against competition? iv. Is our brand identity competitive? v. Can we gain market share?

P

vi. Are our core competencies enabling us to compete?

Q

vii. Are our operational efficiencies enabling us to

R S T U V W X Y Z

compete? viii. Do we have enough resources to compete in this

market? c. Matrix 3: Is it WORTH the investment questions: i. Is the potential NPV worth the investment? ii. Is the risk-adjusted NPV (ECV) worth the

investment? iii. Does the Internal Rate of Return (IRR) potential meet

our standards? iv. Can we balance price with our costs? v. Can we balance our consumption of resources in

alignment with our operating budget?

Real-Win-Worth (RWW) Analysis

565

vi. Is the opportunity there to support our business

financial goals? vii. Is this opportunity worth the risk in comparison to

other opportunities? Step 3.

Assess each RWW question by rating the ability answer it.

A

The response score quantifies the level to which the project team can fulfill (or address) the needs of an RWW characteristic. The value of the score reflects either

B

a. The condition of the relevant marketplace, competition, or risk data that the team currently possesses b. The team’s usage level of supporting tools to gather the required data RWW Strength-of-Fulfillment Scale—Within each matrix, using the rating scale provided, answer each question and record the number in the middle column: a. Score of 1 = based on internal assumptions, guesses, past

experiences, opinions, feelings; or no supporting tool usage by the team b. Score of 3 = based on old, partial data sets and secondary

sources of data (mostly internal interpretations); or some supporting tool usage by the team c. Score of 5 = based on new, complete data sets and pri-

mary sources of data (mostly external interpolations); or extensive supporting tool usage by the team Step 4.

Record any comments. Identify any supporting comments, such as the tools used to collect pertinent data and the date that the data was last gathered. a. Supporting tools There are several supporting tools that the project team can draw upon to gather data on the RWW characteristics. The set of figures that follow show examples of some supporting tools that align with each question. Figure R-5 displays a sample REAL matrix with its supporting tools.

C D E F G H I J K L M N O P Q R S T U V W X Y Z

566

Encyclopedia Real-Win-Worth Analysis REAL Criteria

Ranking Value Comments

Supporting Tools

1

Have we defined a legitimate market and segment arena in which to participate?

Market Definition, Segmentation and Opportunity Identification, SWOT, MPQP

2

Have we qualified the size of the market?

Market Definition, Segmentation and Opportunity Identification

3

Have we quantified the market growth rate?

A

Market Definition, Segmentation and Opportunity Identification, Mkt. Behavioral Dynamics Mapping

4

Have we quantified the purchasing capacity and capability of the market players?

B

Market Definition, Segmentation and Opportunity Identification, SWOT, Porter’s 5 Forces

5

Have we defined clear market trends and customer needs?

C

Market Behavioral Dynamics Mapping, MPQP, VOC Gathering and Processing, Competitive Benchmarking

6

Do the market trends and needs align with our ideas and capabilities (e.g. technologies)?

D

Product Portfolio Architecting, Portfolio Ranking and Balancing, Idea Data base Documentation

7

Does the market have a strategic fit with our business strategy for organic growth and core competencies?

Innovation Strategy, Market Definition, Segmentation and Opportunity Identification, Portfolio Ranking and Balancing, Idea Data base Documentation

8

Does the market have a strategic fit with our business strategy for external partnerships, Joint Ventures and acquisitions?

Innovation Strategy, Market Definition, Segmentation and Opportunity Identification, Portfolio Ranking and Balancing, Idea Database Documentation

E F G

Total REAL Characteristic Fulfillment Score

H I J

0

• 1 = based upon internal assumptions, guesses, past experiences, opinions, feelings; or no supporting tool used. • 3 = based upon old, partial data sets and secondary sources of data (mostly internal interpretations); or some supporting tool used • 5 = based upon new, complete data sets and primary sources of data (mostly external interpolations); or extensive supporting tool used.

Figure R-5: Real RWW Matrix (with Supporting Tools)

K L

Figure R-6 displays a sample WIN matrix with its supporting tools.

M

Real-Win-Worth Analysis

N WIN Criteria

Ranking Value Comments

Supporting Tools

O

1

Portfolio Ranking and Balancing, Product Pipe-line Planning and Activation

P

Can we meet the launch cycle-time requirements to be first or early in the market?

2

Do we have technology and performance superiority relative to our competition?

Pugh Process, Competitive Benchmarking, Portfolio risk Analysis, MPQP, SWOT, idea Data base Documentation

3

Can we establish competitive pricing?

Competitive Benchmarking, Portfolio Risk Analysis, MPQP, SWOT, Business Case Analysis

4

Is our brand identity competitive?

Competitive Benchmarking, MPQP, SWOT, Porter’s Forces

S

5

Can we gain market share?

Market Definition, Segmentation and Opportunity Identification, SWOT, MPQP, Competitive Benchmarking

T

6

Can we compete on price?

Market Definition, Segmentation and Opportunity Identification, SWOT, MPQP, Competitive Benchmarking

U

7

Are our core competencies enabling us to compete?

Market Definition, Segmentation and Opportunity Identification, SWOT, MPQP, Innovation Strategy

V

8

Are our operational efficiencies enabling us to compete?

Portfolio FMEA, Innovation Strategy, Product Pipe-line Planning and Activation

W

9

Do we have enough resources to compete in this market?

Portfolio FMEA, Innovation Strategy, Product Pipe-line Planning and Activation

Q R

X Y Z

Total WIN Characteristic Fulfillment Score

0

• 1 = based upon internal assumptions, guesses, past experiences, opinions, feelings; or no supporting tool used. • 3 = based upon old, partial data sets and secondary sources of data (mostly internal interpretations); or some supporting tool used • 5 = based upon new, complete data sets and primary sources of data (mostly external interpolations); or extensive supporting tool used.

Figure R-6: Win RWW Matrix (with Supporting Tools)

Real-Win-Worth (RWW) Analysis

567

Figure R-7 displays a sample WORTH matrix with its supporting tools. Real-Win-Worth Analysis WORTH Criteria

Ranking Value Comments

Supporting Tools Business Case Analysis, Portfolio FMEA

1

Is the potential NPV worth the investment?

2

Is the risk adjusted NPV (ECV) worth the investment?

3

Does the Internal Rate of Return potential meet our standards?

Business Case Analysis, Portfolio FMEA

4

Can we balance price with our costs?

Business Case Analysis, Portfolio FMEA, MPQP, SWOT

5

Can we balance our consumption of resources in alignment with our operating budget?

Portfolio FMEA, MPQP, SWOT, Product Pipe-line Planning and Activation

6

Is the opportunity able to support our business financial goals?

Market Definition, Segmentation and Opportunity identification, Business Case Analysis, Portfolio FMEA

7

Is this opportunity worth the risk in comparison to other opportunities?

Market Definition, Segmentation and Opportunity identification, Idea Data base Documentation, Business Case Analysis, Portfolio FMEA, Pugh Process

Total REAL Characteristic Fulfillment Score

Business Case Analysis, Portfolio FMEA

0

• 1 = based upon internal assumptions, guesses, past experiences, opinions, feelings; or no supporting tool used. • 3 = based upon old, partial data sets and secondary sources of data (mostly internal interpretations); or some supporting tool used • 5 = based upon new, complete data sets and primary sources of data (mostly external interpolations); or extensive supporting tool used.

Figure R-7: Worth RWW Matrix (with Supporting Tools)

A B C D E F G H I J K

Step 5.

Sum the scores within each RWW characteristic matrix. Within each of the three matrices, add the scores associated with each question found in the middle column to compute the individual Real, Win, and Risk characteristic scores.

Step 6.

Compute the aggregate RWW score. Add the total scores from each individual Real, Win, and Risk matrix to determine the overall RWW strength-of-fulfillment score. The three-score tally reflects the offering’s RWW opportunity, with the largest scores reflecting the least risky investment.

L M N O P Q R S T U V

How to Analyze and Apply the Tool’s Output The RWW analysis can apply to two key process arenas: the strategic portfolio renewal process and the tactical offering development and commercialization preparation process.

W

In the portfolio renewal process, the strategy team probably examines several future offering concepts to select which idea(s) will receive investment funding and thereby activate a development project. Use the RWW analysis in conjunction with the other portfolio renewal tools to help guide the balance of your portfolio. The largest RWW score (representing the lowest risk) is a desirable score but is not necessarily the best score,

Z

X Y

568

A B C D E F

Encyclopedia

but it should not necessarily eliminate offerings with smaller scores. Recall that the RWW Analysis is a judgment enhancement technique. Nonetheless, an RWW analysis is a judgment based on a snap-shot in time, and market conditions and competition inevitably change. The team may be blinded by what it “desires” to be funded and might answer the RWW analysis questions in a biased manner. Hence, a balanced portfolio of RWW scores may represent a less risky position. In summary, when faced with uncertainty, an investment portfolio should represent a mixture of risk and a mixture of scores. The development/commercialization project team will continue to revise the RWW analysis from the portfolio renewal team. At this stage, the technique refines the project team’s understanding of the commercialization efforts needed to win in the competitive environment.

G H I J K L

Examples Sample RWW Analysis Figure R-8 shows the set of three characteristics matrices linked as one RWW analysis for one product concept completed during a portfolio renewal process.

M N

Hints and Tips

O

• Use data driven facts to complete the individual RWW matrices.

P

• Ensure the completed RWW tool is part of an evergreen process

Q

from portfolio renewal through the end of the product develop-

R

ment and market-launch preparation process.

S

• This technique enhances judgment as to the best concepts in which

T

to invest; there is no one right answer. The largest RWW score may

U

not necessarily eliminate a concept that received a lower score.

V

Attempt to build a mix of RWW scores to balance the portfolio of

W

concepts to receive funding as an activated development project.

X Y Z

Supporting or Linked Tools Supporting tools, techniques, and best practices that might provide input when developing an RWW analysis include: • Business Case Analysis • Competitive Benchmarking (See Also “Benchmarking—Avoid Arro-

gance and Lethargy,” in Part III, p. 789 for more information.)

Real-Win-Worth (RWW) Analysis REAL Criteria

569

Ranking Value Comments

Have we defined a legitimate market and segment arena in which to participate?

3

Have we qualified the size of the market?

5

Market Definition and SWOT completed on 6/25/05 Market Definition completed on 6/25/05

5

Segmentation and Opportunity Identification completed on 7/1/05

Have we quantified the purchasing capacity and capability of the market players?

1

SWOT and Porter’s 5 Forces completed on 6/25/05

Have we defined clear market trends and customer needs?

3

Have we quantified the market growth rate?

Market trends tracking completed on 6/1/05

Do the market trends and needs align with our ideas and capabilities (e.g. technologies)? Does the market have a strategic fit with our business strategy for organic growth and core competencies?

5

B

Innovation Strategy documented on 5/9/05

C

5 Innovation Strategy documented on 5/9/05-

Does the market have a strategic fit with our business strategy for external partnerships, Joint Ventures and acquisitions?

3

Total REAL Characteristic Fulfillment Score

30

WIN Criteria

F

5 Pugh Process and SWOT completed on 6/25/05

Do we have technology and performance superiority relative to our competition?

3

Can we establish competitive pricing?

3

Is our brand identity competitive?

3

Can we gain market share?

Competitive Benchmarking completed 7/1/05 Competitive Benchmarking completed 7/1/05

G H I J K

SWOT completed on 6/25/05

L

SWOT completed on 6/25/05

M

3

Can we compete on price?

D E

Ranking Value Comments

Can we meet the launch cycle-time requirements to be first or early in the market?

A

Technical Portfolio Ranking and Balancing completed on 7/15/05

3 SWOT completed on 6/25/05

Are our core competencies enabling us to compete?

5

Are our operational efficiencies enabling us to compete?

3

Do we have enough resources to compete in this market?

3

P

Total WIN Characteristic Fulfillment Score

31

Q

Portfolio FMEA completed on 7/705

Portfolio FMEA completed on 7/705

WORTH Criteria

Ranking Value Comments

Is the potential NPV worth the investment?

5

Business Case Analysis, Portfolio FMEA completed on 7/7/05

Is the risk adjusted NPV (ECV) worth the investment?

3

Business Case Analysis, Portfolio FMEA completed on 7/7/05

Does the Internal Rate of Return potential meet our standards?

3

Business Case Analysis, Portfolio FMEA completed on 7/7/05

5

Business Case Analysis, Portfolio FMEA completed on 7/7/05

Can we balance our consumption of resources in alignment with our operating budget?

5

Business Case Analysis, Portfolio FMEA completed on 7/7/05

Is the opportunity able to support our business financial goals?

3

Business Case Analysis, Portfolio FMEA completed on 7/7/05

Is this opportunity worth the risk in comparison to other opportunities?

5

Total REAL Characteristic Fulfillment Score

29

Can we balance price with our costs?

Business Case Analysis, Portfolio FMEA completed on 7/7/05

• 1 = based upon internal assumptions, guesses, past experiences, opinions, feelings; or no supporting tool used. • 3 = based upon old, partial data sets and secondary sources of data (mostly internal interpretations); or some supporting tool used • 5 = based upon new, complete data sets and primary sources of data (mostly external interpolations); or extensive supporting tool used.

Total RWW Score:

Figure R-8: Sample RWW Analysis

90

N O

R S T U V W X Y Z

570

Encyclopedia • Idea Database Documentation • Innovation Strategy • Market Behavioral Dynamics Mapping

A B C

• Market Definition • MPQP (estimated Market Perceived Quality Profile) (See Also “Mar-

ket Perceived Quality Profile (MPQP),” p. 390)

D

• Offering Pipe-line Planning and Activation

E

• Porter’s 5 Forces (See Also “Porter’s 5 Forces,” p. 464)

F

• Portfolio Architecting, Ranking and Balancing (See Also “Selecting

G H I J

Project Portfolios Using Monte Carlo Simulation and Optimization,” in Part III, p. 921) • Portfolio Risk Analysis (FMEA) (See Also “Failure Modes and

Effects Analysis (FMEA),” p. 287)

K

• Pugh Process (See Also “Pugh Concept Evaluation,” p. 534)

L

• Segmentation and Opportunity Identification

M N O P Q R S T U V W X Y Z

• SWOT (Strengths-Weaknesses-Opportunities-Threats) (See Also

“SWOT (Strengths-Weaknesses-Opportunities-Threats,” p. 699) • VOC Gathering and Processing (See Also “Voice of Customer Gath-

ering Techniques,” p. 737) A completed RWW analysis provides input to tools such as • Launch plan (See Also “Matrix Diagrams—7M Tool,” p. 399) • Communication plan (See Also “Matrix Diagrams—7M Tool,”

p. 399) • Marketing materials and collaterals • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),”

p. 287) • Control plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

Figure R-9 illustrates the link between the RWW analysis and its related tools and techniques.

Regre ssion Analysis Market Definition

Offering Pipeline Planning and Activation

MPQP

Porter’s 5 Forces

571

Portfolio Ranking and Balancing

Business Case Analysis Competitive Benchmarking

RWW Analysis

Idea Database Documentation Innovation Strategy

• Launch Plan • Communication Plan • Marketing materials and collaterals • FMEA • Control Plan

A B C D E

Market Behavioral Dynamics Mapping

F G H

Portfolio Risk Analysis (FME)

Product Portfolio Architecting

Pugh Process

Segmentation and Opportunity Identification

SWOT

VOC Gathering and Processing

Figure R-9: RWW Analysis Tools Linkage

I J K L

Regression Analysis

M N O P

What Question(s) Does the Tool or Technique Answer? What is the cause-and-effect model that describes the process and its critical variables?

Q

Regression analysis helps you to

R

• Model and predict results (of a dependent variable) based on mathe-

matical manipulation of one or more independent variables • Define a statistical model to estimate the causality relationship, thus

saving time, money and resources predicting, optimizing, and controlling a business outcome

S T U V W X Y

Alternative Names and Variations This technique includes different approaches, depending on the available data: • Linear Regression • Multiple Regression • Logistic Regression

Z

572

Encyclopedia

When Best to Use the Tool or Technique After drawing a Scatter diagram to visualize the correlation between two numerical sets of data, conduct a regression analysis to understand any causality relationship. (See Also “Scatter Diagram—7QC Tool,” p. 640) A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Brief Description Regression analysis is a statistical tool that measures the strength of relationship between one or more independent variables with a dependent variable. It develops a quantitative model that relates one or more independent variables (Xs) to a single dependent variable (Y). This relationship is often referred to as “the regression of X on Y” because the independent variable is plotted on the X-axis (horizontal axis), and the dependent variable is plotted on the Y-axis (vertical axis). It builds on the correlation concepts to develop an empirical, databased model. Correlation describes the X and Y relationship with a single number (the Pearson’s Correlation Coefficient (r)), whereas regression summarizes the relationship with a line—the regression line. Regression fundamentally operates on the principle that “if X causes Y and X is known, then Y can be predicted.” If the independent variables (X) can be adjusted and controlled, then the regression model can assist in controlling or optimizing the output (Y). This regression analysis models the Y = f(x) equation (that is, Y is a function of X), where Y is the output or response variable, and the X represents the critical variables in the given process and its inputs that impact Y. The regression model presumes that a system (either physical or behavioral) is governed by physics that can be expressed by the following equation: y

β0

β1x

physics

ε variation (noise)

Notice that this is the algebraic equation for a line (y = mx + b) plus an error-term. The terms β0 is the slope of the line, and β1 is the y-intercept, such that if X=0, then the β1 would fall on the Y-axis. Both β0 and β1 are constants. The error-term is denoted by ε. If the equation had zero error, and X were known, then Y could be calculated. The sample data (X) is gathered to represent a population. Thus, these observations (or sample data) resemble the physics of the population’s system plus any noise. The objective is to predict the population’s system. The predicted value (y-hat, or y^) represents the physics portion of the equation, denoted as the beta β variables and the observation variable (X), namely β0 + β1X.

Regre ssion Analysis

The study of baseball lends itself well to regression analysis. Baseball enthusiasts know the importance of statistics to their sport. Economics professor J.C. Bradbury, of Kennesaw State University in Georgia applies multiple regression to the volumes of baseball historical data and wrote a book (The Baseball Economist: The Real Game Exposed) based on his findings. Bradbury uses statistics to describe baseball facts from fiction and bust some prevalent myths as to what makes a team successful. Bradbury claims to have identified baseball’s critical success factors that if adjusted and controlled can produce a winning season. (See Also “Science Journal: A New Study Sows How Baseball Myths Can Hurt the Game,” in the Wall Street Journal, Friday, February 16, 2007, page B1.) The regression analysis strives to find the best model to fit the data. The best-fit line passes through the data points where the smallest possible total sum of the square of all the residuals. The calculation squares the difference between the residual values and the predicted values that fall on the regression line to accommodate for the distance of the sample data on either side of the line—regardless if it is a positive or negative difference. Thus, the regression analysis (usually) uses the Least Squares method to determine the best fit by selecting the minimum sum of the squared deviations about the regression line. The approach uses the smallest sum of all the squared residual values calculated for each data point to determine the regression line. This model uses a regression line (best-fit line) to estimate the physics in the presence of noise. The noise is estimated by the residual value. The sample data helps make assumptions about the error (ε), such that the observed value (Y) minus the fitted (predicted) value on the best-fit line yields the residual value. The noise is the variation between the observed data and the best-fit line, as shown in Figure R-10.

573

A B C D E F G H I J K L M N O P Q R S

Observed value

error

Variation (noise)

T U V

Y

W A fitted (predicted) value on the best-fit line

X Y Z

X Figure R-10: Fitted Line Plot Illustrating Error

xi

574

A B C D E F G H I J K L M N O P Q R S T

Encyclopedia

Linear Regression Assumptions Hypothesis: The Null hypothesis states that the known X values are unable to predict Y. Thus any change in X has no effect on Y; the Y values remain horizontal—the slope of the regression line is zero. The regression line’s slope is proportional to the correlation coefficient, such that if the one is zero, so is the other. The regression analysis can use the correlation coefficient as a surrogate for the slope of the regression line—called Pearson’s Correlation Coefficient (r), and r=0. Conversely, the alternative hypothesis states that the true relationship between the variables is linear; wherein r = 1 or r = -1. Minimum Sample Size: 20 data points. Normally distributed data: The residuals are assumed to be normally distributed. Residuals: The regression model assumptions about the residuals (error, ε) are three-fold: independence, normally distributed, and have a constant variance. The last assumption of constant variance is the most important of the three. The condition of independence ensures that the values of each variable are a result of random sampling.

Non-Linear Regression There are alternative types of regression modeling other than linear. Multiple regression not only models multiple variables, but also may introduce interactions, quadratic, and/or cubic factors in the equation to better fit the model. An interaction comprehends the effect that two or more variables have on one another. A quadratic is the squaring of a term [X2—multiplying by itself once (X*X)], and a cubic is the cubing of a term (X3—multiplying by itself twice (X*X*X)). Figure R-11 compares these different model types drawn in MINITAB. Linear Model

Quadratic Model

Cubic Model

Y = b 0 + b 1X

Y = b 0 + b 1X + b 2X 2 2 With a squared term (X )

Y = b 0 + b 1X + b 2X 2 + b 3X 3 3 With a cubed term (X )

U V W X Y Z

Figure R-11: Comparing Linear, Quadratic, and Cubic Models

Multiple regression can be dissected into multiple linear regression modeling multiple independent variables and multivariate regression modeling multiple independent and dependent variables. The formula for testing the multiple regression’s Null hypothesis uses the ratio of variances as an F-statistic. (See Also “Analysis of Variance (ANOVA)—7M Tool,” p. 142)

Regre ssion Analysis

575

Logistic regression uses the Logistic or Logit function to model the system’s scenarios involving non-normally distributed data or discrete variables such as categorical, binary, and ordinal variables. Complex forms of regression using the splines technique may be useful. Splines is a statistical technique that uses a mix of different lines and curves to generate a better fitting model. Scenarios that may call for this type of modeling include product lifecycles. A product lifecycle curve may be generalized to encompass four segments. It starts with an S-curve slowly climbing as the product is introduced in the marketplace. Next, (ideally) the product experiences exponential growth. Third, the curve levels off as it matures, and sales growth flattens as the market gets saturated. Last, as the product approaches end-of-life, the curve follows a steep decline along another S-curve. As with simple linear regression, the shape of curve (the mathematical function) is determined by the errors in predicting the dependent variables. However, some of the more complex regression models must use the Maximum Likelihood method to determine the best fit, rather than the Least Squares method used for linear regression.

A B C D E F G H I J K

Hints and Tips Request the assistance of a statistician to conduct these types of complex regression analyses.

L M N O P

Interpreting the Residual Plots Regression analysis can be completed manually, but it is very tedious, thus it is recommend to use a statistical software package such as MINITAB. Graphical output is an essential component of a regression analysis. MINITAB provides a single four-in-one residual plot that assists in evaluating the three regression assumptions about the residuals (independence; normally distributed; and have a constant variance). The four-up graph, show in Figure R-12, includes a Normal Probability plot, histogram, Scatter plot around the regression line, and a Time Series plot of the residuals. (See Also the “Normal versus Non-normal Data” section in the “Control Charts” entry, p. 227, for the Normal Probability plot; “Histogram— 7QC Tool,” p. 330; “Scatter Diagram—7QC Tool,” p. 640; and “Run Chart—7QC Tool,” p. 610 for more detail on each graphical tool.)

Q R S T U V W X Y Z

576

Encyclopedia Residuals Plots for y

A

90

5

50

- 10

1 - 10

H I J K L M N O P Q R S T U V W X Y Z

0

5

10

10

20

30

50

Residuals Versus the Order of the Data

4

10

3

5

Residual

Frequency

40

Fitted Value

Histogram of the Residuals

D

G

-5

Residual

C

F

0 -5

10

B

E

Residuals Versus the Fitted Values 10

Residual

Percent

Normal Probability Plot of the Residuals 99

2 1

0

-5

0

- 10 - 7.5

- 5.0

- 2.5

0.0

2.5

Residual

5.0

7.5

10.0

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15

Observation Order

Figure R-12: Example of a MINITAB Residual Plot (Four-in-One)

Dissect Figure R-12 into its four components. The upper-left quadrant depicts a linear normal probability plot of residuals to support the normality assumption. The lower-left quadrant features a histogram of the residuals to provide another tool to evaluate the normality assumption. The Scatter plot of residuals versus the fitted value, displayed in the upper-right quadrant, portrays a random pattern to support the constant variance assumption. The forth quadrant in the lower-right exhibits a time-ordered plot of residuals with a random pattern to support the independence assumption. In summary, all four graphical tools displayed in Figure R-12 indicate that the regression model is adequate, based on the residual assumptions.

Common Patterns As previously stated, the residual assumption of constant variance calls for a no pattern to be displayed in the Scatter diagram of residuals versus the fitted value. Often this random distribution is referred to as a buckshot pattern, shown in Figure R-13, signaling that the model is adequate. Figure R-13 illustrates the buckshot pattern and three others—the Funnel, Smile (or Frown), and Football patterns. Each of these remaining three Scatter plots contain non-random patterns showing an inconsistent variance of the residuals. Because their patterns are non-random, the regression model is inadequate, and the graphs have an X drawn through them.

Regre ssion Analysis Funnel Pattern

0

Residuals

Residuals

Buckshot Pattern

577

0

A Fits

Smile (or Frown) Pattern

Football Pattern

0

Fitted Values

B C D

Residuals

Residuals

Fits

0

E F

Fitted Values

Figure R-13: Common Patterns of Residuals Versus Fitted Values

In the Funnel pattern, the error increases as Y increases. This model fits better at the low-Y values than at the high-Y values. A transform may be needed.

G H I J K L

In the Smile (or Frown) pattern, the model fits better at the mid-Y values than at low- and high-Y values. A higher-order model maybe needed, such as a quadratic, as shown in Figure R-11.

M

In the Football pattern, the model fits better at the low- and high-Y values than at the mid-Y values. A transform may be needed.

P

Interpreting the Statistical Results After creating a regression model, there are two key results that indicate whether or not the model is a good fit. The regression analysis focuses on the two statistics—the R-Square-adjusted term and the p-value. R-Square-adjusted Given the Scatter plot in Figure R-14, which line best fits the data? The answer is revealed in conducting a regression analysis. The objective to determine the best fitting model is to identify the largest possible RSquare (r2 or R2) value, also called the coefficient of determination. The coefficient of determination is calculated by squaring the correlation coefficient, and has a value between 0 and 1.

N O Q R S T U V W X Y Z

578

Encyclopedia

Scatterplot of Y versus X 50

40

A C

30

Y

B

20

D E F G

10

0 10

20

30

H I

40

50

X

Figure R-14: Which Line Is the Best Fitting?

J K L M N O P Q R S T U V W X Y Z

R-Square increases with every additional predictor (independent variable) added to the model even if the predictor adds little value to the model. To compensate for this effect, the statistical term called R-Squareadjusted (r2adjusted or R2adjusted) is introduced. R-Square-adjusted is like RSquare except that it is allowed to decrease if non-useful predictors are added to the model. R-Square-adjusted reinforces the notion of a parsimonious model—the concept of striving for the most conservative, sparse, frugal, simple model. In general, the R-Square-adjusted value is the best term to assess how good a model is. The formula for R-Square-adjusted (R2adjusted) is as follows:

R2adjusted

1

(1

n R2) n

1 p

,

where n is the sample size, and p is the number of predictors (independent variables) used in the model.

P-value The Predictor’s P-value—When building a regression model with more than one independent variable, often an iterative process of elimination is used to identify the critical few predictors needed to achieve the most parsimonious model. The p-value helps to identify which predictor in the model is useful. Those independent variables deemed useful in the

Regre ssion Analysis

579

regression model will yield a p-value < 0.05, indicating that it is statistically distinguishable from the other predictors. Recall the Null hypothesis states that the known X values are unable to predict Y, and the Alternative hypothesis states that the true relationship between the variables is linear (r=1). The Lack of Fit P-value for ANOVA Residual Error—The p-value also indicates the amount of fit the regression model has applied when examining the resulting Residual Error’s Lack of Fit portion of the Analysis of Variance (ANOVA) test. The Lack of Fit p-value indicates if the linear predictors (Xs) alone are sufficient to explain the error in the model. The small p-value indicates that the predictors (Xs) are not sufficient in explaining the variation in the response (Y). With a p-value < 0.05 for the Residual Error Lack of Fit, the model needs improvement, and interactions of the significant predictors should be used; for example add a quadratic or interaction term one at a time and rerun the model. The Null hypothesis is rejected. Conversely, when the ANOVA Residual Error Lack of Fit’s p-value > 0.05, it indicates that there is no lack of fit. There is insufficient evidence to reject the Null. For the ANOVA Residual Error Lack of Fit test, the Null hypothesis states that the known X values (predictors) alone are able to predict the variance in Y (for example, variation is zero); and the Alternative hypothesis states that the known X values (predictors) are unable to predict the variance in Y. Refer to the following section on terminology for more detail. (See Also “Hypothesis Testing,” p. 335 for more information on the p-value and “Analysis of Variance (ANOVA)—7M Tool,” p. 142)

Predictions Models, such as regression, often are used to predict something, such as a cause-and-effect relationship. Regression attempts to statistically model the effect one or more (independent) variables has on another (dependent) variable. Regression predicts (or estimates) the value of a dependent variable (Y) using independent variables (X) not found in the sample data. Interpolation describes the predicting of an outcome (Y) when the unknown independent variables (X) fall within the range of the known sample data values. Extrapolation defines predicting an outcome (Y) based on unknown independent variables (X) that fall outside the range of the sample data values. Interpolation yields far more accurate predictions, hence it is the preferred approach between it and extrapolation.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

580

A

Encyclopedia

This type of prediction is not necessarily predicting the future. If the independent variable were time, then the regression analysis could only utilize historical data to extrapolate in the future. Predicting the future is difficult and might best be accomplished using statistical tools such as Monte Carlo simulation to forecast possible outcomes. Monte Carlo simulations are non-deterministic, providing a range of probable possibilities. (See Also “Monte Carlo Simulation,” p. 431)

B C D E F G H I J K L M N O P Q R S T U V W X Y Z

How to Use the Tool or Technique Regression analysis can be conducted manually or by using computer software. The following procedure uses MINITAB to illustrate the technique. The procedure covers both simple linear regression and multiple regression. The procedure to run a logistic regression in MINITAB follows a similar approach as that described for the multiple regression. However, logistic regression is not covered in this book. Logistic regression concepts begin to embark on some of the more complex regression models, thus proper setup and analysis may best be done with the assistance of a statistician. Preparation: Determine the relationship that needs to be studied, collect the data, and enter the numeric data has been entered into the MINITAB Worksheet.

Simple Linear Regression Step 1. Graph the data first. a. Graph the relationship on a Scatter diagram and examine the fitted line. To do so, select the following commands from the MINITAB drop-down menu: Stat > Regression > Fitted Line Plot…. b. On the Fitted Line Plot main screen, select the columns containing the appropriate independent and dependent variables to place in the corresponding dialog boxes. Unless the data contains a quadratic or cubic function, keep the Type of Regression Model on its default of Linear, as illustrated in Figure R-15. c. Click OK to generate the subsequent Scatter diagram with the initial regression model and its best-fit line for the sample data, as illustrated in Figure R-13. d. MINITAB provides both the proposed regression model and some statistics to evaluate how well the model fits the data, as depicted in the Scatter diagram found in Figure R-15. In this example, the linear regression model is: Y = 0.033 + 0.9900X. The R-Square is reported at 86.9%.

Regre ssion Analysis

581

A B C Figure R-15: Example of a MINITAB Regression Fitted Line Plot Main Screen and Resulting Scatter Diagram

D E F

Step 2.

Analyze the regression model.

G

a. MINITAB’s resulting session window provides the necessary statistics to evaluate how well this model fits, shown in Figure R-16.

H

b. The coefficient of determination (R-Square) appears reasonably strong at 86.9%, indicating an excellent relationship. (See Also “Rule of Thumb” in the Hints and Tips section of this entry, p. 595, for detail on the strength of the relationship.)

K

c. The ANOVA test reported a p-value of 0.000 for the regression (line), thus less than the conventional threshold of a p-value < 0.05, indicating that the regression line’s slope is statistically different from zero. Based on the p-value of 0.000, the Null hypothesis is rejected.

I J L M N O P Q R S

Note The Null hypothesis states that the regression line’s slope is zero, showing no linear relationship between the independent and dependent variables. MINITAB’s Fitted Line Plot (shown in Figure R-15) also provides the R-Square information, along with R-Square-adjusted and the standard deviation. However, the Session Window provides much more information.

T U V W X Y Z

582

Encyclopedia

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure R-16: Example of a MINITAB Regression Fitted Line Plot Session Window

Multiple Regression Scenario: For purposes of illustration, the following physical system example is used to describe the procedure to conduct a multiple regression. An engineer is developing a medical product. The technology requires heating and cooling a volume of fluid contained in a plastic pouch. The chosen system uses resistive heating and forced air cooling. The engineer thinks that the following variables are critical to determining the time it takes to cool the product. The list of independent variables (Xs) is: Fan Speed, Fluid Volume, Ambient Temperature, Altitude, and %RH (Relative Humidity). The output (Y) is the time to cool the product. This is a smaller is better scenario, whereby the less time it takes to cool the product, the better. Business question: What is the most parsimonious regression model to optimize and control the cool time? (Thus, are all the independent variables critical, or should some be removed to derive the best-fit model?) Preparation: Determine the relationship that needs to be studied, collect the data, and enter the numeric data into the MINITAB Worksheet. Step 1.

Graph the data first. a. Graph the relationship on a Scatter diagram and examine the fitted line. To do so, select the following commands from the MINITAB drop-down menu: Stat > Regression > Regression….

Regre ssion Analysis

583

b. On the Regression main screen, select the columns containing the appropriate independent and dependent variables to place in the corresponding dialog boxes, as illustrated in “Area 1” of Figure R-17. i. Select the column of data with the dependent variable to go into the Response dialog box. In this case, the cool time. ii. Select the appropriate columns containing the different independent variables to go in the Predictor dialog box. In this case, initially all the variables are presumed critical until the model proves otherwise. Thus, all are selected: Fan Speed, Fluid Volume, Ambient Temperature, Altitude, and %RH (Relative Humidity). c. Select the Graphs… button on the Regression main screen to determine the type of graphing output desired and the residual values of interest, as illustrated in “Area 2” of Figure R-17. i. Select the type of Residuals for Plots of interest in the appropriate dialog box. In this case, Regular was selected. ii. Select the type of layout for the Residual Plots in the appropriate dialog box. In this case, Four in one was selected. iii. Enter the independent variables of interest requiring residuals be calculated in the Residual Versus the Variables dialog box. In this case, all of the independent variables were selected, as they each are presumed significant predictors; but not the dependent variable (cool time). iv. Select OK. d. Select the Options… button on the Regression main screen to identify any regression model options, as illustrated in “Area 3” of Figure R-17. i. The Fit Intercept is the default. In this case, leave the Fit Intercept selected. ii. In this case, select the Pure error under the Lack of Fit Tests.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

584

Encyclopedia

iii. Select OK. e. Select the OK button on the Regression main screen to run the regression model, as illustrated in “Area 1” of Figure R-17. A B C D E F G H I J K L M N O P

Figure R-17: Example of a MINITAB Regression Main Screen

Q R S T U V W X Y Z

Step 2.

Analyze the regression model’s Residual Plots Four in one graph, as illustrated in Figure R-18. Examine each of the four residual graphs one at a time to determine if the regression model assumptions for residuals are met. a. Normal Probability Plot of the Residuals (in the upperleft quadrant)—Examine how the data plot appears normal such that the data is plotted close to the line. Recall the regression assumption that the residuals are to be distributed normally. (See Also the “Control Charts” entry in the “Normal Versus Non-normal Data” section, p. 227, for a more detailed discussion about the Normal Probability plot.) In the Figure R-18 example, the residuals appear not to be evenly distributed around the normal probability line, and either extreme data points seem outside what the expected confidence interval might be. It appears the model does not fit the observations well.

Regre ssion Analysis

b. Histogram of the Residuals (in the lower-left quadrant)— Examine the histogram to see if the data appears to be distributed normally, that is, a bell-shaped curve. Recall the regression assumption that the residuals are to be distributed normally. (See Also “Histogram—7QC Tool,” p. 330, for more detail.) In the Figure R-18 example, the residuals appear to be skewed, exhibiting a long tail on the left of the plot, reinforcing that the model does not fit well. c. Residuals Versus the Fitted Values (in the upper-right quadrant)—Examine how randomly the data appears to be distributed around the residual line (indicated as the “zero” line on the plot). Recall the regression assumption that the residuals are to have a constant variance around the regression line. In the Figure R-18 example, the residuals appear to be distributed in a pattern of an upside-down smile, or a frown, with the absence of data below the residual line in the center of the plot. Again, it appears the model does not fit well. d. Residuals Versus the Order of the Data (in the lowerright quadrant)—Examine how randomly the data appears to be distributed around the residual line (again depicted as the “zero” line on the plot). Recall the regression assumptions that the residuals are to have a constant variance around the regression line and independence such that they would be distributed randomly over time around the regression line. In the Figure R-18 example, the residuals appear randomly distributed around the residual line for the firsthalf of the observed sequence but tighten and for the most part move above the line for the second-half—again reinforcing that the model does not fit well. Step 3.

Analyze each of the regression model’s Residual Plots versus each of the independent variables, as illustrated in Figure R-19. Examine each of the four residual graphs one at a time to determine if the data appears to be randomly distributed or displays a pattern.

585

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

586

Encyclopedia Residuals Plots for Cool Time [s] Normal Probability Plot of the Residuals

Residuals Versus the Fitted Values

99

1

A

Percent

90 0 50 -1

B

10

C

1

-2 -2

4

5

6

7

Fitted Value

Residuals Versus the Order of the Data 1

Residual

9

6

0

-1

-2

0

J L

3

3

I K

2

12

Frequency

H

1

Histogram of the Residuals

E G

0

Residual

D F

-1

- 1.6

- 0.8

0.0

2

0.8

4

6

8 10 12 14 16 18 20 22 24 26 28 30 32

Residual

Observation Order

Figure R-18: Example of a MINITAB Residual Plot for a Multiple Regression (Four in One)

M N

Q

0.5 -0.5 -1.5 0

1

2

-2

-1

Residual

Residual

0

1

2

1

2

Vol Residuals Versus Alt (response is Cool Time [s])

0.5 -0.5 -1.5

0.5 -0.5 -1.5

-2

-1

0

1

2

Temp

W

-2

-1

0 Alt

Residuals Versus RH (response is Cool Time [s])

Residual

X Z

-1

Speed

V

Y

-0.5

Residuals Versus Temp (response is Cool Time [s])

S U

0.5

-1.5 -2

R T

Residuals Versus Vol (response is Cool Time [s])

Residual

P

Residual

O

Residuals Versus Speed (response is Cool Time [s])

0.5 -0.5 -1.5 -2

-1

0

1

2

RH

Figure R-19: Example of a MINITAB Residual Plot Versus Each Independent Variable

Regre ssion Analysis

587

In the case of this engineering example shown in Figure R-17, one plot stands out as having strongest pattern of a frown, and that is the graph in the upper-left corner of the Residuals Versus Speed.

Step 4.

The pattern of a frown or smile indicates the need for a quadratic term. (Reference Figure R-11, which illustrates the different model patterns.)

A

Modify the model via an iterative approach.

B

a. Create a quadratic for the variable with the smile/frown pattern. i. In the MINITAB Worksheet, identify an empty column and label it to indicate the new squared-term. In this example, the Speed variable exhibited a smile/frown pattern in its Scatter plot of residuals versus fitted value. Label the new column as Speed2. ii. From the Worksheet use the calculator function to square the data in the appropriate column. Select Calc > Calculator… from its drop-down menu.

C D E F G H I J K L

iii. In the Calculator main menu, select the column in which the new squared-term data should be stored and put it in the Store Result in Variable dialog box.

M

iv. In the Calculator main menu, within the Expression dialog box, enter the column name containing the term to be squared, then select the multiplication function (indicated by an asterisks * ), and then enter a second time the column with the variable to be squared.

P

In this example, the dialog box should read ‘Speed’ * ‘Speed’. Click OK, and MINITAB squares the independent variable and stores the data in the selected new column. Figure R-20 illustrates this step. An alternative calculator dialog box entry that produces the same outcome is the double asterisks2 (**2), signifying the squared-function (to the second-power). Hence the data in the Expression dialog box would be the term to be squared **2. In this example, the dialog box should read ‘Speed’ **2. b. Rerun the regression described in Step 1 (Stat > Regression > Regression…) and check the residuals again.

N O Q R S T U V W X Y Z

588

Encyclopedia

A B C D E F G H I J K L

Figure R-20: MINITAB Calculator to Square the Speed Variable

Step 5.

Analyze the new model by checking the graphical and statistical output.

M

a. Examine the graphical output. Figure R-21 provides the graphical output for this engineering example.

N O P

Residuals Plots for Cool Time [s]

Q

Normal Probability Plot of the Residuals

Residuals Versus the Fitted Values 1.0

S

90

0.5

U

50

10

- 1.0

0.0

0.5

1.0

0.0

1.5

3.0

4.5

6.0

Fitted Value

Histogram of the Residuals

X

Residuals Versus the Order of the Data 1.0

10.0 0.5

Frequency

Z

- 0.5

Residual

W Y

- 1.0

1

V

0.0 - 0.5

7.5

Residual

T

Residual

99

Percent

R

5.0

0.0

- 0.5

2.5

- 1.0

0.0 -0.75 -0.50 -0.25 0.00

0.25

Residual

0.50

0.75 1.00

2

4

6

8 10 12 14 16 18 20 22 24 26 28 30 32

Observation Order

Figure R-21: Rerun Example of the MINITAB Residual Plot for a Multiple Regression (Four in One)

Regre ssion Analysis

589

Referring to Figure R-21, there still appears to be a systematic pattern in the time-ordered plot such that the latter half of the graph reflects a non-random pattern. b. Examine the statistical output in the session window. Figure R-22 shows this engineering example’s session window. A B C D E F G H I J K L M N O P Figure R-22: Rerun Example of the MINITAB Session Window

Q R

Referring to Figure R-22, the engineering example has two insignificant factors included in the regression model, as indicated by their p-values being greater than 0.05— Alt (Altitude) and RH (Relative Humidity).

Note The Null hypothesis states that there is no relationship between the independent and dependent variable. For the two factors, Alt and RH, there is insufficient evidence to reject the Null. To achieve a parsimonious model, these two factors need to be removed from the model and the model rerun.

S T U V W X Y Z

590

Encyclopedia

For the ANOVA Residual Error Lack of Fit, it has a small p-value (less than 0.05). Hence, reject the Null hypothesis, which states that the predictors alone explain the variation in the output. The model still needs improvement by either adding a higher term or an interaction. In conclusion, add interaction terms from the significant predictors (Speed, Volume, Temperature).

A B C

Step 6.

D

a. Using the calculator function, create interactions of each of the significant predictors by repeating Step 4 three times to create three interactions.

E F G

i. In this example, the Expression dialog box should read ‘Speed’ * ‘Vol’.

H I

ii. In this example, the Expression dialog box should read ‘Speed’ * ‘Temp’.

J K

iii. In this example, the Expression dialog box should read ‘Vol’ * ‘Temp’.

L M

b. Rerun the regression model by repeating Step 1 (Stat > Regression > Regression…) and check the residuals again. Figure R-23 illustrates the Regression main screen to rerun the model with the three new interaction factors. Notice, that in the Regression—Graphs screen, for the in the Residual versus the variables dialog box, the significant predictors (X), including the three new interaction terms, are entered. (A similar procedure to Step 1.c.)

N O P Q R S T U V W X Y Z

Modify the model. Explore if any interaction terms improve the model.

Step 7.

Analyze the new model by checking the graphical and statistical output. a. Examine the graphical output. Figure R-24 provides the graphical output for this engineering example. Referring to Figure R-24 illustration, the residual plots look much improved over earlier models. The residuals appear to be distributed normally in both the Normal Probability plot and histogram. The residuals appear more randomly distributed around the regression line in both the Scatter plot and the Time Series plot.

Regre ssion Analysis

591

A B C D E F G H I J

Figure R-23: Rerun Example with Interactions in the Regression Main Screen

K L

Residuals Plots for Cool Time [s] Normal Probability Plot of the Residuals

M

Residuals Versus the Fitted Values

N

99 0.50 0.25

O

0.00

P

Residual

Percent

90

50

Q

- 0.25

10

R

- 0.50

1 - 0.50

- 0.25

0. 00 Residual

0. 25

0. 50

0. 0

Histogram of the Residuals

3.0 Fitted Value

4. 5

6. 0

Residuals Versus the Order of the Data

6. 0

Residual

4. 5

Frequency

1. 5

3. 0

U

0.25

V

0.00

W X

- 0.50

0. 0 - 0.4

- 0. 2

0.0 Residual

0. 2

0.4

T

0.50

- 0.25

1. 5

S

2

4

6

8 10 12 14 16 18 20 22 24 26 28 30 32

Observation Order

Figure R-24: Rerun Example with Interactions—MINITAB Residual Plot, Multiple Regression

Y Z

592

Encyclopedia

b. Examine the statistical output in the session window. Figure R-25 shows this engineering example’s session window. Regression Analysis: Cool Time [s] versus Speed, Vol, ...

A B C D E F G

The regression equation is Cool Time [s] = 5.02 - 0.841 Speed + 0.703 Vol + 0.448 Temp - 0.645 Speed2 + 0.585 Speed*Vol + 0.0754 Speed*Temp + 0.0534 Vol*Temp Predictor Constant Speed Vol Temp Speed2 Speed*Vol Speed*Temp Vol*Temp

Coef 5.01685 -0.84142 0.70333 0.44792 -0.64493 0.58525 0.07537 0.05337

H

S = 0.300051

I

Analysis of Variance

J

SE Coef 0.06709 0.06125 0.06125 0.06125 0.05478 0.07501 0.07501 0.07501

R-Sq = 96.0%

Remove insignificant terms to produce parsimonious model.

R-Sq(adj) = 94.8%

M N

Figure R-25: Rerun Example with Interactions—MINITAB Session Window

L

SS 51.7736 2.1607 0.9082 1.2526 53.9343

P 0.000 0.000 0.000 0.000 0.000 0.000 0.325 0.484

Source Regression Residual Error Lack of Fit Pure Error Total

K

DF 7 24 7 17 31

T 74.77 -13.74 11.48 7.31 -11.77 7.80 1.00 0.71

MS 7.3962 0.0900 0.1297 0.0737

F 82.15

P 0.000

1.76

0.161

No longer any lack of fit.

O P

Referring to Figure R-25, the engineering example has two insignificant interaction terms included in this regression model, as indicated by their p-values being greater than 0.05—Alt*Temp and Vol*Temp.

Q R S T U

Note

V

The Null hypothesis states that there is no relationship between the

W

independent and dependent variable. For the two interaction factors,

X

Alt*Temp and Vol*Temp, there is insufficient evidence to reject the

Y

Null. To achieve a parsimonious model, these two factors need to be

Z

removed from the model and the model rerun.

Regre ssion Analysis

The ANOVA Residual Error Lack of Fit has a p-value greater than 0.05. Hence, there is insufficient evidence to reject the Null hypothesis, which states that the predictors alone explain the variation in the output. The model is adequate, given that the significant predictors are able to explain the variation in the response. Step 8.

Step 9.

Modify the model. Following the procedure in Step 1 (Stat > Regression > Regression…), remove the insignificant predictors (that is, Alt*Temp and Vol*Temp) and rerun the regression model. Check the residuals again. Analyze the new model by checking the statistical output. (Because Step 7.a., Figure R-24, already provided a good set of residual plots, there is no need to look at them again. a. Examine the statistical output in the session window. Figure R-26 shows this engineering example’s session window. Referring to Figure R-26, the engineering example the predictor variables all have p-values less than 0.05. For the ANOVA Residual Error Lack of Fit, it has a pvalue greater than 0.05, at 0.201, which is improved over the earlier run shown in Figure R-25. Hence, there is insufficient evidence to reject the Null hypothesis, which states that the predictors alone explain the variation in the output. The model is adequate, given that the significant predictors are able to explain the variation in the response. The statistics indicate that this is a parsimonious model. The final model’s equation is found at the top of the session window in Figure R-26: Cool Time (Y) = 5.02 - 0.841 Speed + 0.703 Vol + 0.448 Temp - 0.645 Speed2 + 0.585 Speed*Vol.

593

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

594

Encyclopedia Regression Analysis: Cool Time [s] versus Speed, Vol, ... The regression equation is Cool Time [s] = 5.02 - 0.841 Speed + 0.703 Vol + 0.448 Temp - 0.645 Speed2 + 0.585 Speed*Vol

A B C D E F G H I J K L M N O

Predictor Constant Speed Vol Temp Speed2 Speed*Vol

Coef 5.01685 -0.84142 0.70333 0.44792 -0.64493 0.58525

SE Coef 0.06647 0.06067 0.06067 0.06067 0.05427 0.07431

T 75.48 -13.87 11.59 7.38 -11.88 7.88

P 0.000 0.000 0.000 0.000 0.000 0.000

S = 0.297245 R-Sq = 95.7% R-Sq(adj) = 94.9% Analysis of Variance Source Regression Residual Error Lack of Fit Pure Error Total

DF 5 26 9 17 31

SS MS 51.637 10.327 2.297 0.088 1.045 0.116 1.253 0.074 53.934

F 116.89

P 0.000

1.58

0.201

Figure R-26: Rerun Example with Fewer Interactions—MINITAB Session Window

P Q

Hints and Tips

R

The X values represent the independent variables that impact the Y

S

values. (In a Scatter plot, the independent variable is plotted on the

T

X-axis and the dependent variable on the Y-axis.) Be certain to prop-

U

erly assign the data using this convention when conducting a regres-

V

sion analysis.

W

Examine the Scatter plot prior to conducting a regression analysis

X

to visualize the relationship. If the curve reflects a non-linear curve,

Y

a linear regression will not be applicable. The importance of plotting

Z

the data first is discussed further in the Scatter diagram entry. (See Also, “Scatter Diagram—7QC Tool,” p. 640) The Scatter diagrams of the residuals versus the fitted values should display a random (or buckshot-like) pattern. If it reveals any

Regre ssion Analysis

595

potential patterns in the data that might distort the statistical calculations of a regression analysis for example, apply a quadratic to a smile (or frown) pattern and apply a cubic to an S-shape pattern. (Refer to Figures R-11 and R-13.) Rule of Thumb—The strength of the relationship determined by the coefficient of determination (R2 or r2—where it falls between 0% and 100%, or 0 to 1) is as follows: • Excellent relationship: • Mechanical/Physical System yielding an r2 of 91% to 100%. • Behavioral/Human System yielding an r2 of 71% to 100%. • Good relationship: • Mechanical/Physical System yielding an r2 of 76% to 90%. • Behavioral/Human System yielding an r2 of 51% to 70%. • Weak relationship: • Mechanical/Physical System yielding an r2 of 60% to 75%. • Behavioral/Human System yielding an r2 of 30% to 50%.

A B C D E F G H I J K L M N

Statistical Terminology Associated with Regression • Best-fit line—The regression line (derived from the regression model) drawn in a Scatter diagram, as shown in Figure R-12.

O P Q

Note that the slope of the regression line in Figure R-27 indicates a positive relationship as it goes from the lower-left to the upper-right corner of the graph, trending upward. Thus as X increases, so does Y.

R

• Binary logistic regression—Describes the relationship of independent variables with a dependent variable that is a binomial (e.g. good/bad, and yes/no).

U

• Causation—Means that if the value of one (independent) variable is changed, then it induces a change in the dependent variable—a cause-and-effect relationship. • Coefficient of determination (r2)—Determines how well the Scatter plot’s best fit line fits the data. This number falls between 0 and 1, with 1 representing a perfect fit and 0 representing no linear fit. No linear relationship is depicted as a horizontal best-fit line on a Scatter diagram.

S T V W X Y Z

596

Encyclopedia

Scatterplot of Y versus X 11 10 9

A B D E F G H

Y

C

8 7 6 5 4 5.0

7.5

10.0

12.5

15.0

X

I J

Figure R-27: Scatter Plot with a Regression Line

K L M N O P Q R S T U V W X Y Z

The coefficient of determination represents the proportion of Y’s variation that is explained by the regression line. Because most data points fall around, rather than on the regression line, the remainder of the proportion (1-r2) is error. Hence, r2 measures the percent of variability in the response explained the by linear relationship with the predictor (best fit line or model) —ranging between 0% and 100%, or 0 and 1. • Confidence interval—The range of values on either side of the residual line that the represents the confidence within which the true regression line falls. The default usually set at 95%. • Correlation—A metric that measures the linear relationship between two process variables. •A strong correlation does not imply causation (a cause-andeffect relationship). For example, the number of car accidents on a rainy day may be strongly correlated with the sale of umbrellas, but the buying of umbrellas did not cause the accidents. •Lack of correlation does not necessarily mean that no causal relationship exists. The range of data may be too narrow to detect relationship.

Regre ssion Analysis

• Pearson’s Correlation Coefficient (r)—Defines the correlation metric and falls between (-1) and (1), where 0 indicates no linear relationship, (-1) describes a perfect negative correlation, and (1) a perfect positive correlation. Thus, the larger the absolute value of r (either + or -), the stronger the linear relationship. Graphically, this strong relationship is characterized by a tight distribution of data around a best-fit line plotted in a Scatter diagram. Figure R-28 displays different Scatter diagrams with their respective Pearson’s Correlation Coefficient identified with each. The Pearson’s Correlation Coefficient (r) is proportional to the slope of the regression line adjusted for the differences in both X and Y’s standard deviations. Its formula is as follows:

(xi i

x )( yi

y)

C D E F G H

K

n

(xi 1

B

J

1

n i

A

I

n

rxy

597

x )2

( yi i

y )2

1

Square the Pearson’s Correlation Coefficient to derive the coefficient of determination (r2 or R2) used for the regression analysis. (See Also “Scatter Diagram—7QC Tool,” p. 640) • Error—The difference between the observed variation (noise) between the expected value, derived from the average of the entire population (which typically is unobservable). Error technically is unobservable. Errors are independent random variables of each other. Also called residual. • Extrapolation—In a stable process, the model extends beyond the range of two known input variable data values to predict a third unknown independent variable and its response variable. It presumes the unknown independent variable falls outside the range of the known modeled sample data— projecting.

L M N O P Q R S T U V W X Y Z

598

Encyclopedia 0 Mo d er ate Positiv e Co r r elatio n

5

10

Str o n g Positiv e Co r r elatio n

r = 0.64 R2 = 41%

No Co r r elatio n

r = 0.92 R2 = 85%

10.0

r = -0.03 R2 = 0%

7.5

5.0

A

2.5

B

0.0

C

10.0

D

7.5

E F

Mo d er ate Neg ativ e Co r r elatio n

2.5

H

0.0

J

Str o n g , No n lin ear Relatio n sh ip

5.0

G I

Str o n g Neg ativ e Co r r elatio n

r = -0.68 R2 = 46% 0

r = -0.04 R2 = 0%

r = -0.93 R2 = 86% 5

10

0

5

10

Figure R-28: Scatter Plots with Corresponding Pearson’s Correlation Coefficient (r) and Coefficient of Determination (R2) Values

K L M N O P Q R S T U V W X Y Z

• Interaction—Defined as a when the response of one variable is dependent on another, and depicted mathematically as the product combination of those variables (multiplied). Two-way interaction combines two variables, three-way is for three, and so on. (See Also “Analysis of Variance (ANOVA)—7M Tool,” and “Design of Experiment (DOE),” p. 142 and p. 250, respectively.) • Intercept (of a line)—Where the line crosses (or intercepts) the y-axis, and is represented by the constant b in the linear equations Y = mX + b. • Interpolation—In a stable process, the model uses the range of two known input variable data values to predict a third unknown independent variable and its response variable. It presumes the unknown independent variable falls within the range of the known modeled sample data. • Linear regression—A quantitative model building tool that relates one or more independent variables (xs) to a single dependent variable (y). • Logistic ordinal regression—Describes the relationship among ordinal variables (e.g. high-medium-low, or excellent-good-bad).

Regre ssion Analysis

599

• Multiple linear regression—Models the relationship between multiple independent variables (Xs) and a dependent variable (Y). The simplest model (with one independent variable) is written as: Y = b0 + b1X1 + b2X2 + error where b is a coefficient; b1 and b2 are the respective slopes, and b0 is the intercept. • Regression line—The line of best fit drawn through a Scatter plot, as determined by the regression model. • Regression model—A statistical tool that measures the strength of relationship between one or more independent variables with a dependent variable.

A B C D E F G

• Residual—The observable estimate of the unobservable error. Calculated as the difference between the observed value and an estimated value (observable sample average), derived from the model’s line of best fit and the actual data points. Residuals are observable and are dependent of each other.

H

• Scatter diagram—Graphical tool displaying the qualitative strength of the relationship (linear or curvilinear) between two continuous or discrete variables; but not indicating cause-and-effect. The stronger the relationship (positive or negative), the greater the likelihood that the manipulation of the independent variable(s) may affect the dependent variable. (See Also “Scatter Diagram—7QC Tool,” p. 640)

L

I J K M N O P

• Simple linear regression—Models the relationship between a single independent variable (X) and a dependent variable (Y).

Q

The simplest model (with one independent variable) is written as:

S

Y = a0 + a1X + error where a is a coefficient; a1 is the slope and a0 is the intercept.

R T U V

• Slope—Calculated as rise over run; or y divided by x, and is represented by the constant m in the linear equations Y = mX + b.

W

• Positive slope—Indicates that the line of the curve goes from the lower-left to the upper-right corner of the graph, trending upward.

Y

X Z

600

Encyclopedia

• Negative slope—Indicates that the line of the curve goes from the upper-left to the lower-right corner of the graph, trending downward. • Variables—Two main types: A

•Dependent or response variable (Y).

B

•Independent variable (X).

C D E F G H I J

Supporting or Linked Tools Supporting tools that might provide input when developing a regression analysis include • Cause-and-Effect Diagram (See Also “Cause-and-Effect Diagram—

7QC Tool,” p. 173) • Data Collection Sheets (See Also “Data Collection Matrix,” p. 248)

K

• Sampling (See Also “Sampling,” p. 618)

L

• Scatter diagram (See Also “Scatter Diagram—7QC Tool,” p. 640)

M N

A completed regression analysis provides input to tools such as

O

• DOE (See Also “Design of Experiment (DOE),” p. 250)

P

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),”

Q R S T U V W

p. 287) • Control Plan (See Also “Matrix Diagrams—7M Tool,” p. 399)

Figure R-29 illustrates the link between a regression analysis and its related tools and techniques. Cause and Effect Diagram DOE

X Y

Data Collection Sheets

Z

Regression Analysis

FMEA

Sampling Control Plan Scatter Diagram

Figure R-29: Regression Analysis Tool Linkage

Risk Mitigation Plan

601

Risk Mitigation Plan What Question(s) Does the Tool or Technique Answer? How can you best plan for, manage, and mitigate unforeseen risk? A risk mitigation plan helps you to

A B

• Prepare for risk events—through risk identification, quantification,

response planning, and response control • Stimulate thinking when developing the list of the responses to a

potential problem • Organize the planning process and use of various risk tools

C D E F G H

Alternative Names and Variations Risk management tools include

I J K

• Cause-and-Effect diagram; Ishikawa diagram; Fishbone diagram

(See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173)

L M

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

N

• Cause categorization, checklists, or techniques such as

O

5MandP, 6Ms, and 4Ps (See Also “Cause-and-Effect Diagram—7QC Tool,” p. 173)

P Q

• 5-Whys Technique (See Also “5-Whys,” p. 305)

R

• Cause enumeration diagram (See Also “Cause-and-Effect

S

Diagram—7QC Tool,” p. 173) • CEDAC (Cause-and-Effect diagram and Cards) (See Also R

“Cause-and-Effect Diagram—7QC Tool,” p. 173) • Desired-results Fishbone (See Also “Cause-and-Effect

Diagram—7QC Tool,” p. 173) • Process Fishbone (See Also “Cause-and-Effect Diagram—7QC

Tool,” p. 173) • Reverse Fishbone diagram (See Also “Cause-and-Effect

Diagram—7QC Tool,” p. 173) • Time-delay Fishbone (See Also “Cause-and-Effect Diagram—

7QC Tool,” p. 173) • Cause and Prevention Matrix or Cause-Prevention diagram (See

Also “Cause and Prevention Diagram,” p. 198)

T U V W X Y Z

602

Encyclopedia • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),”

p. 287) • Fault Tree Analysis (See Also “Fault Tree Analysis (FTA),” p. 309) • Process Decision Program Charts (PDPC) (See Also “Process DeciA

sion Program Charts (PDPC)—7M Tool,” p. 515)

B C D E F

When Best to Use the Tool or Technique Risk planning, mitigation and planning is a process that identifies, analyzes and responds to risk, and should be done throughout all phases of a project, program and/or operational processes.

G H I J K L M N O P Q R S T U V W X Y Z

Brief Description Risk management is a planning process rather than a tool or technique in and of itself. Risk management tools are covered elsewhere in this book. Risk mitigation planning, in its broadest sense, attempts to prepare for the unknown and respond when it occurs with forethought to either maximize the results of positive events minimize the consequences of adverse events. Yes, risk equals both opportunities and threats. It involves the art and science concerned with identifying, analyzing, and responding to risk. It is part of doing business, given that uncertainty and change exist. Risk management helps to overcome situations wherein uncertainty hinders decision-making. Risk planning should focus on gathering information to proactively No Information Some combat uncertainty. This involves Information talking to both internal and exterComplete Information nal experts and reviewing historical data about past similar Amount of Risk scenarios, with the intent of Total Certainty heightening awareness as to the possible sources of risk. Figure R30 illustrates the inverse relationTotal Uncertainty ship of information and uncertainty. Information comes Figure R-30: The Inverse Relationship of from two primary sources. The Information first is objective sources of recorded experiences and data on current and past events. The second is subjective sources based on knowledgeable experts. This latter source is particularly helpful with early planning. Risk planning should be a continuous, formal activity, conducted on a regular basis, throughout a project, program, or operation.

Risk Mitigation Plan

603

Warning Less successful risk planning tends to focus on past experiences or past incidences. People tend to relate risk to things that are highly chancy or hazardous, but many risks are so commonplace we scarcely notice them. Rarely does risk planning address uncertainty explicitly or formally; however, it should so as to avoid too many surprises.

A B C D

Establishing Common Terminology Risk is defined as an element or factor that involves uncertain hazards. It connotes probability of an event and its consequences—the confidence level with which an outcome will occur. Probability ranges from the objective (informed by data or experimentation) to the subjective (relying on judgment or opinion). Certainty might be defined as having availability of all the information necessary to make the right decision, such that the outcome can be predicted with confidence. Uncertainty is the complete absence of information, where nothing is known about the possible outcomes. With the lack of information, probability distributions based on experience cannot be developed; they would be done only in the abstract. Risk management is the formal process in which risk factors are systematically identified, assessed, and controlled. It is a systematic method to identify and control risk events that potentially cause unwanted change or improve any positive results. Risk management quantifies the risk level and its impact of an action so it can be related to the organization’s tolerance level. The tolerance level may vary depending on the type of impact, for example financial, legal, regulatory, quality, safety.

Risk Management Process The risk management process consists of four phases: identification, quantification, response, and control. Sometimes risk identification and quantification are treated as a single process called risk analysis or risk assessment. Risk response is sometimes called risk mitigation. And risk response development and risk response control sometimes are treated as a single process called risk management. Regardless of how the phasing, they are iterative and may happen in parallel. To unleash the power of this process and its structure, it should become ubiquitous with the ongoing operations or work and not treated as an independent, one-time event or function. Risk Identification This phase determines which risks are likely to have the most significant impact and ascertains their characteristics. Along with identification,

E F G H I J K L M N O P Q R S T U V W X Y Z

604

Encyclopedia

documenting the essential risk characteristics and communicating them are essential elements of risk planning. The process owner typically is the person accountable for this phase and drives the refresh and renewal process to keep the information current. A

Risk identification is a systematic process involving the following approach.

B

Step 1.

C

Risk classification. Classify potential risks according to their cause or source, not the effect or impact. Understanding the potential root cause dictates the appropriate countermeasures. Sources of risk vary by business type or business model. Identify and categorize appropriate sources for your business. Considerations might include

D E F G

a. External unpredictable causes: government regulations,

H

natural hazards.

I

b. External predictable causes: raw materials availability,

J

business and financial risk.

K

c. Internal (non-technical): labor stoppages, cash flow issues,

L

safety issues, poor estimates, requirement changes, poor planning, poorly defined roles or processes.

M N

d. Technical (systems, software): changes in technology,

O

design issues, operations, and maintenance issues.

P

e. Legal: licenses, patent rights, lawsuits, contractual issues.

Q

Risk classification takes on additional dimensions other than its cause. The appropriate classifications needed for your risk planning depends on factors such as your business size, scale, and scope. However, at minimum, utilize risk source as a primary classification.

R S T U

Another risk classification is business versus insurable risk. Business risk encompasses an opportunity of profit or loss, whereas insurable risk represents a chance for loss only. Given the onesided nature of insurable risk, it is also known as pure risk. Other risk classifications include level of uncertainty, magnitude of impact, and nature of risk (defined by its characteristics). Risk characteristics consist of the probability of occurrence, range of possible outcomes, expected timing, and anticipated frequency.

V W X Y Z

Step 2.

Rank risks. Define an appropriate rating system for the situation that weights or prioritizes the identified categories.

Risk Mitigation Plan

The rating system might represent an integration of the three classic FMEA scales of severity, frequency of occurrence, and detect-ability and integrate them into one risk priority number (RPN)—or the scale could rank the ability to manage effective responses. For example, consider the impact that a cumulative effect might have. If several risk events occur in conjunction, the risk impact of the combination may be more severe than the individual events alone; thus, managing this cumulative risk response could present more challenges. Step 3.

605

A B C

Identify tools and techniques.

D

Determine what additional information is needed and the best way to organize and present the content for purposes of response planning and decision-making. Outputs from risk identification are four-fold:

E

a. list of potential risk events b. risk symptoms

F G H I J

c. sources of risk

K

d. inputs to other processes

L

Risk Quantification This phase evaluates risk to assess the range of possible outcomes. It determines the probability or likelihood of occurrence of a risk event or several risk events in combination, as well as the magnitude of impact if it occurred. Its primary purpose is to determine which risk warrants a response. The person or role accountable for this phase depends on the organization and type of risk involved. More than likely, this phase’s activities would be parsed out to several people or functions to quantify and qualify the entire portfolio of risk opportunities. If that were the case, the process owner serves as the overall customer for this phase. Risk quantification collects inputs from key stakeholders as to their risk tolerances, various sources of risk, potential risk events, cost estimates, and event duration estimates. This phase uses several tools and techniques such as statistical data, expected monetary value, decision trees, simulation, and expert judgment. Given the emphasis on probability in this phase, statistics plays a critical role. Descriptive statistics (mean, median, mode, range, variance, standard deviation) help to describe critical parameters. Moreover, statistical concepts such as normal distribution and Monte Carlo simulations add a depth of perspective to the probability of an event or set of events. A Monte Carlo simulation models the behavior or performance of a system and produces a statistical distribution of expected results. Financial Quantification: The expected monetary value (EMV) of a risk event is calculated by multiplying the probability of occurrence times the

M N O P Q R S T U V W X Y Z

606

A

Encyclopedia

estimated financial gain or loss if that event occurs. Rather than simply relying on a deterministic estimate (one number), a Monte Carlo simulation provides more information with a range of probable outcomes. (See Also “Monte Carlo Simulation,” p. 431 and “Selecting Project Portfolios Using Monte Carlo Simulation and Optimization,” in Part III, p. 921 for more detail on the technique and its various applications.)

B C

Hints and Tips

D

A way to explore the impacts of a decision and its multiple paths of

E

possibilities is to use a decision tree. It is not as robust as a Monte

F

Carlo simulation, but it does diagram key interactions among deci-

G

sions and associated chance events as understood by the decision-

H

maker and calculates the expected monetary value (EMV) of the

I

different paths. Each branch of the tree represents either a decision

J

or chance event, wherein the tree, in aggregate, represents the sum

K

total of 100% of the possibilities, as shown in Figure R-31.

L

500 records x 0 .96 x $ 2,000 per test = $ 960,000

M

Pass

N

9 6%

O P

$6.42M = $ 5 M + $ 960 K + $ 460 K

Q R S T U V

500 records x 0 .04 x $ 23,000

ST

4% F per repair = $ 460 ,000 ailu re

00

0,0 $1 5M sX =$ d r d r o ec reco 0r 50 ost / c

TE

Test or Don’t Test 100% of the records ?

Do

$7M = $0 + $7M

Fail 500 records x 0 .96 x $ 0 per test = $ 0

Pass

NO TT es

t

9 6% 500 records x 0 .046 x $350,000 per repair & reinstall = $7,000,000

W X

4% F ailu re D ecision Tree Flow

Y Z

.

Fail C ost C alculation

.

Figure R-31: Decision Tree Example

Schedule Quantification: A schedule may represent either business risk or insurable risk. Two simple tools to assist in schedule risk calculation are the Activity Network Diagram (AND) and PERT (with its most likely, pessimistic and optimistic values to triangulate a quantification estimate).

Risk Mitigation Plan

607

A more sophisticated tool available to understand the probability of a range of schedule outcomes is the Monte Carlo simulation. (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127; “PERT (Program Evaluation and Review Technique) Chart,” p. 453; and “Monte Carlo Simulation,” p. 431) Risk Qualification: If risk quantification is difficult, a qualification rating system should be developed using an ordinal scale. Develop a qualification risk rating system appropriate for your business. An example of a qualification system is as follows: • High—Risk represents significant probability of cost, schedule, qual-

ity, and/or performance disruption. • Medium—Risk represents moderate probability of disruption, and

the difficulties could possibly be overcome. • Low—Risk represents a small disruption potential, and if it

occurred, the resulting difficulties probably could be overcome. Outputs from risk quantification are two-fold: 1) opportunities to pursue (or those threats to respond to) and 2) opportunities to ignore (or those threats to accept).

Risk Response This phase determines the appropriate course of action for each individual risk opportunity of interest. Risk planning often calls for a prioritization of risk opportunities, based on the quantification and qualification of risk. Those prioritized risks at minimum receive the benefit of subsequent response planning. Typically the process owner is the person accountable for this phase, to ensure ongoing management and monitoring of appropriate response activities. Risk response falls into three main categories: • Risk Avoidance—The elimination of a specific threat, usually by

eliminating the cause. • Risk Mitigation—The reduction of the expected monetary value of

a risk event by minimizing the probability of occurrence and/or the risk event value. Another mitigation approach is the transference of all or a portion of the risk to another party (outsourcing, partnering, insurance are examples). Insurance transfer examples include: • Direct property damage (for property such as auto collision,

equipment coverage, fire insurance) • Indirect consequential loss (contractor protection for debris

removal, equipment replacement)

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

608

Encyclopedia • Legal liability (such as design errors, workplace injuries, per-

formance failures) • Personnel (worker’s compensation, employee replacement

costs) A B C D

• Risk Acceptance (or Retention)—The acceptance of the conse-

quences of a risk event by either actively or passively allowing it to occur without intervention. Active acceptance may result from developing a contingency plan. The consequence likely would either be positive or moderately negative.

E F

Note

G

Idea generation tools and techniques help to keep the response plan-

H

ning creative and fresh. Brainstorming techniques and interviewing

I

of experts can yield appropriate potential risks and corresponding

J

response approaches. A powerful interviewing method utilizes a

K

panel of experts to develop convergent solutions, called the Delphi

L

technique, and often convenes periodically over a period of years.

M

Other techniques to explore are procurement strategies, contingency

N

planning, alternative strategy planning, and insurance strategies.

O P Q R S T U

Key questions to ask about the possible risk events that warrant response planning include • How could this risk be avoided? • Can the risk be reduced?

V

• Can the risk be shared or transferred (to insurance for example)?

W

• Should the risk be accepted, and if so, does it warrant a financial

X Y Z

reserve or schedule allowance or contingency? • Can the risk be contained?

Outputs from risk response development may include: 1) risk management plan, 2) inputs to other processes, 3) contingency plans, 4) reserves, and 5) contractual agreements. A risk management plan documents the procedures to manage risk. It covers the roles and responsibilities associated with managing risk and managing its evergreen process. It defines how contingency plans will be implemented and how reserves are to be allocated, if necessary.

Risk Mitigation Plan

Risk Control This phase responds to changes in risk and keeps the risk planning process alive. This phase involves executing the risk management plan to respond to a given risk event. The approach used to manage risk should be documented and communicated widely. Risk control monitors the environment for change and information both internally and externally and responds to changes in risk as it occurs. Typically the process owner is the person accountable for this phase, to ensure ongoing management and monitoring of appropriate response activities. Even if the monitoring, reporting, and controlling tasks are delegated, the process owner provides the direction, validates its necessity, and energizes the process. Workarounds are a type of risk response control. They are unplanned responses to negative risk events. This type of response is not defined in advance of the risk event occurring, but it may trigger a best practice that would be important to document and communicate. Additional response development is another important technique that reinforces the importance of actively managing and monitoring for information and changes. It is triggered when unanticipated risk events occur or when the effect of the event exceeds the expected magnitude. A supporting tool to this technique is the Process Decision Program Charts (PDPC). (See Also “Process Decision Program Charts (PDPC)—7M Tool,” p. 515) There are a host of tools available to aid in the various stages of risk mitigation planning. Some of the soft tools are featured as entries in this book and also are found on the preceding list of variations. In addition to those soft tools, mostly matrix-like tools, there are several software application tools to assist with risk management. Outputs from risk response control are two-fold: 1) corrective action (implementing contingency plans or workarounds) and 2) updates to the risk management plan.

609

A B C D E F G H I J K L M N O P Q R S T U

Hints and Tips How do you know when to take a risk? Consider the following dimensions to help answer the question: •Quantify the risk level. •Determine its impact (magnitude)—positive or negative. •Define the risk tolerance level for the organization.

V W X Y Z

610

Encyclopedia

Additional Resources or References • A Guide to the Project Management Body of Knowledge. Pennsylvania:

Project Management Institute, Inc., 2000 (for more on risk management). A B C

• U.S. Defense Acquisition University, Acquisition Community Con-

nection—for a list of Risk Management tools (Risk Management Systems; Standalone Tools; Analysis Tools; Quick Links to other Risk Topics); https://acc.dau.mil/CommunityBrowser.aspx

D E F G H I

Rolled Throughput Yield See Also “Process Capability Analysis,” p. 486

Run Chart—7QC Tool

J K L M N O P Q R

What Question(s) Does the Tool or Technique Answer? How does the data look over time? Are the data randomly distributed over time? Does the process look stable and random over time? A run chart helps you to • Graphically display a data set to detect trends, shifts, cycles, or

extreme values in the data • Determine if the process is stable over time and randomly

distributed

S T U V W X Y

Alternative Names and Variations This tool is also known as • Time Plot; Time Series plot

Variations on the tool include • Line graph

Z

When Best to Use the Tool or Technique A Time Series plot should be one of the first graphical tools used to display the key output variable(s) in the sequence it was produced to examine if the process appears to be stable over time and if the data are randomly distributed over time.

Run Chart—7QC Tool

611

Use a run chart versus a control chart when the data set contains less than the 20 data point minimum for a control chart.

Brief Description A run chart displays data in a time sequence by placing the units of measure on the Y-axis (vertical axis) and time on the X-axis (horizontal axis). It is a simple, easy to create, easy to read line chart of continuous or discrete data representing a product or service characteristic. The graph plots the process output data as it is produced in the same sequence. Thus, it reveals any trends, shifts, cycles, or extreme values in the data over time— the absence of which indicates process stability and randomly distributed data. The run chart’s trend line represents the process data. Sometimes it features a second line to represent the average of the data for that specific period of time. This average line serves as a reference point to identify any data patterns. Essentially a run chart is a limitless control chart (no control limits). A trend or shift may indicate a process influx, either improving or degrading, and need to be brought under control. Extreme values may indicate a special cause event needing to be eliminated. The cyclical nature may indicate different work procedures from shift-to-shift, or operator-to-operator, or location-to-location. Regardless, eliminate data patterns to steady the process. Utilization of other statistical tools such as process capability and Hypothesis testing assume process stabilization as a prerequisite. The run chart is a member of the revised 7QC Tools, (or seven “Quality Control” tools), attributed to Dr. Kaoru Ishikawa. The 7QC Tools sometimes are called the seven basic tools given that they were the first set of tools identified as the core quality improvement tools. Ishikawa’s original 7QC Toolset includes: Cause-Effect diagram; Check sheet (or checklist); Control charts; Histogram; Pareto chart; Scatter diagram; and Stratification. Recently, the 7QC Toolset was modified by substituting the Stratification technique with either a flowchart (or Process map) or a run chart (or Time Series plot).

A B C D E F G H I J K L M N O P Q R S T U V W X

How to Use the Tool or Technique Use the following procedure to develop a run chart using MINITAB. Note that MINITAB refers to a run chart as a Time Series plot. Step 1.

Given that numeric data has been entered into the MINITAB Worksheet, select the following commands from its drop-down menu: Graph > Time Series plot…. Figure R-32 displays sample MINITAB screens to create a run chart, which is displayed in Figure R-33.

Y Z

612

Encyclopedia

Step 2.

In the Time Series plot main screen, select the appropriate graph type (“Area 1”) and click the OK button. In this example, the simple plot was selected.

Step 3.

The Time Series plot—Simple screen opens. Select the appropriate process variable data of interest (Yield, for example), shown in “Area 2” of Figure R-32.

Step 4.

Click the Time/Scale button to select the appropriate time scale to be displayed along the X-axis, select the column containing the X-axis time data and click the OK button, as shown in “Area 3” of Figure R-32.

Step 5.

Click the OK button on the Time Series plot—Simple screen, shown in “Area 2” of Figure R-32, and MINITAB generates a run chart. Figure R-33 illustrates a sample run chart output.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure R-32: Example MINITAB Run Chart Main Screen

Run Chart—7QC Tool

613

Time Series Plot of Yield 95

Yield

90

A B

85

C D 80

E F

75 1/1

1/6

1/12

1/18

1/24

1/30

2/5

2/11

2/17

2/23

Date

Figure R-33: Example MINITAB Run Chart

Based on the Figure R-33 graph, the data appears to be randomly distributed over time, thereby indicating stability.

G H I J K L M

Hints and Tips Run charts are simple to create and read. Use them as a preliminary quick view of process stability and randomness, particularly when the data set contains less than the required 20 data point minimum for control charts. Run chart data must maintain the same time sequence as to when it was produced and observed. Run chart data provides less information than a control chart. A control chart’s upper and lower control limits enable the calculation of different out-of-control statistical tests and the standard deviation. The confirming tests of stability and randomness require statistical analysis, such as that used in control charts. If possible, use a control chart when analyzing the performance of a process. (See Also “Control Charts—7QC Tool,” p. 217)

N O P Q R S T U V W X Y Z

614

Encyclopedia

Supporting or Linked Tools Supporting tools that might provide input when creating a run chart include • Data gathering plan to collect the appropriate metrics. (See Also A

“Data Collection Matrix,” p. 248)

B

• Performance charts and dashboards.

C

A run chart can provide input to tools such as

D E

• Control charts (See Also “Control Charts—7QC Tool,” p. 217)

F

• Cause-Effect diagrams (See Also “Cause-and-Effect Diagram—7QC

G H I J K L M

Tool,” p. 173) • Histograms and other graphical tools (See Also “Histogram—7QC

Tool,” p. 330 and “Graphical Methods,” p. 323) • FMEA with follow-on action planning (See Also “Failure Modes

and Effects Analysis (FMEA),” p. 287) Figure R-34 illustrates the link between a run chart and its related tools and techniques.

N O

Control Chart

P Q R S T U

Data Gathering (metrics) Run Chart Performance Charts and Dashboards

V

Y Z

Histogram and other graphical tools

FMEA

W X

Cause-Effect Diagram

Figure R-34: Run Chart Tool Linkage

7M—Seven Management Tools

615

S 7M—Seven Management Tools A

What Question(s) Does the Tool or Technique Answer? What tools can you use to analyze qualitative data?

B

This categorization (or acronym) helps you to double-check which tools apply to qualitative data.

D

C E F

Alternative Names and Variations This category is also known as • 7MP or seven management and planning tools

Variations on the category include • 7QC Tools, or seven quality control tools

G H I J K L

When Best to Use the Tool or Technique To recall what fundamental quality tools are used to analyze, organize, manage, and plan using qualitative data.

M N O P

Brief Description In the mid-1970s, the JUSE (Union of Japanese Scientists and Engineers) assembled a suite of seven management tools, also known as the 7M Tools, to ensure that projects comprehended qualitative information. The toolset focused on creativity, communication, and planning with qualitative data. Most of the 7M Tools are attributed in part to Dr. Shewhart who introduced and popularized many of the “traditional quality tools” to analyze quantitative data. The 7M Toolset includes • Activity Network diagrams or Arrow diagrams—A technique that

depicts the process flow to evaluate the time needed to complete a project. (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127) • Affinity diagrams—A technique that dissects and organizes verbal

data into its natural groupings to summarize themes and help to make the input actionable. (See Also “Affinity Diagram—7M Tool,” p. 136) • Interrelationship digraphs or Relations diagrams—A diagramming

tool used to plot a complex situation and to parse it into its natural cause-and-effect relationships. (See Also “Interrelationship Diagram— 7M Tool,” p. 369)

Q R S T U V W X Y Z

616

Encyclopedia • Matrix diagrams—A set of different shaped grids that display rela-

tionships between and among different factors to align, dissect, and rank. (See Also “Matrix Diagrams—7M Tool,” p. 399) • Prioritization Matrices—Often replacing the more complex Matrix A B C

data analysis. A set of matrices that rank relationship strengths to aid in the selection process. (See Also “Prioritization Matrices—7M Tool,” p. 470) • Process Decision Program Charts (PDPC)—A risk mitigation tech-

F

nique that charts what might go wrong during a development project and the corresponding countermeasures to prevent or mitigate the impact. (See Also “Process Decision Program Charts (PDPC— 7M Tool,” p. 515)

G

• Tree diagrams—A technique that displays the hierarchical relation-

H

ship among topics. (See Also “Tree Diagram—7M Tool,” p. 712)

I

The Quality Toolbox, authored by Nancy Tague, presents the 7M Tools ranked from those used for abstract analysis to detailed planning: Affinity diagram, Relations diagram, Tree diagram, Matrix diagram, Matrix Data analysis (commonly replaced by a more simple Prioritization Matrix), Arrow diagram, and Process Decision Program Chart (PDPC).

D E

J K L M N

7QC—Seven Quality Control Tools

O P Q R

What Question(s) Does the Tool or Technique Answer? What basic tools best assure quality control?

S T U V W X Y Z

Alternative Names and Variations This category is also known as • 7QC or seven quality control tools • Seven basic tools

Variations on the category include • 7M or seven management tools • 7MP or seven management and planning tools

7QC—Seven Quality Control Tools

617

When Best to Use the Tool or Technique To recall the minimal set of tools used to analyze, improve, and maintain quality control.

Brief Description Kaoru Ishikawa purported that “…as much as 95% of quality related problems…can be solved with seven fundamental quantitative tools.” Ishikawa, the creator of the Fishbone diagram (sometimes referred to by his last name), promoted a set of seven basic tools to ensure quality control. While the list of tools are rather simple, in aggregate they are powerful in analyzing, improving, and maintaining quality control. The original 7QC Toolset includes

A B C D E F

• Cause-and-Effect diagram (or Fishbone, or Ishikawa diagram)—

It is used to analyze, organize, and illustrate the cause-and-effect relationship to a problem. (See Also “Cause-and-Effect Diagram— 7QC Tool,” p. 173) • Checklists or Check sheets—They serve as simple vehicles to record

data that provides a snapshot of the frequency patterns and to remind people of standard operating procedures. (See Also “Checklists—7QC Tool,” p. 204) • Control charts—They are used to monitor how a process operates, if

it behaves consistently and predictably, or if any special cause variation events occur. (See Also “Control Charts—7QC Tool,” p. 217) • Histogram—A graphical tool displaying frequency to investigate

centeredness (location), variation, and shape of the data. (See Also “Histogram—7QC Tool,” p. 330) • Pareto chart—A graphical tool that ranks causes from greatest to

G H I J K L M N O P Q R S T

least magnitude, to analyze the percent contribution of one item over another. (See Also “Pareto Chart—7QC Tool,” p. 445)

U

• Scatter diagram—A graphical tool that displays the relationship

V

between two variables (X and Y). (See Also “Scatter Diagram—7QC Tool,” p. 640) • Stratification*—A technique that culls out the different sources of

data to better evaluate patterns. (See Also “Stratification—7QC Tool,” p. 697) See Also “Process Map (or Flowchart)—7QC Tool,” p. 522 and “Run Chart,” p. 611 * The 7QC tools often now replace the Stratification technique with either Flowchart (for example, Process map) or run chart, depending to the type of data or topic that is of interest.

W X Y Z

618

Encyclopedia

Sampling What Question(s) Does the Tool or Technique Answer? What is the best method of collecting a representative sample of the population? A B C

Sampling helps you to • Collect a portion of all the data

D

• Use that portion of the data to make conclusions

E

• Save on time, resources, and money

F G H I J K

When Best to Use the Tool or Technique Use sampling when the population size is too large to study each member, but some decisions or conclusions are needed. Sampling saves on time, money, and resources versus studying each member of an entire set of data from a population.

L M N O P Q R S T U V W X

Brief Description Sampling is an efficient and effective alternative to looking at all the data. The term population represents all items or observations of a particular group of interest. A sample is a representative subset used to estimate a larger group—either a population or process. Studying the entire population can be expensive, take time, and require resources. If the group is particularly large, just because of its sheer size, studying it in total may be difficult or near impossible (that is, the entire population of the United States or China). Depending on the type of study being conducted, the test may need to destroy the item (destructive testing), such as taste tests and some quality tests such as a car crash-test to check the safety mechanisms. Destructive testing is expensive and consumes much of the tangible item to be tested, and so is thereby prohibitive to conduct on the entire population. In summary, general reasons behind sampling involve

Y

• Economic factors

Z

• Time factors • Population size • Partially inaccessible populations and/or processes • Destructive nature of the observation • Accuracy

Sampling

Another advantage in sampling is that the averages from several samples will tend to distribute normally. This describes the Central Limit Theorem, which when plotting the sample averages produces a predictable normal distribution that serves as a base assumption for a large majority of the popular statistical tools. (See Also “Statistical Tools,” for a discussion on the Central Limit Theorem, p. 684) Sample characteristics can be studied and analyzed and used to infer or approximate a population characteristic, called a parameter. Sound conclusions can often be drawn from a relatively small amount of data. Given appropriate sampling techniques and sample size, the sample information can accurately portray a population. Use of both descriptive and inferential statistics describe and relate a sample and its population. Inferences are made with a certain degree of confidence, referred to statistically as a confidence interval. Confidence intervals support the inferred conclusions about a population. (See Also “Hypothesis Testing,” p. 335, for more discussion on confidence intervals; and “Statistical Tools,” p. 684, for a discussion on descriptive and inferential statistics.)

Representative Sampling Constructing a sample that is representative is the first priority. For conclusions to be valid, samples must be representative of the population or process. Representative means that the sample’s characteristics reflect that of the population’s. If the focus is a subset of characteristics within a population, the representative sample reflects all of those characteristics. There should be no systematic differences between the data in the sample and the data not in the sample (that is, inaccuracy of exit polling data in recent elections).

619

A B C D E F G H I J K L M N O P Q R

A representative sample requires planning. A data collection plan helps organize sampling activities, ensuring that all of the needed data is gathered and ensuring data integrity. Consider the following questions when planning to collect samples:

U

• Are there natural groupings in the data? If so, what proportion of

V

each group should be sampled?

S T

W

• When is it best to sample?

X

• How frequent should you sample?

Y

• Where is the appropriate place to sample?

Sampling involves an understanding of the types of data needing to be collected to ensure data integrity. Data integrity requires attention on collecting random, unbiased data, adhering to the time sequence of the data production or observation to capture and record the data in that order. If

Z

620

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

the item’s characteristics change over time, record the measurement or classification as quickly as possible after production and again after the stabilization period. Inspect for data recording, entry, or transfer errors. Validate the operational definition. Calibrate and analyze the measurement system and select the appropriate sampling technique to ensure accuracy, precision, and unbiased data. (See Also “Data Collection Matrix,” p. 248 and “Measurement System Analysis (MSA),” p. 412 for discussion on operational definition, accuracy, precision, and bias.) Appropriate sample size addresses reliability. The larger the sample size, the more reliable the sample. Sample size will be discussed later in this entry. A sampling plan involves consideration about the technique of extracting samples from a population. Those techniques vary by circumstance— the population to be studied and what is known about it. A sampling plan attempts to avoid sampling error. Sampling error results from an issue of validity (issues of bias), not reliability. Sampling strategies include the following techniques.

Population Versus Process Sampling Valid conclusions require a representative sample of the population or the process. However, population sampling and process sampling have different objectives and approaches. The sampling strategy depends partially on whether the population is static or dynamic. (The term static refers to a relatively fixed number of members in the group, such as the number of annual flights per airline in and out of Chicago O’Hare airport; whereas, dynamic refers to a frequently changing number, such as the human population.) Dynamic refers to a process; wherein the population is an outcome of a process. Static populations require considerations of appropriate sample size. Sample size formulas use assumptions based on a static population and what is known about it. See the “sample size”later in this section. Most sampling applications involve sampling from dynamic processes.

Population Sampling When clear distinct boundaries can be defined to describe a population in its entirety, population sampling techniques can be used. Population sampling requires that each item contained in that population can be identified and numbered. Population sampling technique describes the characteristics of the items within a known population group. For example, if sampling were required of an organization’s employees or of a complete batch of cookies, population sampling could be used because the population boundaries

Sampling

621

are known, and each member of the population can be identified and counted. Figure S-1 illustrates the population sampling technique. Population

X X

X X

X X X

X X

X

X

X X

X

X

Sample

XXX

X

A B C D

X X X

E F

Figure S-1: Population Sampling

G

Process Sampling Process sampling involves a dynamic scenario wherein all units cannot be identified as some may not yet exist (that is, items that will be produced from the process tomorrow). In this case, the process sampling technique should be used. It characterizes a process so as to predict future outcomes and/or behavior and to make future improvements, if necessary. For example, a customer call center process might want to examine the total resolution time of customer calls. Figure S-2 illustrates the process sampling technique.

H I J K L M N O P

Time Plot of Process Output

Q

Yesterday Day 4

Today Day 5

Tomorrow Day 6

R

X

Ordered Process Output Y

X X

XXX XXX ??? 1

2

X 3

? X ?

4

5

6

Day

S T U V W X

Figure S-2: Process Sampling

Process stability is an important consideration whether to sample or not. Different conclusions can be drawn based on when in the process a sample is extracted. If the process were unstable, then timing determinations become difficult. Consider the different run charts shown in Figure S-3 and determine when best to sample in the four process scenarios illustrated.

Y Z

622

Encyclopedia

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure S-3: Run Charts of Different Processes

Depending on at which point a sample is taken, the conclusions may vary. The run charts in Figure S-3 provide an unstable picture for Processes A, B, and C. Process A shows signs of special causes, given its three spikes. A sample containing these three spiked observations would give an inflated mean and standard deviation. Conversely, if the sample did not include any of those three spiked points, the mean and standard deviation would be deflated and not be representative of the entire process scenario. Process B contains a shift at observation 16. Conclusions drawn for a sample taken in the first 15 observations would be vastly different from a sample taken after the last 15 observations. Process C portrays an upward trend over time. Conclusions again would depend on where the samples are taken. Process D displays a stable, random process; it appears predictable within a band of natural ebb and flow variation. Various sampling methods would provide a representative sample from Process D.

Sampling Methods Sampling approaches fall into four main categories—random sampling, stratified random sampling, systematic sampling, and subgroup sampling. Both random and stratified random sampling techniques apply to population sampling. Process sampling methods use either systematic or subgroup sampling. Furthermore, market research also uses quota

Sampling

623

sampling and self-selected sampling techniques. The different scenarios determine which method best ensures optimal representation. The sampling technique defines the different kind of samples produced.

Random Population Sampling Random sampling assumes a homogeneous population. Regardless of which item is selected, it poses similar characteristics of interest as the remaining items in the population. Random sampling requires that each population item has an equal likelihood of being selected. Sampling must be representative and not just the result of what was easy to obtain. Often emphasis is placed on the mechanics of using a sampling plan and not on the identification and selection itself. However, both aspects of sampling are important. Sampling without randomness compromises a good sampling plan. Statistically, random sampling has a different definition than haphazard. Haphazard means at the whim of the sampler. If the sampler determines the sampling process, it contains inherent bias. Examples of inherent sampling include closing your eyes and throwing a dart or pointing, picking a number out of your head, sampling what is convenient. Random sampling avoids sampling bias. It employs the use of a statistical random number table or random number generation software that either produces a table or selects data randomly. An example of how to create a random numbers table in MINITAB can be found in the subsequent “How To Use the Tool or Technique” section of this entry.

Stratified Random Population Sampling Stratified random sampling takes into account a stratified population, with a distinctive mix of segregated characteristics. The characteristics of interest determine subgroups within the population. This is a sophisticated technique that requires careful consideration of sample size. Use this technique when the sampling needs to preserve the relative proportion of each distinct subgroup to the population. This technique deliberately samples a proportionate number of items from each sub-population randomly. The random sampling technique alone tends to arrive at the same result; however, the stratified random sampling technique ensures it. Figure S-4 compares the difference between random sampling and stratified random sampling concepts.

Systematic Process Sampling Systematic sampling selects items according to a pattern rule. The sample selection follows a non-random procedure, such as pick every tenth item produced in a process. Notice only a single item is selected at a time. The key is that the selection maintain the relative sequence as the process’ occurrences or production. This relates to manufacturing processes as well as service processes. For example, every fourth customer caller could receive a satisfaction survey after contacting a customer call center.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

624

Encyclopedia Random Sampling Population X X X X

A B C D E F G H I J K

X

X X X X X

Stratified Random Sampling Population B

Sample

X

X X X X X X X X X X X X X

A A A

XXXXXXXX

C

One third of the population is sampled

N O P Q R S T U V W X Y Z

Sample C

C C C C CC C C C C C C C

ABBCCCCC One third of each subgroup is sampled

Figure S-4: Comparison of Random and Stratified Sampling

Process knowledge and sampling needs (size of sample and throughput of process) define the pattern rule. Systematic sampling ensures a sample is representative in the presence of process drift (frequency of sampling needs to be often enough to capture drift). Thus, systematic process sampling can be as good as, or even better than, a random sample. Figure S-5 illustrates the systematic process sampling method. Time-Ordered Process Output 1

2

3

4

5

6

7

L M

B B B B B

Sample 147 Sampling every third item

Figure S-5: Systematic Process Sampling Method

Subgroup Process Sampling is very similar to the systematic process sampling method, but instead of subgroup sampling one item at a time several samples are taken from a batch or subgroup. This approach often is used in high-volume processes, such as the manufacturing of pharmaceutical tablets. Subgroup sampling involves the same size sample from each batch, and selects the samples at predetermined regular time intervals. Subgroup sampling relies on the assumption that a rational subgroup exists. A rational subgroup requires that the items occur close in time proximity to one another. The items within a subgroup tend to be more similar to one another than items between subgroups. For example, a high volume commercial bakery bakes several cookies in a week. Sampling rational subgroups of cookies could be defined by when a batch’s ingredients were mixed to test for thoroughness of mixing, or when the cookies were baked to check the conveyor belt timing through the ovens and test for doneness. In a pharmaceutical plant, sample a set number of tablets every half hour to measure the mass and compare to the product requirements.

Sampling

Quota Sampling This is a form of convenience sampling used by some survey techniques such as door-to-door, mall surveying, or election polling after voting. Quota sampling empowers the person responsible for collecting the survey data to select the potential survey subjects. Each surveyor is assigned a quota of surveys to complete. The potential subject may decline; however, the selection as to when to survey and who to survey within a set of guidelines or restrictions, are left up to the convenience of the surveyor. Different data collectors may employ different collection techniques. To prevent bias, the more data collectors the better. Unfortunately, what the surveyors perceive as convenient tends to be more similar than not, leading to issues of bias. This approach should be used with caution, and the guidelines should be carefully designed and communicated to the survey collectors. Self-selected Sampling This is a form of convenience sampling wherein the subjects decide whether or not to participate in the sampling process. This technique is used frequently with either inbound or outbound call centers. For example, if a customer places a phone order or asks for help, at the end of the call, the caller may choose to participate in a satisfaction survey (inbound); or if the recipient of an unsolicited call agrees to participate in a survey (outbound). In general, the more self-selection inherent in the sampling process, the more potential for bias. In the case of Internet and television polling, they employ self-selection sampling and have the inherent bias of the person first selecting to watch the program or go to that web site in the first place, before deciding whether to participate in the survey. A large sample size might help to minimize some bias; however, proceed with caution when drawing conclusions and inferring the findings onto a population. Some say these polls are “not scientific.”

Sample Size One of the most frequently asked question of statisticians is, How many measurements are required to have a statistically representative sample of the population? Surprisingly to many, there is no correct answer because often the answer is a matter of tradeoffs. The size of the population and what is known about the population (What kind of data can be collected? Are there subgroups?) factor into the calculation of sample size. The tradeoffs factor in risk and economics. Many of these considerations must be determined in advance of calculating the sample size. Sample size does matter. Because sample data is used to make inferences about populations, how is the validity of that projection checked? Sample size, denoted by n, is a key factor. Large sample sizes minimize inference error. However, new students of statistics typically are surprised to learn that there is no one correct answer when it comes to sample size.

625

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

626

A B C D

Encyclopedia

Sample size calculations are structured to minimize risks and maximize the ability to find useful differences. In addition, more is always better when it comes to sample size. However, the economics of time and money play a part in tempering the size of a sample. How many samples are needed depends on the required confidence level (related to Type I and Type II Errors) and the type of data available (variable data is measured, attribute data is counted). In general, various tradeoffs are considered among four factors that affect sample size, which are • Standard Deviation (denoted by s)—A direct relationship, such that

E

as standard deviation increases, the sample size also increases. This tradeoff relates to the value of the population standard deviation.

F

• Detectable Difference (known as delta or ∆)—An inverse relation-

G H I J K L M N O P Q R S T U V W X Y Z

ship, such that as the detectable difference decreases, the sample size increases. This tradeoff relates to the ability to detect a difference between values represented by the Null and Alternative hypotheses. • Type I Error (alpha or α)—An inverse relationship, such that as

alpha decreases, the sample size increases. The default value is set at an alpha of 0.05. This tradeoff relates to risk of rejecting a true hypothesis. • Type II Error (known as beta or β)—An inverse relationship, such

that as beta decreases, the sample size increases. The default value is set at a beta of 0.2.

This tradeoff represents the risk of accepting a false Null hypothesis when the alternative hypothesis is true. • Alternately, this can be stated as a direct relationship by using Power, which is defined as 1–beta, such that as Power increases, the sample size increases. (See Also “Hypothesis Testing,” p. 335 for additional information on Type I and Type II Errors.) The sampling plan examines these four factors: standard deviation, detectable difference (or delta), alpha, and beta to determine the sample size prior to beginning the sampling process. This is called prospective planning. Using historical data requires a retrospective sample size calculation. Much of Hypothesis Testing involves a retrospective scenario, wherein assessing the Power and detectable difference occurs after a sample has already been collected and analyzed. Often a Hypothesis test shows no significance, and the follow-on question is asked, “What was the Power of the test?” If the Power is low, then having additional sample data would increase Power, which is the ability to detect a difference if one truly exists.

Sampling

627

Sample size is not linearly related to precision, as shown in Figure S-6. For example, a statistically significant survey of likely voters found that 45% ( +4% ) would vote for the incumbent. The sample size for this survey was 601. 10000

A

9000

B

Sample Size (n)

8000

C

7000

D

6000

E

5000

F

4000

G

3000

H

2000

I

1000

J

0 0%

2%

4%

6%

8%

10%

Precision

K L M

Figure S-6: Sample Size for Proportional Data

N

The formulas for calculating the minimum sample size (n) vary by the data type and thus the type of Hypothesis test. The various formulas include the following:

O P Q

• Normal distribution, where the standard deviation is known:

R

• Two-tail test: n = (Z

/2 +

Zβ) (σ/δ)

S

• One-tail test: n = (Z

/2 +

Zβ)2 (σ/δ)2

2

2

T

These formulas use critical values of the standard normal distribution. • Student t-distribution, where the population standard deviation is

U V W X

unknown: • Two-tail test: n = (t

/2 + tβ)

(s/δ)

• One-tail test: n = (t

2 /2 + tβ)

(s/δ)2

2

2

Y Z

628

Encyclopedia

Recall that the critical values for a t-distribution (tcritical) rely on knowing the degrees of freedom. And degrees of freedom depend on the sample size, which it is yet to be determined. • Estimate average daily yield of a chemical or mixture process with a A B C D E

95% confidence interval (two-tail test) involves three equations. Yield avg. − δ δ=

F G H I J K L

n

µ σ n

Yield avg. + δ

Z0.025

1.96 δ

2

σ2

Tables exist to look up the minimum sample size depending on the type of Hypothesis test. However, statistical software packages with these various formulas already embedded make it easy for one to determine the minimum sample size with simple what if scenarios. The following “How to…” section describes calculating sample size using MINITAB.

M N O P Q R S T U V W X Y Z

How to Use the Tool or Technique The next set of procedures use MINITAB’s Power and Sample Size capabilities and different Hypothesis tests to answer questions about sample size or any of the other four factors that affect sample size. (See Also “Hypothesis Testing,” p. 335) The process to calculate sample size in MINITAB basically follows a similar “what if” scenario for the various Hypothesis tests. Enter the known summary information into the appropriate dialog boxes for the four sample size factors (standard deviation, detectable difference, alpha, and beta), leave the sample size dialog box empty, and MINITAB provides the minimum sample size. If sample size is known along with any three out of the four dependent factors, the fourth factor can be calculated.

Determine Sample Size: 1-Sample t-Test Example scenario background: You are responsible for a soft drink bottling process. Your current equipment operates at a throughput of 600 bottles per minute on average. There is a new model coming on the market that touts a throughput of 700 bottles per minute. A business case

Sampling

shows that at that speed it would be cost effective to make the purchase. You plan to go to the supplier to verify the throughput on your product. Based on your current equipment, the machine-to-machine variability (standard deviation) is 50 bottles per minute. How many machines would you request the supplier to demo for you to be confident of this new capital purchase? Assumptions:

629

A B

• Question: What is the sample size of new machines to demo? • Differences = 700–600 bottles per minute = 100. • Power Values = (1-Beta); in this example the default is used;

(1–0.2) = 0.8.

C D E F G

• Standard Deviation (s) = 50 bottles per minute.

H

• Significance Level (Alpha) = 0.05.

I

• Alternative hypothesis = greater than; assume that the new machine is

J

better, so conduct one-sided test. MINITAB procedure: Step 1.

Step 2.

From MINITAB’s main drop-down menu, select the following sequence of commands, Stat > Power and Sample Size > 1Sample t…

K L M N O P

Enter the assumption data into the appropriate dialog boxes, as illustrated in Figure S-7.

Q

a. In the main screen of the Power and Sample Size for 1-

R

Sample t, enter the following three items—Differences, Power values, and Standard deviation. Keep empty the Sample Sizes dialog box.

T

b. Click the Options button to open the Power and Sample

Size for 1-Sample t Options window. Select the appropriate Alternative hypothesis. Enter the appropriate Significance Level in its dialog box. Click the OK button. c. Click the OK button on the main screen.

S U V W X Y Z

630

Encyclopedia

A B C D E

Figure S-7: MINITAB Screens for 1-Sample t-Test Sample Size

F G H

Figure S-8 provides the MINITAB output in its session window, using the soda bottle example data.

I J K L M N O P Q R S T U V W

Figure S-8: Bottle Example in MINITAB Session Window for 1-Sample t-Test Sample Size

X Y Z

MINITAB determined that four machines should be demonstrated. If they had a throughput greater than or equal to 700, with a standard deviation (s) no larger than 50, there is a 90% chance of concluding that the new model can achieve 700 bottles per minute.

Sampling

Determine Power: 2-Sample t-Test Example scenario background: You are considering a second supplier for pizza boxes. You have a standard test to ensure the box can hold 5lbs. of weight because the moisture, heat, and oil weaken the cardboard material. It is determined that if this critical deflection force for the new supplier differs by more than 50mm from the current supplier, there would be excessive variability in the box design. If the new supplier has 10 boxes available for testing, what would be the power for standard deviation (s) = 30mm and alpha = 0.05? Assumptions: • Question: What would be the power for standard deviation (s) = 30mm, and alpha = 0.05? • Sample Sizes = 10 boxes. • Differences = 50mm.

631

A B C D E F G H I

• Standard Deviation (s) = 30 mm.

J

• Significance Level (Alpha) = 0.05.

K

• Alternative hypothesis = not equal test.

L M

MINITAB procedure:

N

Step 1.

From MINITAB’s main drop-down menu, select the following sequence of commands, Stat > Power and Sample Size > 2Sample t…

O

Enter the assumption data into the appropriate dialog boxes, as illustrated in Figure S-9.

R

a. In the main screen of the Power and Sample Size for 2Sample t, enter the following three items—Sample size, Differences, and Standard deviation. Keep empty the Power values dialog box.

T

Step 2.

b. Click the Options button to open the Power and Sample Size for 1-Sample t Options window. Select the appropriate Alternative hypothesis. Enter the appropriate Significance Level in its dialog box. Click the OK button. c. Click the OK button on the main screen.

P Q S U V W X Y Z

632

Encyclopedia

A B C D E F

Figure S-9: MINITAB Screens for 2-Sample t-Test for Power

G H I

Figure S-10 provides the MINITAB output in its session window, using the pizza box example data.

J K L M N O P Q R S T U V

Figure S-10: Pizza Box Example in MINITAB Session Window for 2-Sample t-Test for Power

W X Y Z

MINITAB determined that for a sample size of 10 boxes, we have very good power (0.94). If the boxes differ by at least 50mm, we will have a good chance of detecting it with the standard test.

Determine Sample Size: 2-Proportion Example scenario background: A candy company makes a brand of candy that contains five different colors in one bag. It publishes the proportion of different colors of its candies on its web site. According to the company information, the proportion of blue candies is 0.24. The proportion of pink candies is 0.2. If you wanted to test the company’s assertion, how many samples would be required to detect this difference for alpha = 0.05?

Sampling

633

Assumptions: • Question: How many samples are required to detect an alpha level

of 0.05? • Proportion 1 values = 0.2 pink candies. • Power values = 0.8.

A B

• Proportion 2 = 0.24 blue candies.

C

• Significance Level (Alpha) = 0.05.

D

• Alternative hypothesis = Less than test; less pink candies than blue

E

candies.

F G H

Note If blue candies (0.24) were placed in the Proportion 1 value dialog box, and the pink candies (0.20) in the Portion 2, then the Alternative hypothesis test would need to be greater than.

I J K L M

MINITAB procedure: Step 1.

Step 2.

From MINITAB’s main drop-down menu, select the following sequence of commands, Stat > Power and Sample Size > 2Proportions… Enter the assumption data into the appropriate dialog boxes, as illustrated in Figure S-11. a. In the main screen of the Power and Sample Size for 2

Proportions, enter the following three items—Portion 1 values, Power values, and Portion 2. Keep empty the Sample Sizes dialog box. b. Click the Options button to open the Power and Sample

Size for 2 Proportions—O… window. Select the appropriate Alternative hypothesis. Enter the appropriate Significance Level in its dialog box. Click the OK button. c. Click the OK button on the main screen.

N O P Q R S T U V W X Y Z

634

Encyclopedia

A B C D E F G H I

Figure S-11: MINITAB Screens for 2-Portions Sample Size

Figure S-12 provides the MINITAB output in its session window, using the soda bottle example data.

J K L M N O P Q R S T U V W X Y Z

Figure S-12: Candy Example in MINITAB session window for 2-Portions Sample Size

MINITAB determined that a sample containing at least 1326 of each color of candy is needed to detect a difference of just 4% between the number of blue and pink candies. Blue candies represent 20% of the total

Sampling

635

in the bag. So the total number of candies needed to sample is approximately the five different colors times the required sample size, which is 6630 individual candies (5 x 1326 = 6630); now that is a lot of candy.

Determine Sample Size—Retrospective Scenario Example scenario background: A sales rep was considering two different routes for a new target list. She tested each route five times throughout the day to better understand the different traffic patterns. The mean for route 1 was 22.4 minutes. The mean for route 2 was 24.8 minutes, and the standard deviation for both was about three minutes She performed a 2-Sample t-test in MINITAB using the commands of Stat > Basic Statistics > 2 Sample t (Test and Confidence Interval)… loaded this five-run data from her two different routes, selected Summarized data, and got the following results in MINITAB’s session window, as shown in Figure S-13.

A B C D E F G H I J K L M N O P

Figure S-13: MINITAB Session Window for 2-Sample t-Test for the Sales Rep Example

Q R

The sales rep was disappointed with the results. She was hoping that one route would prove to be faster, but the T-Test of difference is zero. Moreover, the p-value is 0.242, indicating that there is insufficient evidence to reject the Null hypothesis, meaning that the two route times were indistinguishably different. (See Also “Hypothesis Testing,” p. 335 for a discussion about p-value.) Calculate the power of the test. If the times for the routes remain stable, how many additional trips would need to be made in order to show a difference with a reasonable power? Assumptions: • Question: How many additional trips are needed to indicate a differ-

ence with reasonable power? • Differences = 24.8–22.4 = 2.4 minutes.

S T U V W X Y Z

636

Encyclopedia • Power values—Try multiple alternatives by entering all three num-

bers with a space in between each = 0.7 0.8 0.9. • Standard deviation = 3.0 because times remain stable, same as before. • Alternative Hypothesis = Not equal test, hoping route 1 is faster. A B C

MINITAB procedure: Step 1.

From MINITAB’s main drop-down menu, select the following sequence of commands, Stat > Power and Sample Size > 2Sample t…

Step 2.

Enter the assumption data into the appropriate dialog boxes, as illustrated in Figure S-14.

D E F G H I J

a. In the main screen of the Power and Sample Size for 2Sample t, enter the following three items—Differences, Power values, and Standard deviation. Keep empty the Sample Sizes dialog box.

M

b. Click the Options button to open the Power and Sample Size for 2-Sample t - Options window. Select the appropriate Alternative hypothesis. Enter the appropriate Significance Level in its dialog box. Click the OK button.

N

c. Click the OK button on the main screen.

K L

O P Q R S T U V W X Y Z

Figure S-14: MINITAB Session Window for 2-Sample t-Test for the Sales Rep Example

Figure S-15 provides the MINITAB output in its session window, using the sales rep example data.

Sampling

637

A B C D E F G H I Figure S-15: Sales Rep Example in MINITAB Session Window for 2-Sample t-Test and Power

J K

MINITAB determined that a total of 26 trips were necessary to have adequate Power of 80%. The sales rep has already taken five, so there are 21 remaining trips yet to take.

ANOVA Sample Size Calculating the sample size for a variety of Hypothesis tests is similar to the preceding examples. (See Also “Analysis of Variance (ANOVA)—7M Tool,” ‘p. 142) If a three-level ANOVA test needed to be conducted, and the business required the defaults of an 80% Power level and 5% alpha level, the sample size can be calculated for different values of detectable difference.

L M N O P Q R S T

From MINITAB’s drop-down menu, select the sequence Stat > Power and Sample Size > 2-Sample t….

U

For example purposes, enter in the following data into the different dialog boxes on the Power and Sample Size for One-Way ANOVA main screen, as illustrated in Figure S-16:

W

• Number of levels = 3 • Sample sizes = (Leave this box empty.) • Values of the maximum difference between means = 20 30 40 50 60 • Power values = 0.8

V X Y Z

638

Encyclopedia • Standard deviation = 20 • Options button: Leave the default of 0.05 in the Significance level

dialog box. A B

Select the OK button, and the following session window appears, as shown in Figure S-17. Notice that there is an inverse relationship between sample size and the detectable difference; such that as sample size increases, the detectable difference decreases.

C D E F G H I J K L M N O P Q R S T U

Figure S-16: MINITAB One-Way ANOVA for Sample Size One-way ANOVA Alpha = 0.05 Assumed standard deviation = 20 Number of Levels = 3

V W X Y Z

SS Means 200 450 800 1250

Sample Size 21 10 6 5

Target Power 0.8 0.8 0.8 0.8

Actual Power 0.814770 0.817278 0.805317 0.883287

Maximum Difference 20 30 40 50

Figure S-17: MINITAB Session Window for One-Way ANOVA—Sample Size

Random Numbers Generation MINITAB can create a random numbers table. Select the following sequence from its drop-down menu: Calc > Random Data > Normal….

Sampling

Decide how many columns and rows of data you wish to be generated. Determine the mean and standard deviation needed. Select the OK button. MINITAB will populate its worksheet with the amount of random numbers requested, which yield the specified mean and standard deviation. Figure S-18 shows a sample MINITAB main screen for random number generation of normally distributed data.

639

A B

Note

C

Non-normal random numbers may also be generated by specifying the

D

appropriate parameters, not the mean and standard deviation.

E F

Select Calc > Random Data and then make the appropriate selection.

G H I J K L M N O P Q R S T U V W

Figure S-18: MINITAB Random Number Generator for Normal Distribution

Because every random number generation table would be different, an illustration of the final MINITAB Worksheet would not be useful.

X Y Z

640

Encyclopedia

Scatter Diagram—7QC Tool What Question(s) Does the Tool or Technique Answer? Are these two factors correlated? What is the relationship between these two factors? A

A Scatter diagram helps you to

B C D

• Visualize the relationship between two variables—either linear or a

curve

E F G H I

Alternative Names and Variations This tool is also known as • Scatterplot

J

• X-Y graph

K

• Correlation chart or graph

L M N O P

When Best to Use the Tool or Technique Examine the Scatterplot prior to conducting a regression analysis to visualize the relationship. If the curve reflects a non-linear curve, a linear regression will not be applicable.

Q R S T U V W X Y Z

Brief Description A Scatter diagram is a graphic display of a set of ordered pairs of numerical data points (X and Y). By convention, the first variable in the pair is the independent factor (X) plotted along the horizontal axis, and the second is the dependent variable (Y) plotted on the vertical axis. Analysis of a Scatterplot evolves around the paired X-Y’s relationship. If the data have an inverse relationship (as the X factor increases, the Yvariable decreases), then they are said to have a negative linear relationship. A negative linear relationship displays its points in a downward trending slope from upper-left to lower-right corner of an X-Y graph. If they have a direct relationship, such that both X and Y increase simultaneously, it is a positive linear relationship. A positive linear relationship displays its points in an upward trending slope from lower-left to upperright corner of an X-Y graph.

Scatter Diagram—7QC Tool

641

If X increases and Y holds fairly constant, then it is said they have no relationship. No linear relationship displays its points in a flat, horizontal, or zero slope, parallel with the X-axis of an X-Y graph. The Scatter diagram can graph both linear or non-linear relationships. It supports two statistical tools—the correlation analysis for linear relationship and the regression analysis for non-linear relationships.

Linear Relationships—Use with Correlation Analysis A Scatter diagram describes the correlation between two factors. Correlation is a metric that measures the linear relationship between two process variables. The tighter the display of data points, the stronger the relationship. A strong relationship, or correlation does not imply causation (a cause-and-effect relationship). For example, the number of car accidents on a rainy day may be strongly correlated with the sale of umbrellas, but the buying of umbrellas did not cause the accidents. Lack of correlation does not necessarily mean that no causal relationship exists. The range of data may be too narrow to detect relationship. Consider the graph in Figure S-19 if it simply revealed the center band, it would appear to display no relationship. However, if the perspective expands to encompass the entire graph, it represents a moderately negative relationship.

A B C D E F G H I J K L M

Scatterplot of y vs x 25

N

75

100

y

O

90

P

80

Q

70

R

60

S

50

T

40

U

30

V W

20 0

20

40

60

x

Figure S-19: Scatterplot with Bands

80

100

X Y Z

642

A B C D E

Encyclopedia

A poor-man’s test for a strong linear relationship is called the pencil test. Once the data is individually plotted on an X-Y graph, sit in a chair at a table or desk and place the graph on top of the flat surface. Hold up a pencil at arms-length, if the pencil covers up the dots, it is said to be a strong relationship. If dots can be seen on either side of the pencil, the relationship is moderate to weak. Other than the pencil test, another visual technique is to look in the opposite corners to see how much white space exists. The more white space in the opposite corners from the data, the more likely a strong linear relationship exists. Figure S-20 illustrates the different types of linear relationships and their strength.

F

Strength of Relationship

G

J

Moderate

None

S ca tte r p lo t o f Y _ p o s2 v s X

S ca tte r p lo t o f Y _ p o s v s X 20

24 22

18 20

Positive

Y_pos2

I

Type of Relationship Y_pos

H

Strong

18 16

16

14 S ca tte r p lo t o f Y _ n o n e v s X

14 24

12 12

22

K

10

10

12

14

16

18

20

22

10

24

L

Scatte rplot of Y _ neg v s X

M O P Q

14

16

18

20

22

24

X S ca tte r p lo t of Y _ n e g 2 v s X

20

18 16

14

18 12

18 16

10

Negative

Y_neg2

16

Y_neg

N

12

X

Y_none

20 10

14

12

10

14

12

14

16

X

18

20

22

24

12

10 10 8 8 6 10

12

14

16

18

20

22

24

10

12

14

16

X

18

20

22

24

X

Figure S-20: Scatter Diagrams with Different Relationships

R S T U V W X Y Z

There is a statistic that also determines the strength of a linear relationship between two quantitative variables, called the Pearson’s Correlation Coefficient (r). It defines the correlation metric and falls between (-1) and (1), where 0 indicates no linear relationship, (-1) describes a perfect negative correlation, and (1) a perfect positive correlation. Thus, the larger the absolute value of r (either + or -), the stronger the linear relationship. Graphically, this strong relationship is characterized by a tight distribution of data around a best-fit line plotted in a Scatter diagram. The Pearson’s Correlation Coefficient (r) is proportional to the slope of the regression line adjusted for the differences in both X and Y’s standard deviations. Its formula is as follows: n

(xi i

rxy

x )( yi

n

n

(xi i

y)

1

1

x )2

( yi i

1

y )2

Scatter Diagram—7QC Tool

643

Correlation analysis and its Pearson’s Correlation Coefficient (r) is useful in describing the strength of a linear relationship. It is a relatively simple, straight-forward, quick analysis compared with its cousin, the regression analysis. Recall that Scatter diagrams support both types of statistical analyses and graph both linear and non-linear relationships.

Non-Linear Relationships—Use with Regression Analysis The regression analysis applies to both linear and non-linear relationships. Use the correlation analysis for a simple test of magnitude for a linear relationship. However, if modeling is required to predict (extrapolate or interpolate) results, then regression analysis should be used. And for non-linear relationships, use regression and its R-squared value (R2) [for example, the square of the Pearson’s Correlation Coefficient (r)], to describe the relationships. (See Also “Regression Analysis,” p. 571, for more detail on Pearson’s Correlation Coefficient, correlation, and linear and non-linear relationships.) Figure S-21 illustrates several Scatter diagrams with their corresponding Pearson’s Correlation Coefficient (r).

A B C D E F G H I J K

0 Mo d er ate Po sitiv e Co r r elatio n

5

L

10

Str o n g Po sitiv e Co r r elatio n

r = 0.64

r = 0.92

Mo d er ate Neg ativ e Co r r elatio n

Str o n g Neg ativ e Co r r elatio n

10.0

M

7.5

N

5.0

O

2.5

P

0.0

Q

No Co r r elatio n

r = -0.03

10.0

R

7.5

S

5.0

T

2.5

U r = -0.68

0.0 0

5

Str o n g , No n lin ear Relatio n sh ip

r = -0.93 10

V

r = -0.04 0

5

10

W X

Figure S-21: Scatter Diagrams with Corresponding Pearson’s Correlation Coefficient (r)

The Scatter diagram is a useful tool because it plots either a linear or non-linear relationship. It also can depict a quadratic or cubic relationship. As a result, subsequent regression analyses yield non-linear models with

Y Z

644

Encyclopedia

additional terms in the equation. Figure S-22 compares these different relationship Scatter diagrams drawn in MINITAB. Linear Relationship

Quadratic Relationship

Cubic Relationship

A B C D

Figure S-22: Comparing Linear, Quadratic, and Cubic models

E F G H I J K L M N O P Q R S T U V W X Y Z

The Scatter chart is a member of the 7QC Tools, (or seven Quality Control tools), attributed to Dr. Kaoru Ishikawa. The 7QC tools sometimes are called the seven basic tools because they were the first set of tools identified as the core quality improvement tools. Ishikawa’s original 7QC toolset includes: Cause-and-Effect diagram; Check sheet (or checklist); Control charts; Histogram; Pareto chart; Scatter diagram; and Stratification. More recently, the 7QC toolset is modified by substituting the Stratification technique with either a flowchart (or Process map) or a run chart (or Time Series plot).

Importance of Plotting Data Before a Regression Analysis Unlike the correlation analysis, regression analysis models several different types of relationship, not simply linear relationships. Before conducting a regression analysis, it is important to plot the data using a Scatterplot, to understand what type of data relationship exists before identifying the potential regression model terms. If the relationship is linear, the model is relatively straight-forward. If the model involves quadratics, cubic, or interaction relationships, the appropriate calculations must be applied to the data before testing a potential equation. (See Also “Regression Analysis,” p. 571) Brief Scenario: We are pressed for time. We have four pair of X-Y data to examine that we believe come from the same population and don’t want to take the time to graph them prior to running a regression analysis. We have been to a Lean Six Sigma course, and we know that if samples have the same means that they most likely are from the same population. We run a quick analysis in MINITAB to check that the four sets have the same mean and the same standard deviation; and they do. (MINITAB’s Display Descriptive Statistics can be found from its dropdown menu by following this sequence: Stat > Basic Statistics > Display Descriptive Statistics….)

Scatter Diagram—7QC Tool

645

We run another quick analysis to verify that the correlation statistics are the same; and they are. (MINITAB’s Display Descriptive Statistics can be found from its drop-down menu by following this sequence: Stat > Basic Statistics > Correlation….) Figure S-23 displays the MINITAB session window containing both the Descriptive Statistics and Correlation information for this example.

A B C

Descriptive Statistics: X1, Y1, X2, Y2, X3, Y3, X4, Y4 Variable X1 Y1 X2 Y2 X3 Y3 X4 Y4

N 11 11 11 11 11 11 11 11

D

Mean StDev 9.00 3.32 7.501 2.032 9.00 3.32 7.501 2.032 9.00 3.32 7.500 2.030 9.00 3.32 7.501 2.031

All Xs have the same mean and standard deviation. All Ys have the same mean and standard deviation.

E F G H I J K

Correlations: X1, Y1, X2, Y2, X3, Y3, X4, Y4

Y1 X2 Y2 X3 Y3 X4 Y4

X1 0.816 1.000 0.816 1.000 0.816 -0.500 -0.314

Y1

X2

0.816 0.750 0.816 0.469 -0.529 -0.489

0.816 1.000 0.816 -0.500 -0.314

Y2

0.816 0.588 -0.718 -0.478

Y3 X4 TheX3 correlations among the four datasets are the same.

L M N O

0.816 -0.500 -0.314

-0.345 -0.155

P 0.817

Q R

Figure S-23: Example MINITAB session window of Descriptive Statistics and Correlation

We feel very confident about our data and launch right into running regression analyses. We would expect that the regression equations for the four X-Y pairs also would be the same; the intercepts are the same; the slopes are the same; and the Coefficients of Determination (R2) are the same. But much to our amazement, the R-square and R-square-adjusted tell us that our model is inadequate. We are getting frustrated and running out of time. In general, use the value of R-square-adjusted to assess how good a model fits. R-square increases with every additional predictor (or variable) added to the model even if the predictor adds little value to the model. (See Also “Regression Analysis,” p. 571 for more information.)

S T U V W X Y Z

646

Encyclopedia

We decide to plot the data to see if graphs display any unusual patterns in the four different data sets. Figure S-24 shows the Scatter diagrams for each set, and we find that in fact they are very different.

Scatterplot of Y1 vs X1, Y2 vs X2, Y3 vs X3, Y4 vs X4

A

5

B

Y 1 * X1

10

15

12

C

10

D

8

E

6

F G H

4 Y 3 * X3

10 8

J

6

L M N

Y 4 * X4

12

I K

20

Y 2 * X2

4

5

10

15

20

Figure S-24: Example Scatterplot of Four Data Sets with the Same Means and Correlations

O P Q R S T U V W X Y Z

How to Use the Tool or Technique Use the following procedure to develop a Scatter diagram using MINITAB. Given that numeric data has been entered into the MINITAB Worksheet, select the following commands from its drop-down menu: Graph > Scatterplot…. In the Scatterplot main screen, select Simple, for a simple X-Y plot. Click the OK button. In the Scatterplot—Simple main screen, enter the appropriate columns of X and Y data in the corresponding cells of the dialog box. Click the OK button. Figure S-25 illustrates both the MINITAB main screen and the corresponding output graph.

Scatter Diagram—7QC Tool

647

A B C D E F G H Figure S-25: Scatter Diagram—MINITAB Main Screen and Graph

I J

Figure S-25 shows a moderate positive relationship between Profit and FTE (Full-time Equivalents).

K

The same plot can be drawn with a regression (or fitted) line. From MINITAB’s drop-down menu, select the following sequence: Graph > Scatterplot….

M

L N

In the Scatterplot main screen, select With Regression for a simple X-Y plot and fitted line. Click on the OK button.

O

In the Scatterplot—With Regression main screen, enter the appropriate columns of X and Y data in the corresponding cells of the dialog box. Click the OK button. Figure S-26 illustrates both the MINITAB main screen and the corresponding output graph.

Q

Correlation Analysis—for Linear Relationships Other than calculating the Pearson’s Correlation Coefficient (r) manually, MINITAB also calculates it. Using MINITAB to determine the correlation statistic for the data plotted on a Scatter diagram, select the following sequence from MINITAB’s main drop-down menu Stat > Basic Statistics > Correlation….

U

Enter the appropriate X and Y variables in the Correlation dialog box and click the OK button to calculate the Pearson’s Correlation Coefficient (r), which MINITAB provides in its session window. Figure S-26 shows the MINITAB Correlation main screen and the session window for the same example used in Figure S-26. The Profit and FTE correlation results indicate that the correlation is positive and moderately strong with an r of 0.621 (with 1.0 being very strong) and that the two variables are statistically correlated with a p-value of 0.000.

P R S T V W X Y Z

648

Encyclopedia

A B C D E F G H I J K

Figure S-26: Scatter Diagram with Fitted Line—MINITAB Main Screen and Graph

L M N O P Q R S T

Note The Null hypothesis for correlation: H0 states that the correlation equals zero (r = 0); but in this case because the p-value is low, there is sufficient evidence to reject the H0 and accept the Alternative Hypothesis that the pair of variables are correlated. Recall that when p is low, H0 can go. (See Also “Hypothesis Testing,” p. 335, for details on p-value and Null hypothesis.)

U V W X Y Z

Figure S-27: Fitted Line Plot—MINITAB Main Screen and Graph

Scatter Diagram—7QC Tool

Fitted Line Plot or Regression Line The same Scatter diagram exhibited in Figure S-26 can also be drawn using MINITAB’s Fitted Line Plot. The main difference is that using the Fitted Line Plot feature, MINITAB also provides additional information with the graph, namely the regression equation [found under the graph’s title), the standard deviation (s), R-square (R-Sq), and R-square-adjusted (R-Sq(adj.)]. (See Also “Regression Analysis,” p. 571) From its main drop-down menu, select the following sequence Stat > Regression > Fitted Line Plot…. In the main screen, select the appropriate columns of data that correspond with the X and Y data and enter each in the corresponding dialog box. Unless the relationship between the paired data is known, select linear relationship under the Type of Regression Model option. Click the OK button. Figure S-28 illustrates both the MINITAB main screen and the corresponding output graph.

649

A B C D E F G H I J K L M N O

Figure S-28: Fitted Line Plot—MINITAB Main Screen and Graph

Subgroup Stratification MINITAB’s Scatterplot can distinguish among different subgroups to display stratification. Within the FTE category, subgroups of Region office location exist. From MINITAB’s drop-down menu, select the following sequence: Graph > Scatterplot….

P Q R S T U V

In the Scatterplot main screen, select With Regression and Groups for a stratified X-Y plot and a fitted line. Click the OK button.

W

In the Scatterplot—With Regression and Groups main screen, enter the appropriate columns of X and Y data in the corresponding cells of the dialog box. Click the OK button. Figure S-29 illustrates both the MINITAB main screen and the corresponding output graph.

Y

X Z

650

Encyclopedia

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Figure S-29: Scatterplot with Regression and Groups—MINITAB Main Screen and Graph

Figure S-29 shows that North region (indicated by the squares) has a more positive linear relationship between FTE and profit than the West region. A logical next step might be to ask why the difference in relationships exist and conduct a Cause-and-Effect analysis.

Matrix Plots MINITAB offers a useful tool for when several different data exist, called a matrix plot. A matrix plot is several individual Scatter diagrams displayed via one graph. This tool is meant for analysis purposes only and is not a good presentation tool because it can become an eye-chart in that is packed with a lot of detail, making it difficult to read quickly; it requires careful scrutiny to glean any messages. To illustrate this point, see Figure S-30. However, it serves as a quick peek at a lot of data to identify if any relationships of interest exist. A matrix plot narrows the field of possible paired relationships of interest before spending time on creating individual graphs or conducting regression analysis. To create a matrix plot, select the following sequence from MINITAB’s drop-down menu: Graph > Matrix Plot…. In the Matrix Plot main screen, select With Groups. Click the OK button. In the Matrix Plot—Matrix of Plots, With Groups main screen, enter all the columns of data that may be of interest in the dialog box.

Scatter Diagram—7QC Tool

Notice that MINITAB allows a range of data columns to be selected as indicated by a hyphen between two data column names as a shortcut, rather than listing each individual column name. To select the range, click in the first data column name in the far left window and hold the shift key down while clicking the last data column name of interest. Click the Select button, and the range of names will display in the dialog box. Enter the subgroup data name to stratify. Click the OK button. Figure S-30 illustrates both the MINITAB main screen and the corresponding output graph.

651

A B C D E F G H I J K L M N O P Q R

Figure S-30: Matrix Plot with Groups—MINITAB Main Screen and Graph

S T

Figure S-30 illustrates an eye chart; however, it is intended only as a quick and easy tool to narrow the array of potential relationships and identify some of interest for further exploration and analysis. A matrix plot is an analysis tool, not a presentation tool.

U V W X

Hints and Tips • The X values represent the independent variables that impact

the Y values. (In a Scatter plot, the independent variable is plotted on the X-axis, and the dependent variable on the

Y Z

652

Encyclopedia

Y-axis.) Be certain to properly assign the data using this convention when conducting a regression analysis. • Correlation does not mean cause-and-effect relationship. A

• If the data shows no relationship, determine if subgroups

exist and examine if stratification changes the findings. (See

B C D E F

Also “Stratification—7QC Tool,” p. 697) • Correlation equates to the strength or magnitude of a rela-

tionship between the X and Y variables. • Plot data in a Scatter diagram before running regression

analysis to understand if any patterns exist that may distort

G

the statistical analysis, as shown in Figures S-23 and S-24.

H I J K

Supporting or Linked Tools Supporting tools that might provide input when developing a Scatter diagram include

L M

• Data Gathering (See Also “Data Collection Matrix,” p. 248)

N

• Performance Charts and Dashboards

O

A completed Scatter diagram provides input to tools such as

P Q R S T

• Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram—

7QC Tool,” p. 173) • Correlation analysis (See Also “Cause-and-Effect Prioritization,”

p. 188)

U

• Regression analysis (See Also “Regression Analysis,” p. 571)

V

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),”

W X Y Z

p. 287) Figure S-31 illustrates the link between a Scatter diagram and its related tools and techniques.

Score cards

653

Cause-Effect Diagram Data Gathering (metrics)

Correlation Analysis Scatter Diagram

Performance Charts and Dashboards

A Regression Analysis

B C D

FMEA

Figure S-31: Scatter Diagram Tool Linkage

E F G H

Scorecards

I J K

What Question(s) Does the Tool or Technique Answer? What is the progress of the project and project team or the process and process players? Scorecards help you to

L M N O

• Monitor the healthiness of a task, project, process, or entire business

P

• Plan and guide the team’s work

Q

• Understand if the project improvement team members, or process

players, are using the right tools at the right time

R S T U

When Best to Use the Tool or Technique Use a system of scorecards throughout the lifecycle of the improvement project or process to monitor and manage the flow of work.

Brief Description Similar to an adventurer’s reliance on a compass, a business person’s most powerful guiding tool to ensure the business stays on course, or adheres to its plan, is a scorecard. A scorecard is the primary predictive tool for both in-process measures and performance results. Regardless of the type of work, people should measure progress against their goals. There are two distinct reasons to establish formal methods for measuring team performance:

V W X Y Z

654

Encyclopedia • Manage risk and make key decisions at reviews and key project

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

milestones to ensure adherence to the plan or to explore indications of critical changes requiring adjustments to stay on plan. To avoid going off plan, make changes based on leading indicators to stay on plan. Changes should be made to fulfill the plan, rather than being made too quickly. The tools and tasks are what change, not the deliverables and requirements, which are immutable. • Pay for performance in accordance with specific requirements and

deliverables. A tool is as good as the information that is supplied to it. The old adage, “garbage in, garbage out,” could not be more applicable. Scorecards must track the right information to be useful to a data-driven leader. Taking the time to determine the critical risk accrual requirements defines the appropriate information to design into and track in a scorecard. The requirements may vary based on the tool, task, or gate deliverable. Hence, a system of scorecards works best to address: 1) the right tool, 2) applying the right tool to the right task at the right time, and 3) delivering the right summary data for risk management and decision-making. Requirements are the business questions asked before a phase of work is conducted. The goal is to design measurable work in light of the requirements before you start measuring. In general, there are two types of tracking tools: checklists, the simpler version, and the scorecard. More complex scorecards also can be referred to as dashboards, depicting that all the essential key indicators—the critical parameters—needed to “drive” a business are together in one spot. Checklists assess two states of task completion—done or not done. Checklists serve well as a reminder tool rather than as a predictive tool. Checklists prompt people to recall what needs to get done and track if it is done. The checklist catalogs the expected requirements, tasks, or deliverables. Checklists monitor if an item has been completed or not (binary yes or no response) to help avoid duplication of effort. Checklists suffer from a lack of discrimination. They fail to provide information about quality or quantity. Checklists lack predictive specifics about how well a tool was used or the percent completion of a task against its original target, and they fail to discriminate on the details of risk accrual. (See Also “Checklists—7QC Tool,” p. 204) Scorecards help to scrutinize the healthiness of the business in terms of its performance and risk accrual. The scorecard is a tool that drills down into the quality of processes or deliverables against respective targets and requirements. Scorecards can predict trends and alert business people to potential changes that require a response to maintain the planned course. A system of scorecards presents those vital few indicators needed to understand how well a business is performing against its

Score cards

655

goals; whereas a subset of scorecards probes deeper into an individual process or functional task/tool application area. The concept of scorecards applies to any discipline to monitor the healthiness of a task, project, process, or entire business. Use both a system of scorecards and a dashboard, as they complement one another but are unique. The dashboard contains the critical success metrics that require monitoring on a regular basis to maintain the healthiness of the ongoing operations. The dashboard takes a snapshot of the business appropriate leading and lagging indicators and organizes things typically in one place (a document and/or web site). The system of scorecards examines the work approach used by the process players, monitoring the interconnected tool-task-deliverables link to the requirements. The system of scorecards informs the review processes not only for ongoing operations (as with a dashboard), but also the phase-gate governance process of a project. Dashboards refer to the data critical to driving the business. The literature on dashboards is voluminous; however, little is written on the scorecards system. An organization should utilize both systems to aid in monitoring its business’ state of health. (See Also “A Process for Product Development,” p. 887 in Part III, for more detail on the governance and phase-gate review process.) A four-level, hierarchical flow starts with the criteria or requirements that dictate the deliverables and then steps down to the tasks needed to complete the right outputs at the right time. Then it flows down to the appropriate tools, techniques, and best practices that support the tasks. Figure S-32 illustrates these four criteria in sequence.

A B C D E F G H I J K L M N O P Q

Requirements

R

Deliverables

S

Tasks Tools, Methods and Best Practices

T U

Figure S-32: Hierarchical Flow of Work

This four-level flow is well-suited to measurement by using an integrated system of scorecards. A system of scorecards can be designed and linked so that each subordinate level adds summary content up to the next level. Use several types of scorecards to measure performance and managing risk along the work hierarchy: Tool Scorecards, Task Scorecards, and two types of Gate Review Scorecards.

Tool Scorecards The most basic level of scorecard is filled out by team members or process players who are actually using specific sets of tools, methods, and best

V W X Y Z

656

A B

Encyclopedia

practices to help complete their tasks. This is called a Tool Scorecard. They are easy to fill out and should not take more that 20 minutes or so to complete at the end of a tool application. Typically they are filled out in collaboration with the team leader for supervisory concurrence. At the conclusion of using a tool, the person or team members responsible for applying the tool to a task should account for the following measurable items: quality of tool usage, data integrity, and results versus requirements. Table S-1 provides a sample Tool Scorecard to track these items.

C Quality of Tool Use

D E F G H I J K L M N O

Data Integrity

Tool

Results versus Requirements

Average Score

Data Summary (incl. Type & Units)

Task Requirements

1. 2. 3. 4.

Table S-1: Sample Tool Scorecard

The first column of the Tool Scorecard simply records the name of the tool used. The Quality of Tool Usage column can be scored on a scale of 1 to 10 based on the following suggested criteria and levels, as shown in Table S-2. Adjust these rankings as appropriate for your applications.

P Q R S T

Rank

Right Tool

Fullness of Use

Measurement System Capability

10

X

High

High

9

X

Medium

High

High

Medium

8

X

U

7

X

Low

High

V

6

X

Medium

Medium

W

5

X

Low

Medium

X

4

X

High

Low

Y

3

X

Medium

Low

Z

2

X

Low

1

Wrong Tool

Table S-2: Suggested Quality of Tool Usage Ranking

The Integrity of the Data produced by the tool usage can be scored using the following suggestions shown in Table S-3.

Score cards

Rank

Right Type of Data

Proper Units

Measurement System Capability

% Data Gathered

10

Excellent

Direct

High

High%

9

Excellent

Direct

High

Medium%

8

Excellent

Direct

Medium

High%

A

7

Good

Close

High

High%

B

6

Good

Close

Medium

Medium%

C

5

Good

Close

Medium

Low%

4

Weak

Indirect

Medium

High%

D

3

Weak

Indirect

Low

Medium%

2

Weak

Indirect

Low

Low%

1

Wrong

Wrong

None

N/A

Table S-3: Suggested Integrity of Data Ranking

Adjust the nature of the scoring criteria as appropriate for your applications. The key is to make clear delineation between various levels of measurable fulfillment of the criteria. This scoring stratification transforms business requirements into actionable performance criteria, thereby clarifying expectations. The ability of the Tool Results to Fulfill the Task Requirements is scored with the help of the criteria found in Table S-4. Scale

657

Criteria

10

The task is complete in all required dimensions, a well-balanced set of tools have been fully applied to 100% completion.

9 to 8

The task is approximately 80-90% complete, some minor elements of the task are not fully done,a well-balanced set of tools have been used but some minor steps have been omitted.

7 to 4

The task is not complete somewhere across the range of 40-70%,

3 to 1

The task is approximately 80-90% complete, some minor elements of the task are not fully done,a well-balanced set of tools have been used but some minor steps have been omitted.

E F G H I J K L M N O P Q R S T U V W

Table S-4: Suggested Tool Results To Fulfill the Task Requirements Scale

X Y Z

658

A B C D E F G

Encyclopedia

The rating scale in Table S-4 accounts for how well the data fulfills the original requirements. It is acceptable to find that a full set of data cannot meet the requirement a task was designed to fulfill. If this is the case and the results are bad news, congratulations for communicating the truth. The intent is to avoid false positives and false negatives when making decisions about a project’s viability. This metric helps quantify the underdevelopment of data and facts that can lead to poor decisions.

Task Scorecards Task scorecards have the capability of discriminating performance relative to its requirements at both the aggregate level of tool completion and the summary level for each major task. A sample task scorecard is shown in Table S-5. Table S-5: Sample Task Scorecard

H I

Task

J

1.

K

2.

L

3.

M N O P Q R S T U V W X Y Z

Average Tool Score

% Task Fulfillment

Task Results versus Deliverable Requirements

Red

Yellow

Green

Deliverable Requirements

X X X

4.

The Average Tool score averages the tool scores together that align with each major task. A very insightful metric for each major task within a phase is the Percent Complete or Percent Task Fulfillment; wherein the real mechanics of cycle-time are governed. If a team is overloaded with projects or not given enough time to use tools, it is almost certain that they will not be able to fully complete their critical tasks. Under-completed tasks usually are a leading indicator that the schedule likely will slip. Especially if too few tools are being used, and the ones that are being used are not being fully applied. So there is a double effect: 1) poor tool use leading to incomplete data sets and 2) tasks that simply are unfinished. The Average Tool Score will also tend to be low. This means risky decisions are being made on the wrong basis. It is fine to face high risk situations in projects but not because the team is too busy to do things right. Task incompletion is a major contributor to why teams make mistakes and fail to grow on a sustainable basis. Table S-6 suggests a ranking scale from 1 to 10 to quantify the level of risk inherent in the percent of uncompleted tasks. The column comparing the Task Results versus Deliverable Requirement identifies how well the work satisfies project requirements. If 100% of the critical tasks is completed, and a balanced set of enabling tools are used to underwrite the integrity of the deliverables, the team is doing its

Score cards

best to control risk. Outstanding deliverables can be produced with full integrity and clarity, but simply fail to meet the requirements for the project and its business case. The team has great data saying that the goals cannot be met. This is a good reason to kill a project with confidence. Not many companies kill projects very often and even fewer do it with tooltask-deliverable confidence. Scale

Criteria

10

The task is complete in all required dimensions, a well-balanced set of tools have been fully applied to 100% completion.

9 to 8

The task is approximately 80-90% complete, some minor elements of the task are not fully done,a well-balanced set of tools have been used but some minor steps have been omitted.

7 to 4

The task is not complete somewhere across the range of 40-70%, moderate to major elements of the task are not done; tool selection and use has been moderate to minimal; selected tools are not being fully used, significant steps are being skipped.

3 to 1

The task is not complete across the range of 10-30%, a very few tools have been selected and used, their steps have been heavily truncated, and missing major steps altogether.

Table S-6: Suggested Score on Risk Level Inherent in Uncompleted Tasks

659

A B C D E F G H I J K L

Indicate fulfilled requirements with a positive green light to continue investing in the project. Depict unfulfilled requirements using a caution yellow or a negative red light that signals a re-directing of the project or an outright discontinuance. A color-coded scheme of classifying risk can be defined as follows:

M N O P Q

• GREEN—100% of the major deliverables are properly doc-

R

umented and satisfy the Gate requirements. A few minor deliverables may be lagging in performance but present no substantive risk to the success of the project on three accounts: time, cost, and quality.

T

• YELLOW—A very few major deliverables are incomplete

or falling short of fulfilling their requirements. A corrective action plan is documented, and there is a very high probability the problems can be overcome in a reasonable and predictable amount of time. • RED—One or more major deliverables is unfinished or fails

to meet requirements, and no corrective action plan exists to close this gap. The project is to be killed, redirected, or postponed until a corrective set of specific actions is defined and a predictable path to project timing is in hand.

S U V W X Y Z

660

Encyclopedia

Table S-7 suggests a set of ranking values to quantify the risk associated with varying levels of mismatch between what a major task delivers and the requirement. Scale

A

10

9 to 8

Results deliver a major portion of the data necessary to support the fulfillment of lack of fulfillment of the requirements.

7 to 4

Results deliver a moderate portion of the data necessary to support the fulfillment or lack of fulfillment of the requirements.

3 to 1

Results deliver a very limited amount of the data necessary to support the fulfillment or lack of fulfillment of the requirements

B C D E F G

Criteria Results deliver all data necessary to completely support the fulfillment or lack of fulfillment of the requirements.

H I

Table S-7: Suggested Score on Mismatch of Task to Requirements

J K L M N O P Q R S T U V W X Y Z

In summary, the Tool Scorecard documents the quality of tool use, integrity of the data, and the fulfillment of a task requirement. The next level of scoring risk is a Task Scorecard, where the project leaders quantify how well one or more enabling tools have contributed to completing a major task; what percentage of the task has been completed; and how well the deliverables from the task fulfills the gate requirements. Scorecards help to answer the questions, “How well did the team do in meeting the requirements?” and “Is the team done and ready to prepare for a Review, or not?” A positive affirmation to these questions indicates the following conditions: “Tools fulfilled the Task Requirements” and the “Tasks fulfilled the Gate Requirements.” We are now ready to look at the final summary scorecards that a Gate Keeping team will use to quantify risk accrual at the end of a phase of work.

Review Scorecards The project sponsor or process owner uses a set of Review Scorecards to quantify accrued risk and make decisions at each Review or major milestone. Figure S-33 illustrates the two kinds of reviews: 1) Functional-Level Reviews and 2) Executive Level Reviews. Functional Reviews are detailed and tactical in nature and prepare the team for an Executive Review. An Executive Review is strategic in nature and looks at macro-level risk. The risk management at the executive level delves into how a particular project contributes to the overall portfolio of commercialization projects or how it manages risk in the post-launch environment. Thus, functional reviewers worry about micro-details within their particular project or process, while the executive reviewers

Score cards

worry about accrued risk across all projects that represent the future growth potential for the business as a portfolio. The former looks at alignment of risk within the specific project’s tactics, while the latter looks at alignment of project risk across the business strategy. Functional Reviews can be done for technical teams and marketing teams as independent events. Executive Reviews are summary presentations that should integrate both technical and marketing perspectives, as well as any other applicable macro-gate deliverables.

661

A B C

Types of Reviews

D E

Functional • Micro • Tactical • Preparation for Executive Review

F

Executive • Micro • Strategic • Enterprise-wide Risk Management

G H I

Figure S-33: Types of Reviews

J

Table S-8 shows an example of a generic template for a Functional Gate Review.

K L M

Deliverables

1. 2. 3.

Grand Average Tool Score

Summary of Tasks Completion

Summary of Task Results versus Deliverable Requirements

Red

Yellow

Green

Accountable

Target Completion Date

Corrective Action & Risk Assessment

N O P

X

Q

X X

4.

Table S-8: Sample of a Functional Gate Review Scorecard

R S T U

The integrated system of Tool, Task, and Deliverable Scorecards provide a control plan for quantifying accrued risk in a traceable format that goes well beyond simple checklists. Control plans are a key element in Six Sigma. The Task Scorecard feeds summary data to this functional form of a Deliverable Scorecard. The next review format is used for Executive Reviews where deliverables are summarized for rapid consumption by the executive and sponsor team. Figure S-34 provides a sample of a scorecard used at the executive level.

V W X Y Z

662

Encyclopedia Project Financial Goals:

Project Cycle -time Status:

NPV:

This Phase

ECV:

A B C

Next Phase

IRR:

Project Resource Balance:

Op. Income:

Tech.: req’d Mktg.: req’d

ROI:

Integrated Program Plan:

actual actual

Gate Deliverables & Risk Score & Color Ratings: Gate Deliverable:

D

Confidence in Data Risk:

Results vs. Reqt Risk:

Score

Score

Color

Gate Review Risk:

Color

Risks/Issues/Decisions/ Corrective Actions:

Color

E F G H I J

Contribution to Portfolio Status:

K

RWW Value:

New Technology & Design Capability Growth Status:

Tech. Probability of Success:

Technology CFR CGI:

Mktg. Probability of Success:

Design CFR CGI:

L M N O P Q R S T U V W X Y Z

Overall Gate Status:

Figure S-34: Sample Executive Review Scorecard

The template in Figure S-34 is a common element employed by numerous companies that have a commitment to strategic portfolio management. The executive team looks at a number of these scorecards to balance risk across its portfolio of projects while driving growth to the top-line in accordance with the business strategy.

Integrated System of Scorecards Modify the proposed integrated system of scorecards to suit your organization’s process. Figure S-35 illustrates this interlocking scheme. Each hierarchical level should feature a scorecard designed to mirror the respective business model from one of the following four areas: 1) your tool-task groups (clusters), 2) your deliverable-requirement groups as they are aligned with your phases of your portfolio renewal process, 3) product commercialization process, and 4) your post-launch line management process. (See Also “Six Sigma for Marketing (SSFM),” in Part I, p. 67) A system of scorecards should be used to manage the portfolio renewal process, with the first three being sufficient to manage risk within a project. As each product commercialization project is activated from the portfolio renewal team, an Executive Summary Scorecard is initiated and updated during commercialization.

SIPOC (Supplier-Input-Proce ss-Output-Customer)

Quality of Tool Use

Data Integrity

Tool

Tool Use

Results versus Requirements

Average Score

Data Summary

663

Task Requirements

(incl. Type & Units)

1. 2. 3. 4.

Task Completion

Task

Average Tool Score

% Task Fulfillment

Task Results versus Deliverable Requirements

Red

1.

Yellow

Green

Deliverable Requirements

A

X

2.

X

3.

X

B

4.

Deliverables

Deliverable Fulfillment for Funtional Reviews

Grand Average Tool Score

1. 2.

Summary of Tasks Completion

Summary of Task Results versus Deliverable Requirements

Red

Yellow

Green

Accountable

Target Completion Date

C

Corrective Action & Risk Assessment

D

X X

3.

E

X

4.

Project Financial Goals:

Project Cycle-time Status:

Integrated Program Plan:

This Phase

ECV: IRR: Op. Income: ROI:

Next Phase

Project Resource Balance: Tech.:req’d_____ actual______ Mktg.:req’d_____ actual______

Gate Deliverables & Risk Score & Color Ratings: Gate Deliverable:

Executive Summary of Risk

Confidence in Data Risk:

Results vs. Reqt Risk:

Score

Score

Color

Color

F G

NPV:

Gate Review Risks/Issues/Decisions/ Risk: Corrective Actions: Color

H I

Contribution to Portfolio Status: RWW Value:

New Technology & Design Capability Growth Status:

Tech. Probability of Success:

Technology CFR CGI:______

Mktg. Probability of Success:

Design CFR CGI:______

Overall Gate Status:

Figure S-35: Summary of Scorecards

As the old saying goes, “That which gets measured gets done.” Monitor performance with scorecards. Use checklists as reminders of what to use or complete, not how well that expectation is being met. A custom designed scorecard set spanning the four-level hierarchy of tool-taskdeliverables-requirements serves as both a quantitative and a qualitative barometer. Scorecards identify not only what to measure (the critical performance elements), and how well these critical parameters satisfy a set of criteria (or requirements), but also when to measure so that you have a much higher probability of preventing down-stream problems as you seek to sustain growth across your enterprise.

J K L M N O P Q R S T U

SIPOC (Supplier-Input-Process-Output-Customer)

V W

What Question(s) Does the Tool or Technique Answer? What is the scope of the project or process? The SIPOC technique helps you to • Communicate a high-level view of the project scope or process • Prevent scope creep in a project

X Y Z

664

Encyclopedia

When Best to Use the Tool or Technique At the beginning of a project or to provide a big picture perspective of a process.

A B C D E F G H I J K L M N

Brief Description The acronym SIPOC [pronounced “sie-POCK”] stands for Supplier-InputsProcess-Outputs-Customer. The tool provides a high-level summary of the project or process. It is a good communication tool that sums up the focus of a project—the process of interest and its related elements. Based on the input-process-output model (IPO), it recaps the process-related detail from the project charter and defines the project boundaries to help prevent scope creep. It diagrams these five key elements in a column-structured chart to provide information as follows: • Suppliers—The key functions (roles or people) that produce the

inputs • Inputs—The key process information, parts, components, decisions

contributions required prior to beginning or completing an activity or task • Process—The high-level process activities (typically three to eight

steps) that transform the inputs to produce the outputs

O

• Outputs—The key process deliverables or tangible outcomes

P

• Customers—The key customers (both external and internal) request-

Q R S T U V W X Y Z

ing the outputs (or deliverables) Often this tool is built backward, starting with the customer (the “C” of SIPOC, as if it were named COPIS) and working back through to the suppliers, given that it is the customers that determine or define the deliverables of interest. Customers provide the Voice of the Customer (VOC). The process outputs represent the key deliverables (both final and inprocess) of interest that may require improvement. Hence, the outputs are measurable (either variable or attribute data). The high-level process elements bound the scope of the project or process and contain elements considered as potential root cause sources. The process inputs also represent potential sources of root causes. The key suppliers should be engaged in the improvement initiative. Figure S-36 illustrates a SIPOC example for dining at a restaurant. Notice that the process activities are high-level and are sometimes referred to as a high-level Process map. The suppliers, inputs, and outputs are directly related to the process and the customer(s). (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

SMART Problem and Goal Statements for a Proje ct Charter Supplier • Newspaper

Input • Critiques and Reviews

Process

Output

Going to a Restaurant to Dine

• Hunger Satisfied • Liked Meal • Ambience enjoyed

• Phone Company

• Phone Service • Yellow Pages • Discount coupons

• Select Restaurant • Order Drinks and

• Friends

• Word of Mouth

• Eat and Drink • Pay Bill

Recommendation

• Taxi Service

665

Customer • Restaurant Patrons

Meal

• Ride to restaurant

Figure S-36: Dining at a Restaurant SIPOC

A B C D E

A SIPOC serves as a high-level snapshot and serves as an effective communication tool to explain what could be a complex process in simple terms.

SMART Problem and Goal Statements for a Project Charter

F G H I J K

What Question(s) Does the Tool or Technique Answer? What is the most succinct description of the project’s goal and problem statements? SMART helps you to • Communicate a clear, concise, complete description of a project’s goal • Identify the problem it addresses

L M N O P Q R S

When Best to Use the Tool or Technique During the planning stage that establishes and defines a project, use this technique to better communicate its main points and goals.

T U V W

Brief Description The SMART acronym stands for Specific-Measurable-Achievable (but Aggressive)-Realistic-Time-bounded. Some literature defines the SMART acronym with the “A” representing Agreed to or Achievable, but Aggressive or Attainable and the R as Relevant to the business, therefore it is agreed to by management/sponsor. In summary, the acronym translation is fluid, but the key elements remain constant—specific, measurable, time-related and some combination of aggressive, but realistic. The following options include bolded terms to indicate the more common translation:

X Y Z

666

Encyclopedia • S—specific, significant, stretching • M—measurable, meaningful, motivational • A—achievable, but aggressive, attainable, agreed to, acceptable,

action-oriented A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

• R—realistic, relevant, reasonable, results-oriented • T—time-bound, timely, time-based

This technique is applied to both a project’s goal statement and its problem statement to ensure that both contain the essential elements to succinctly communicate their vital elements. Specific describes precise, concise language. Measurable requires a metric, target, or unit of measure. Achievable but Aggressive describes the objectives, which are Realistic. The last element, Time-bounded or Timely takes on different dimension for the goal and problem. With respect to the goal statement, time-bounded seeks a specific date (day, month, year) that the project targets for its completion. Vague language providing the month, quarter, or year that the project intends to achieve its objective leaves room for ambiguity and misunderstanding between the project team and its sponsor. Ambiguity leads to improper expectation setting, wherein the project team could intend the end of a month, or worse yet a quarter, while the sponsor anticipates completion at the beginning of the month, or quarter. Misalignment leads to frustration, disappointment, and at times more serious consequences. If the project completion date slips, the team is better off communicating that to the sponsor and why, rather than hoping for a miracle. The time element for the problem statement describes the duration of the problem. It defines how long this pain has persisted—is it a onetime event or a systemic problem. The time dimension bolsters the team’s case for change when trying to secure project funding and resources and implementing and sustaining the improvement. It adds a sense of magnitude and is best accompanied with a control chart or run chart illustrating the key metric plotted over time. (See Also “Control Charts—7QC Tool,” p. 217 and “Run Chart—7QC Tool,” p. 611) The problem statement describes what is wrong over time. Alternatively, at times the problem statement is replaced by an opportunity statement that describes the possibilities. An example of a problem statement for a restaurant might be The number of restaurant patrons have declined over the last six months, the number of restaurants in the area have increased, and food costs have increased 7% over last year.

SMART Problem and Goal Statements for a Proje ct Charter

The goal statement is part of a larger purpose, as a key element of a project charter. It articulates the team’s improvement objective—project success criteria and duration (how long it should take to achieve). It determines when the project team knows it has achieved its mission and gauges what “good” looks like. However, it is not the solution or the answer on how to achieve the desired state. An example of a goal statement for a restaurant might include Within the next two months, increase the number of services offered that can yield last year’s 15% target profit margin. The example is specific (= increased offerings), measurable and achievable (= prior year’s 15% profit margin), time-bounded (= within in next two months), and relevant to the business (= restaurant services). However, the time element ideally should be more specific and give a specific target date the new services would either be identified or implemented. The same example rewritten more tightly is:

667

A B C D E F G H I

By May 20, 2008, recommend new services offerings (with a business plan) showing how they can help us achieve last year’s 15% target profit margin within two months of launching them. Upon approval, new services to be operational by July 1, 2008.

Project Charter A project charter is a sponsor document. It essentially is a contract for work between the project sponsor and the project team. This document authorizes the project, names its key stakeholders, and defines its scope to address a business need on a temporary basis. (A project has a finite duration with a definite beginning and end.) It drafts the people to work on the project and defines the project budget and timeframe. A project charter is incomplete until the project sponsor approves it and communicates it to the appropriate parties. Developing and publishing a project charter is the first step in commissioning a project—before any work begins. Often the charter or a subsequent memo issued to the organization formally recognizes the initiation of a project. The project charter describes the high-level requirements for project success and provides the project manager the authority to apply organizational resources to project activities. A charter, its layout, and format are unique to an organization, but ideally are standardized throughout the organization. In general, charters contain any combination or all of the following components (the more commonly included categories are bolded): • Header Information—Project Title, sometimes a tracking number,

issuance date, and Project Sponsor establishing the project. • Overview Statement—Aligning the project with overall company strategy/vision.

J K L M N O P Q R S T U V W X Y Z

668

Encyclopedia • Problem Statement (or Opportunity for Improvement). • Goal Statement [with the critical parameters, improvement target,

A B C D E F G H I J K L M N O P Q R S T U V W

timeframe (target completion date), and sometimes the critical deliverables]—It may describe the approach to achieve the opportunity statement or solve the problem statement defined in the project scope, but not provide the answer or the solution. • Project Scope—Describes the boundary conditions; identifies key

parameters covered or not covered by the project—what’s inbounds/out-of-bounds. • Target Audience—Defines who the project scope addresses (that is,

target populations, functions, disciplines, departments) or who the customer is. • Project Deliverables—Defining the key project outputs, sometimes

listed separately from the goal statement. • Project schedule—The proposed high-level timeline with project

milestones (including Milestone Review Meetings or phase-gates) and target completion date for each (key milestones). (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127, and “PERT (Program Evaluation and Review Technique) Chart,” p. 453) • Business Case—Definition of the quantifiable benefit of the project

and how it aligns with business strategy or goals. • Project Budget—The proposed project funding required to complete

the deliverables. • Project Resources: • Project Manager • Project Team Members (usually only the core team members,

even if sub-teams also are involved) (See Also “RACI (Responsible, Accountable, Consulted, Informed) Matrix,” p. 554) • Approval Signatures—Project Sponsor and any critical Key Stake-

holders, if applicable

X Y Z

Who are Stakeholders? The stakeholder is defined as anyone (or any function) impacted by the project, its outcome, or the process. Examples include management; upstream and downstream process players (including suppliers and third party partners); regulatory; process players not participating in the project, but who have representatives on the project team.

SMART Problem and Goal Statements for a Proje ct Charter Project Charter

Approval Date:

669

Target Completion Date:

Project Title: Team Members and Roles: Project Manager: Sponsor: Key stockholders:

A

Problem Statement:

B C

Goal Statement:

D E

Business Case: Project Scope:

In Scope

F

Exclusions

G

Financial Benefits: (in 000th) Financial Assumptions: Unquantifiable Benefits:

H

Project Budget

Risks/Constraints Assumptions:

Key Deliverables/Milestones:

Target Completion Date:

Comments:

I J K

Approval Signature Attached Item Checklist:

L M N

Figure S-37: Sample Project Charter Template

O P

There are five basic requirements for conducting a successful project. In order of importance, they are as follows: choosing the right people; choosing the right people; choosing the right people; setting up the right organization; and using the appropriate methodology, tools, and systems.

Q

Additional information that is beneficial to add to the project charter or provide as addendums include

T

• SIPOC—Serves as a summary document that captures a higher level

snapshot in presentation format. (See Also “SIPOC (Supplier-InputProcess-Output-Customer),” p. 663) • High-level Process map. (See Also “Process Map (or Flowchart)—

7QC Tool,” p. 522) • Project RACI—Identifies who is accountable for the key deliver-

ables. (See Also “RACI (Responsible, Accountable, Consulted, Informed) Matrix,” p. 554) • Project Communication Plan—(See Also “Matrix Diagrams—7M

Tool,” p. 399 for a brief discussion on a communication plan.)

R S U V W X Y Z

670

Encyclopedia • Project Team—People assigned to project, key role, project structure

(particularly if sub-teams exist), and time allocation. •

Project Team member list to include: sponsor(s), key stakeholders, project leader (or manager), core project team members, any support members or subject matter experts, and any sub-team members.

•

Sometimes the project sponsor designates a representative to oversee the project direction; this delegate may be referred to as a project “champion” working on behalf of the sponsor.

A B C D E F G H I J K L M N O P Q R S T U V

• Supporting summary data—Highlights the problem (particularly

over time). (See Also “Run Chart—7QC Tool,” p. 611) • Key Project Success Metrics—What the sponsor agrees are the criteria

(metrics) for success over and above the goal statement, project budget, and timeframe (for example, quality, process adherence, and so on). • Project Approach—Method the project will follow (that is, DMAIC,

UAPL, DMADV, and so on). • Business Case—Describes why the particular project has been cho-

sen. (See Also “Cost/Benefit Analysis,” p. 238) • Key Assumptions and Constraints—Defines any project limitations

or conditions.

Summary A project charter and its problem and goal statements are inseparable. Together they summarize and document the fundamental business thinking about the problem and the overall project roadmap to address the problem. The charter defines project boundaries—in scope, out-of-scope, constraints (that is, budget). It identifies who is accountable to resolve the problem and what roles (or departments) are affected. The charter defines when the project will be completed; based on what criteria, and in what timeframe.

W X Y Z

How to Use the Tool or Technique Develop a project charter and its problem and goal statements using the following procedure. A common project charter template should be deployed consistently across an enterprise or at minimum, a division. A common approach to completing a project charter includes Step 1.

Draft charter content based on current understanding of project scope and on lessons learned from prior projects of similar scope, scale, and complexity.

SMART Problem and Goal Statements for a Proje ct Charter

671

a. Start with the problem statement, drafting a concise statement that articulates the pain that the issues have been causing the business, the magnitude (or effect), and for how long, using the SMART technique. b. Define the goal statement, applying the SMART technique.

A

Review charter with sponsor, make modifications, and gain approval.

B

Sponsor should publish and distribute the project charter, naming the project lead and key project members, to the organization(s) affected (particularly the stakeholders).

D

Step 4.

Project charter is reviewed and revised on an as-needed basis.

G

Step 5.

Project charter is revisited in Stage Gate Reviews or distributed as pre-read for a Stage Gate Review.

H

At close of project, the project charter gets its final review to ensure it reflects the actual project events and includes any lessons learned. The final document is filed with the other administrative documents to close the project.

J

Step 2. Step 3.

Step 6.

Hints and Tips

C E F

I K L M N O

• Examine the problem statement to ensure that the SMART

P

technique has been applied. Double-check that the described

Q

pain and its magnitude create a rallying point for the project

R

team to focus on to resolve. The magnitude, in part, should

S

include a time component describing how long this problem

T

has persisted.

U

• Ensure that the goal statement fits the SMART by specifying

V

a specific day of the month and year for the project comple-

W

tion date and that it describes what “good” looks like or how

X

to approach solving the problem—but avoids providing a

Y

solution.

Z

• Verify that the sponsor approved the charter.

672

Encyclopedia

• Verify that the project charter has been published and distributed to the key stakeholders (including project team members and respective managers). A B C D E F G H I J K L M N O P Q R S T

• Use in Gate Reviews to ensure project remains within scoped boundaries,

within

budget,

and

producing

on-time

deliverables. • Modify project charter if “new” circumstances change or better define the original conditions. • At project close, ensure that an updated, accurate project charter is filed with the project papers. • Reference charters from previous projects to determine if they had similar problems, scope, scale, budget, and so on and whether the historical knowledge in their project files suggest applicable and invaluable lessons learned. • A project charter template should be approved by the head of the organization and deployed as a standard throughout the enterprise. • The project charter template may be developed as part of an intellectual capital management system (for example, a project database) or by using application software such as Microsoft Word or Excel.

Solution Selection Matrix

U V W

What Question(s) Does the Tool or Technique Answer? Which solution option best meets requirements?

X

A Solution Selection Matrix helps you to

Y Z

• Prioritize the solution options and select the most optimal one • Communicate the decision made in the selection process and why

Solution Sele ction Matrix

673

Alternative Names and Variations Variations on the tool include • Prioritization Matrix (Also See “Prioritization Matrices—7M Tool,”

p. 470) • Pugh Concept Evaluation (See Also “Pugh Concept Evaluation,”

p. 534)

A B C

When Best to Use the Tool or Technique Selecting from a set of solution options, apply this tool to rank and select the best option, after requirements are thoroughly understood, the problem, current state and root causes are analyzed, and potential solution concepts have been generated.

D E F G H I

Brief Description The Solution Selection Matrix structure and purpose is very similar to a Prioritization Matrix; however, its purpose is specifically to rank solution candidates and select the most appropriate. The criteria used to evaluate the candidates should tie to the critical-to-quality (CTQ) characteristics of the customer and business. Selection Criteria The selection criteria used to evaluate the candidates could be any appropriate combination of customer and business requirements. Typically the voice of the customer (VOC) prioritizes or serves as the first filter in the selection process. The voice of the business (VOB) typically dominates the final selection to ensure that the end result aligns with the business goals. By convention VOB criteria include three main considerations: • Feasibility, which encompasses

• Designing, developing, implementing, and maintaining the solution • Alignment with the organization’s mission, goals capabilities and capacity [of funds, resources, technology, and knowledge (know how)]. • Effectiveness in addressing either a key area of opportunity

(potential benefits), root cause of a problem, or a specific set of requirements • Cost to design, develop, implement, and maintain the solution

J K L M N O P Q R S T U V W X Y Z

674

Encyclopedia

The criteria also may include the ease of implementation and potential risk (pitfalls or adverse side effects).

A B C D E F G H I J K

Defining the criteria is a critical step in using this tool. Developing a common understanding among those who will evaluate the candidates eliminates any inconsistency in the scoring. Moreover, document this criteria definition to communicate to the stakeholders why a certain candidate was selected over the others.

Evaluation and Selection The simple matrix approach helps to organize and focus the process of evaluating each option against a common set of criteria. This is best done with a cross-functional, multi-disciplined team to provide different perspectives. Either the consensus or equal voting approach can produce a meaningful outcome. Use the consensus approach when the team make-up is comprised of similar talkativeness, experience, and reporting relationship. Each team member discusses and recommends a proposed criteria weight and relationship score, and then the team collectively arrives at a single number per appropriate cell, as illustrated in Figure S-38.

L

Criteria and Weights

M N O P Q R S T U V W X Y Z

Easy

Quick

Tech

Hi Impact

Customers

Solution

8

7

5

9

5

Total

A

1

9

3

9

3

182

B

0

9

3

9

3

174

C

1

9

1

3

3

118

D

0

3

1

9

3

122

Figure S-38: Sample Solution Selection Matrix Using the Consensus Approach

Use the equal voting technique when the team includes members with dissimilar experience, mixed reporting levels, and when some members tend to dominate the conversation more than others. Each member votes for a criteria weight and relationship score, either silently or aloud. The respective votes for each type are tallied (added) and divided by the number of voters to calculate the average score. That single average number is entered per appropriate cell. (An alternative to the average score would be to use the total score; however, sometimes the large numbers it can create seem unwieldy.) Figure S-39 shows a Solution Selection Matrix created with the total scoring approach to the equal voting technique.

Solution Sele ction Matrix

Criteria

Bill

Amir

Julie

Total

1. Ease of Implementation

4

1

2

7

2. Time to Implement

1

5

2

8

3. Customer Notices Change

4

3

5

12

4. Defect Reduction

2

1

2

5

A

Criteria and Weights

B

Easy

Time

Customer

Defect

Solution

7

8

12

5

A

1,3,3 49

9,3,3 120

1,0,3 48

3,1,3 35

252

B

3,9,9 147

3,3,3 72

1,1,3 60

9,3,9 105

384

C

1,0,1 14

3,1,3 56

9,9,3 252

3,1,1 25

14 = 7 x (1 + 0 + 1)

675

C

Total

D Individual Scores Bill, Amir, Julie

E F G H

347

347 = 14 + 56 + 252 + 25

Figure S-39: Sample Solution Selection Matrix Using Equal Voting Technique with Total Scores

A third type of evaluation technique matches a solution against its original root cause to ensure the solution candidate’s effectiveness. This root cause technique focuses the team better than the other two approaches. It encourages brainstorming during the idea generation step, but the ideas stay tied to the root cause, rather than symptoms. Freewheeling, far-fetched ideas appear less frequently. After the initial creation of ideas, this approach encourages a synthesis step to hone in on more practical candidates. Moreover, this technique uses a unique threeway criteria scoring that multiplies feasibility, effectiveness, and cost to calculate an overall score. The team still needs to define and rank the specific criteria within the three categories. Figure S-40 shows a Solution Selection template for the root cause technique.

Selection Consider the highest scoring solutions first and then weigh other intangible factors that may lead to a secondary solution candidate being considered and perhaps further tested. As with all soft tools, such as matrices, avoid blindly selecting the top score getter as the final answer without further scrutiny. Sometimes the final scores are close and require evaluation and refining of the concepts.

I J K L M N O P Q R S T U V W X Y Z

676

Encyclopedia

Total FEC Score

Action (Y/N)

Cost

Practical Solution

Effectiveness

Solution Candidates Original Brainstorm List

Feasibility

Root Cause

Evaluation

Solution to Address: _____________________________

A B C D E F G H

Definitions of Evaluation Feasibility Criteria Scale: (high #= feasible) Effectiveness Criteria Scale: (high #= feasible) Cost Criteria Scale: (high #= inexpensive) Criteria

I J K L M N O P Q R S T

Figure S-40: Solution Selection Matrix template using the Root Cause Technique

Other Solution Selection Methods If the Solution Selection Matrix fails to provide a clear choice (or perhaps even if it does), consider getting more information about the final selected solution or set of valid solution candidates. For example, model or simulate the top or some of the top solution candidates to identify additional features or dimensions to modify and help confirm the final selection. Trials (small scale tests) also provide additional insight about the proposed option(s). If an organization currently utilizes a similar solution, observe their situation and gather any best practices and lessons learned.

U V W X

How to Use the Tool or Technique Develop a Solution Selection Matrix using the following procedure: Preparation:

Y Z

Step 1.

Review what you know about the process and the confirmed root causes.

Idea Creation. Create potential solutions, if not available already. a. List ideas sparked by the discoveries made while analyzing the process.

Solution Sele ction Matrix

677

b. Generate other possible solutions that address the root causes via i. Various brainstorming and creativity techniques,

such as challenge the rules, get rid of excuses, challenge paradigms. (See Also “Brainstorming Technique,” p. 168)

A

ii. Utilizing Lean concepts. (See Also “Lean and Lean

B

Six Sigma,” in Part I, p. 29) iii. Other potential sources of ideas including bench-

marking, best practices, literature search, trade journals, prior experiences of people, industry experts, databases, and surveys. iv. Affinity diagrams are useful for grouping possible solutions into categories. Step 2.

Create a matrix using an L-shaped structure and list the ideas down the first column, to serve as row headings. Construct the matrix template in Excel, Word, or on a flip chart, or any other appropriate media as long as the team members can collectively and simultaneously view the document.

Step 3.

Define the selection criteria and criteria weight and document the output. a. Agree on the selection criteria and their respective definitions. b. Determine the weight for each criteria, using a numeric scale where the highest number is the most preferable or most important, using either the consensus or equal voting technique. Common criteria weight scales are 1 to 3, 1 to 5, or 1 to 10. However, we recommend stack ranking the criteria, with the largest number representing the most important to avoid tie weightings to ensure better discrimination among the various solutions. c. Document the criteria and respective weights as a reference key for the matrix. d. List the criteria and their respective weights across the very top row of the matrix to serve as the column headings.

C D E F G H I J K L M N O P Q R S T U V W X Y Z

678

Encyclopedia

Step 4.

Prioritize the solution candidates. a. Use the scale of 9-3-1 to define the relationship score for each solution candidate and the criteria and record the rating in each appropriate cell, using either the consensus or equal voting technique.

A

The scale translates such that nine equals strong relationship, three represents moderate, one connotes weak, and zero or a blank cell indicates no relationship. Another way to define this scale might include cost savings such as 1 for < $100K, 3 for $100-500K, 9 for > $500K.

B C D E

The 9-3-1 scale helps to better discriminate among the options versus a 3-2-1 scale.

F G

b. Multiply the cell relationship score times the corresponding criteria weight to calculate the weighted criteria relationship.

H I J

c. Sum the weighted criteria relationship scores for a given solution candidate to calculate the total candidate score.

K L M N O P Q R S T

Step 5.

Select the optimal solution(s). a. Identify the highest total candidate score, which will probably be the preferred solution. b. Avoid turning your brain off and search the matrix for the next closest candidates to see if any of them are possible candidates as well. Consider combining the concepts to create a stronger proposal. c. Collect more information about the solution concept before doing a full-fledged implementation.

U V W

Hints and Tips The Solution Selection Matrix contains three distinct areas: idea cre-

X

ation, selection criteria, and the actual prioritization and selection of

Y

a final solution or set of viable solutions to be further tested. Find the

Z

Hints and Tips for the latter two topics embedded within the “Brief Description” section of this entry.

Solution Sele ction Matrix

Idea Creation: • Involve the right people in the idea generation activity; be inclusive. • Although this is a team exercise, individuals generate ideas, not teams. Ensure each team member contributes. • Focus on one root cause at a time. • Prioritize the root causes and start with the one determined to have the biggest impact on the desired response (big Y). • Generate as many potential solutions as possible for each root cause because the more potential solutions available to consider, the more opportunities for uncovering ideas for improvement. Exhaust all the ideas that the team can produce. • Consider combining ideas or the best characteristics from multiple solution ideas to generate a stronger solution. • Work with far-fetched ideas by modifying and alternating perspectives to uncover a potentially hidden, more practical solution candidate.

679

A B C D E F G H I J K L M N O

Supporting or Linked Tools Supporting tools that might provide input when developing a Solution Selection Matrix include • Benchmarking (See Also “Benchmarking,” p. 160) • Brainstorming (See Also “Brainstorming Technique,” p. 168) • Fishbone diagram (See Also “Cause-and-Effect Diagram—7QC

Tool,” p. 173)

P Q R S T U V W

• RWW (See Also “Real-Win-Worth (RWW) Analysis,” p. 560)

X

• SWOT (See Also “SWOT (Strengths-Weaknesses-Opportunities-

Y

Threats),” p. 699) • VOC/VOB (See Also “Voice of Customer Gathering Techniques,”

p. 737)

Z

680

Encyclopedia

A completed Solution Selection Matrix provides input to tools such as • DOE (See Also “Design of Experiment (DOE),” p. 250) • Pilots A

• Simulations (See Also “Monte Carlo Simulations,” p. 431)

B

• Transition and control plans (See Also “Matrix Diagrams—7M

C D E

Tool,” p. 399) Figure S-41 illustrates the link between a Solution Selection Matrix and its related tools and techniques.

F G

Bencharking

H I

DOE

Brainstorming

J K L

Fishbone

M N

Pilots Solution Selection Matrix

RWW

Simulations

O P

SWOT

Q

Transition and Control Plans

R S T U

VOC / VOB

Figure S-41: Solution Selection Matrix Tool Linkage

V W X Y Z

Variations Prioritization matrix (See Also “Prioritization Matrices—7M Tool,” p. 470) Pugh Concept Evaluation (See Also “Pugh Concept Evaluation,” p. 534)

Stakeholder Analysis What Question(s) Does the Tool or Technique Answer? Who supports the project, who doesn’t, and why?

Stakeholder Analysis

681

Stakeholder analysis helps you to • Assess the root cause behind why someone opposes or is indifferent

to the project, its mission, its recommendation, and/or its ongoing implementation, utilization, management, and control—plus develop a specific action plan as a countermeasure • Enhance the ownership of the project’s success (including sustaining

the improvement) among the stakeholders

A B C

• Improve communications

D

• Avoid political landmines

E F

When Best to Use the Tool or Technique Develop a stakeholder analysis at the beginning of a project and implement, revise, and monitor it throughout the life of the project.

G H I J

Brief Description The stakeholder analysis tool examines details around the Voice of the Business (VOB), to evaluate the key stakeholders’ acceptance of the project scope, boundaries, deliverables, and timing. The technique is meant for the project sponsor and project manager to assist in strategizing the optimal change management and communication approaches. (This tool assumes the project sponsor holds a senior rank relative to the project manager/Six Sigma black or green belt.) The technique helps to focus attention on an action plan grounded in business issues, rather than nonactionable discussions about individual personalities. A stakeholder is anyone impacted by the project; however, the project sponsor and project manager need to identify the key stakeholders needed to support, promote and sustain the project and its improvement. The executive sponsors often fund the project but are not the only stakeholders. Stakeholders also include the process owner, who may be a different person or role, and the suppliers to the process and customers of its outputs. Select only the vital few stakeholders who have key implications on (or influence on) the project. Because most key stakeholders’ organizational reporting relationship will be on par with or higher than the sponsor and the project manager/Six Sigma black or green belt, it is imperative to involve the sponsor in this activity. Personalize this process by identifying actual people, not roles within an organization. Conduct the analysis by person so that the diagnosis and countermeasure is tailored for that one individual. For this reason, keep the stakeholder analysis tool confidential. Given the importance of personalization, try to understand the given stakeholder’s position by direct

K L M N O P Q R S T U V W X Y Z

682

Encyclopedia

input, if possible. Determine current stakeholder positions by actually collecting data; talk to them at a conceptual level about the project and how it could impact them. If a direct, personal conversation is not possible, ask a close colleague to have the discussion. Be careful not to assume—do not guess. A B C D E F G H I J

The analysis technique has two parts—the acceptance chart and influence planning chart: Acceptance chart—Identifies current level of stakeholder support and defines the desired level of stakeholder support for the project to be successful. The chart is structured as an L-shaped matrix, with the key stakeholders as row headings and the different support positions as column headings. Typically, the chart graphically depicts the current level of stakeholder support by an icon and indicates the desired support level by a second icon. The convention for the support positions comprise the following four categories: strongly against, lets it happen (or neutral), helps it happen, and makes it happen (strongly supports). Figure S-42 shows a sample stakeholder acceptance chart. (See Also “Matrix Diagrams—7M Tool,” p. 399, for a discussion on different shaped matrices.)

K L M

Key Stakeholder

N

Helps It Happen

Makes It Happen (Strongly Supports)

Eileen

O

George

P

Neeta

Q

Sam

R

Current

S T

Strongly Against

Lets It Happen (Neutral)

Needed

Figure S-42: Sample Stakeholder Acceptance Chart

U V W X Y Z

Key Stakeholders Recognize that all stakeholders do not have to be in the Makes It Happen category. Some stakeholders may be just as effective in the Let it Happen category, based often on the fact that their functional area is not directly related to the area. For example, if a product development project were about to launch, perhaps the functional areas of finance or human resources might fall into the Let it Happen category.

Stakeholder Analysis

Influence planning chart—Tracks the current and desired positions. This chart indicates the source of resistance and strategy to move the stakeholder to desired level of support. Take the time to properly diagnose the type of resistance so as to employ the appropriate response strategy. The resulting action plan will either change the resistor’s way of thinking or modify the project or solution. Some stakeholders may exhibit more than one type of resistance; if so, identify the most dominant type and address that one first. Figure S-43 illustrates a sample influence planning chart.

683

A B C D

Key Stakeholder

E

Type of Resistance

Underlying Issue

Strategy

Eileen

Technical

Lack of Knowledge

Educate; Involve in meetings.

George

Political

Loss of Control

Emphasize project benefits.

Organizational

Control

Stakeholders get credit for solution.

K

Individual

Overworked

Educate over time via colleagues.

M

Neeta

Sam

Figure S-43: Sample Stakeholder Influence Planning Chart

Notice some of the types of resistance identified in Figure S-43 and the corresponding root cause (or underlying issue) and its countermeasure. For example, some strategies may include

F G H I J L N O P Q R S

• Lack of knowledge—Educate on project and personally involve in

review meetings. • Threat of loss of control—Focus on the project benefits to the

organization. • Desire to control—Credit the stakeholders for designing and imple-

menting the solution. • Overworked, seen as additional work—Educate over time and examine

possible process redundancies or gaps that may cause workload issues. Customize the strategy and action plan for the specific individual and corresponding reason for resistance. As a follow-on, a more general communication plan may result from a stakeholder analysis. This more general outcome would complement the individual strategies. (See Also “Matrix Diagrams—7M Tool,” p. 399, for a brief discussion on a communication plan matrix.)

T U V W X Y Z

684

Encyclopedia

Statistical Tools What Question(s) Does this Tool or Technique Answer? What statistical tools are available to use, and when should you use them? A B C D E F G H I J K L M N O P Q R S T U V W X Y

Brief Description The various different Six Sigma methodologies view work through a consistent set of lenses that examine three core elements: process, variation, and data to help make business decisions. Statistical thinking views all activities as part of process and that all processes have inherent variability. It uses data to understand variation and to drive decisions to improve these processes. (See Also Part I for a discussion on the four main categories of Six Sigma methodologies.) Statistics help examine these three core elements of work with less economic burden (time, cost and resources). Statistics enable quicker and cheaper analysis of a system. There are several statistical concepts that involve a combination of numeric and graphical techniques. A general overview of the most common ones used by Six Sigma include the following.

Descriptive Statistics Descriptive statistics are used to describe a population, process, and sample and their characteristics, such as location (central tendency), spread (variation), and shape (skewedness and peaks). Common terms include the following: Central Tendency This describes the location of data. It is measured three different ways to describe the values around which the data tends to congregate. • Mean—The sum (Σ) of the measurements (X) divided by the num-

ber of measurements (N _or n), sometimes referred to as the arithmetic average, denoted by µ or X. The median is affected by extreme values, such that one extreme value pulls the mean toward it. The formula for a population and sample mean where x is the number of observations (data), N is the population size, and n is the sample size):

Z n

µ

i

n 1

Xi

N

Population

i

1

Xi

n Sample

Statistical Tools

685

• Median—The value representing an equal number of observations

(data) above and below it when the data are arranged in sequence or rank order (increasing or decreasing). Half of the data exist above and half below the median. If there are an odd number of observations, the median is the middle value. If there are an even number of observations, the median is the average of the two middle-most values (for example, two middle-most values divided by two). The median is denoted by M or Md.

The median is unaffected by extreme values, so it is a good measure of central tendency for highly skewed data. • Mode—The value measuring the most frequently occurring obser-

vations (count) in the data set, denoted by Mo. • Sample Calculations—For the following set of data (2, 2, 2, 3, 3, 4, 5,

7, 10, 12), the central tendency values include • Mean = 50/10 = 5

A B C D E F G H I

• Median = 3.5

J

• Mode = 2

K

Figure S-44 illustrates the mean, median, and mode on a right-skewed distribution. The central tendency characteristics of a normal distribution possess equal mean, median, and mode. Median

Mode

Mean

L M N O P Q R S T U V W

Figure S-44: Mean, Median, Mode of a Right-Skewed Distribution

X Y

Variation These measurements indicate how the data is dispersed or spread out. When evaluating a model, variability measures quantify the response’s robustness to sources of variation (both controlled and uncontrolled). It can be measured in three ways:

Z

686

Encyclopedia • Range—The difference between the data set’s extreme values (the

A B C D E

largest and smallest measurements), denoted either by an R or r. The unit of measure is the same as that of the data set. The formula is maximum value minus the minimum value of process data, or R = Xmax –Xmin. The range, minimum and maximum provide a simple way of measuring the amount of consistency in a data set—the spread. • Variance—Measures the sum of the squared deviations from the

mean, divided by the number. It is expressed as the data’s squared units of the standard deviation. Variances can be algebraically added and subtracted.

F G H I J K L M N O P Q R S T U V W X Y Z

n

σ2

i

1

(Xi − µ)2 N

Population

n

s

2

i

1

(Xi − )2 n−1

Sample

Variation is tolerated if the process or system remains within or on target and stable over time. Thus, it is acceptable if within the process specifications or control limits. Minimizing, controlling, or eliminating variation requires an understanding of its sources. Classifying variation helps to understand its sources. Control charts help to isolate between the two types of variation—common cause and special cause. In Six Sigma, variation is the enemy. Variation aims to identify and eliminate special cause variation and minimize and control common cause variation. (See Also “Control Charts—7QC Tool,” p. 217) • Common Cause Variation—Random, stable, and consistent over time. It represents the natural ebb-and-flow of a process or system. There will always be some variation in all processes—natural, inherent oscillation. Everyday examples of variation occur in human or behavioral systems as evidenced by handwriting, tone of voice, and strength. In mechanical systems, variation happens in weight, volume, size, and shape of an item. Other examples include a person’s experience level, an individual’s illness, daily server demand, and daily issues. Within a work environment, common cause variation is viewed as the responsibility of management. Management owns and creates the process, and only management can intervene to change or improve the system. An estimated 85% of the reasons for failure to meet customer expectations are related to deficiencies in systems and processes rather than the employee. • Special Cause Variation—The opposite of random, and it appears erratic or unpredictable over time. On average, 5 to 15% of variation

Statistical Tools

687

is special cause. Examples include worker errors, computer problems, phone system issues, procedural changes, workers out on holiday, and uncommon, unusual, unplanned, often uncontrollable occurrences or circumstances.

Business views special cause variation as a local workforce issue that should be handled by those involved in the process if they have proper tools and operation environment. • Standard Deviation—Measures the positive square root of the vari-

ance. It is expressed in the data’s units of measure. Standard deviation cannot be algebraically added or subtracted.

A B C D E

N

σ

σ2

i

1

(Xi − µ)2

n

s

s2

i

(Xi − )2 n−1

N Population

1

Sample

• Sample Calculations—for the following set of data (2, 2, 2, 3, 3, 4, 5,

7, 10, 12); the variation calculations include

F G H I J K

• Range = 12–2 = 10

L

• Variance = 114/(10–1) = 12.67

M

• Standard deviation = 12.67 = 3.56

N

Shape This describes what the curve of the data looks like, whether it’s normal (and standard normal), t-Distribution (for less than 20 values), bimodal, exponential, or quadratic. Measures of skewedness and kurtosis (peakedness) help to describe the shape of the curve. (See Also “Histogram— 7QC Tool,” p. 330 and “Regression Analysis,” p. 571, for more discussion on shapes and corresponding illustrations.)

O P Q R S T U

• Normal Distribution—Represents a symmetric shape and is uni-

modal, often referred to as a bell-shaped curve. Two parameters define it—the mean and standard deviation. Figure S-45 illustrates a normal distribution’s symmetry by using the number of standard deviations on either side of its mean. Note that the standard deviation also refers to the sigma level and the percent area under the curve.

V W X Y Z

688

Encyclopedia

99.9997%

99.73%

A B

95.44%

C D

68.26%

E F G H I J K L M N

-6

-5

-4

-3

-2

-1

0

1

2

3

4

5

6

Standard Deviations away from the Mean.

Figure S-45: Normal Distribution

Normal distribution is important because of the relationship between the shape of the curve and the standard deviation. Over time many processes follow a normal distribution; thus, it assists in predicting performance over time.

Q

The applications of a normal distribution are many. Not only is it helpful in sampling techniques, it also applies in quality management when constructing the center line, the upper and lower control limits of a control chart, and the PERT chart when examining the probability of completing the project.

R

• Standard Normal Distribution—A special case of a normal distribu-

O P

S T U V W X Y Z

tion where its mean equals zero, and its standard deviation equals one. It serves as a base distribution to compare other distributions. The Standard Normal curve represents a conventional scale for a normally distributed bell-shaped curve that has a central tendency of zero, and a standard deviation of one unit, wherein the units are called sigma (σ). The standard deviation (sigma) is the unit of measure, and the mean, mode, and median all equal zero sigma. Figure S-45 illustrates a standard normal curve, with the Z-scale plotted along the horizontal axis. For every increment in the Z-scale, for every standard deviation, it represents one sigma (σ) unit. Six sigma (6σ) spread translates into 99.9997% of the area under the curve. Hence, the standard normal curve with its Z-scale represents the Six Sigma philosophy. This standard normal scale is used as a common scale to which all other normal distributions with different scales can be translated. An

Statistical Tools

analogy might be that the standard normal scale serves as a common currency. As an individual travels the globe, the currency of their home country needs to be converted into that of the other countries to conduct commerce. Upon return home, the traveler converts the other nations’ currency back to that of the home country. The standard normal scale serves as that common currency for all normal distributions. Regardless of the unit of measure, normal distributions can be re-scaled, or converted, into a standard normal scale to compare distributions, areas under the curve, and proportions between and across other distributions, using a common sigma scale, as shown in Figure S-46.

689

A B C D E F G H

68.26%

I J

95.46%

K

99.73%

L

σ=

-3

-2

-1

0

1

2

3

$=

59

71

83

95

107

119

131

M N O P

¥ = -1025

-562

-99

364 827

1290

1753

Figure S-46: Standard Normal Curve

Q R S

The specific bell-shaped distribution (for example, the shape of the normal data) remains constant, hence the proportion of the area under the curve remains constant. The area in a Standard Normal curve is delineated in sigma units (recall that standard deviation is one). If two points on its X-axis are known, they demarcate the corresponding probability that a given variable will fall between them, as shown in Figure S-46. Hence, regardless of the scale (or units of measure), the shape of the Standard Normal distribution can be used to calculate the probability of an event and predict the range of its value—statistical inference. The various values of a standard normal curve are called Z-scores. A Z-score can be calculated or looked up in a Standard Normal table. The formula to convert a distribution into a standard normal distribution is as follows, regardless of units of measure. (It is a similar procedure to converting currency from one denomination to another.)

T U V W X Y Z

690

Encyclopedia

Zi

A B C D E F G H I

(Xi − µN ) σN

Wherein, Z is the variable of interest minus population mean, divided by the population standard deviation. (See Also “Standard Normal Distribution,” in Appendix B, p. 975) • Student t-Distribution—Used for small sample sizes of 30 or less

and when the standard deviation (sigma) is known. Its shape is affected by the sample size, such that as sample size increases, the degrees of freedom (dof) increase, and the shape approaches a normal distribution, as shown in Figure S-47. The values cluster near a single most frequently occurring value, and 95% of the area under the curve is covered in 1.96 sigma. The t-distribution is used in confidence intervals and t-tests (for Hypothesis testing). (See Also “Analysis of Variation (ANOVA)—7M Tool” and “Hypothesis Testing,” p. 142 and p. 335, respectively for details on degrees of freedom and t-tests.)

J 0.4

K L

0.35

M N O P Q R

0.3

dnorm (z, µ , σ)

0.25

dt (z, dof1) dt (z, dof5 )

0.2

dt (z, dof20) 0.15

S T U V W X Y Z

0.1

0.05

0

7

6

5

4

3

2

1

0

1

2

3

4

5

6

7

Figure S-47: Student t-Distribution Compared to a Standard Normal Distribution

• Binomial Distribution—Models two outcomes of discrete data as

part of the same distribution. The equation involves a constant probability of success, denoted by p and a constant probability of failure, denoted by q, which equals 1–p. The equation for the probability of X number of successes in n trials is:

Statistical Tools

691

n! p x(q n − x) x!(n − x)!

P(x)

This concept applies in quality management, to create control charts for attribute data, namely the np-chart and the p-chart. (See Also “Control Charts—7QC Tool,” p. 217) • Poisson Distribution—Models random occurrences over time of

discrete data with essentially no upper bound, where the number of occurrences of an event during an interval is independent among intervals. Examples include time, distances, volumes, area, rates, and flows. This applies to systems where it is difficult to count all non-occurrences (that is, defect counts), to approximate the binomial distribution. The probability of occurrence is constant among intervals, and in a small interval is small and proportional to the length of the interval. The equation is: P(x)

λxe−λ x!

A B C D E F G H I J K

, where e = 2.71828.

L

This concept applies in quality management in constructing control charts for attribute data based on the Poisson distribution, namely the c-chart and the u-chart. (See Also “Control Charts—7QC Tool,” p. 217) • Down-Side Variance (DSV)—Examines the sample values smaller

than the mean (Xis). Its formula is: n

DSV

i

O P

(Xis − )

R

n−1

S

, where n is the number of data values in the entire set. • Skewness (SK)—Examines the shape of a sample distribution. Its

formula is the following ratio, where s2 is the sample’s variance:

SK

N

Q 2

1

M

S2 2(DSV )

If SK > 1; then the distribution is skewed to the right (positively skewed). If SK < 1; then the distribution is skewed to the left (negatively skewed). If SK = 1; then the distribution is symmetrical.

T U V W X Y Z

692

Encyclopedia

When using MINITAB and Crystal Ball software packages, note that they utilize a different formula to calculate skewedness; hence, the skewedness determination becomes zero: n

A B C D

SK

i

1

(Xi − )3

(s3 N)

If SK > 0; then the distribution is skewed to the right (positively skewed).

E

If SK < 0; then the distribution is skewed to the left (negatively skewed).

F

If SK = 0; then the distribution is symmetrical.

G H I J K L M N O P Q R S

• Pearson’s Coefficient of Skewness (γ–gamma)—another formula to

calculate skewness using population data, where the population mean (µ), the median (M), and population’s standard deviation (σ) are used:

3(µ − M )

γ

σ

Central Limit Theorem This theorem states that regardless of the shape of the population distribution, the distribution of the means calculated from several samples tends to become normal as the sample size increases. This concept can be expressed as an equation: σ

2

σx2

n

T

, where n is the number of observations.

U

Thus, with increasing sample size, the mean of the distribution of sample means eventually will equal the mean of the population, as illustrated in Figure S-48. The standard deviation of the distribution of sample means is equal to the population standard deviation divided by the square root of sample size.

V W X Y Z

Figure S-48 illustrates how the graph of the means from several samples is a normal distribution and tighter around the mean than the plot of individual data points. The significance of the Central Limit Theorem on control charts is that the distribution of sample means approaches a normal distribution, regardless of the shape of the parent population. If a control chart of sample averages is plotted, on average only about 0.27% of the time the sample mean falls outside the control limits if the process remains unchanged. This is calculated as 1–99.73% for three sigma on either side of the mean.

Statistical Tools Sample 2 of size n = Sample 1 of size n; Average =

693

X2

X1

Xi A

X

X

Distribution of all the Sample Averages

Distribution of Individual Observations

C D

Histogram of C10, Xbar Normal

E

Var iab le C10 Xb ar

2000

Frequency

B

Mean StDev 99.84 10.09 99.99 2.025

1500

N 10000 10000

F G

1000

H

500

0 70

80

90

100

110

120

130

Data

I J K

Figure S-48: Central Limit Theorem

L

Inferential (or Comparative) Statistics Inferential statistics are used with sampling to infer knowledge about the sample and its characteristics onto that of the population. (See Also “Sampling,” p. 618) Parameters This refers to the summary measures for the population. Parameters are usually represented by lower case Greek letters, for example • Population or Process Mean, denoted by µ (pronounced mu). • Population or Process Variance, denoted by σ2 (pronounced sigma-

squared).

M N O P Q R S T U V

• Population or Process Standard Deviation, denoted by σ (pro-

nounced sigma).

W X

• Population Size, denoted by N.

Sample Statistics This set of measures provides summary information about the sample. They are usually represented by lower case English letters, for example: _

• Sample Mean, denoted by X (pronounced x-bar). • Sample Variance, denoted by s2 (pronounced s-squared).

Y Z

694

Encyclopedia • Sample Standard Deviation, denoted by s. • Sample Size, denoted by n.

A B C D E F G

Hypothesis Testing This is a robust topic, as statistical analysis represents multiple types of tests. A unique set of tests cover enumerative data (counted data), such as attributed, classification, or categorical data. Enumerative Hypothesis tests include Chi Square, binomial, and Poisson distribution. The schematic found in Figure S-49 serves as a summary guide to the various questions and considerations needed when selecting the appropriate Hypothesis test based on the type of data being analyzed. Figure S-49 also includes the MINITAB commands to execute the various Hypothesis tests. (See Also “Analysis of Variation (ANOVA)—7M Tool,” p. 142, and “Hypothesis Testing,” p. 335, for more detailed information.)

H I J

Hypothesis Testing Guide

Levene's Test H O : σ 1 = σ 2 = σ 3 ... H A : σ i ≠ σ j for i ≠ j

L

Stat>ANOVA> Homogeneity of Variance

N

Fail to reject HO

1 level

O

2 levels or > 2 levels?

Test median or sigma?

P

2 or more levels

1, 2 or more levels?

Q

Chi-Square Test

HO : M 1 = M 2 HA : M 1 ≠ M 2

H O : σ1 = σ t HA : σ 1 ≠ σ t

R

Stat>Non-parametric> Mann-Whitney

W X Y

2 or more levels

Mood's Median Test (used with outliers) H O : M 1 = M 2 = M 3 ... H A : M i ≠ M j for i ≠ j

Kruskal-Wallis Test (assumes outliers) H O : M 1 = M 2 = M 3 ... HA : M i ≠ M j for i ≠ j (or at least one is different)

Stat>Non-parametric> Kruskal-Wallis

2+ Factors?

2+ Factor

HO : F A Independent FB HA : F A Dependent FB

Is data normal?

1 Factor

Variable

1 Factor

1, 2 or more Factors?

1 or 2 Samples

1-Proportion Test HO : P 1 = P t HA: P 1 ≠ P t

1 Sample

t = target

Stat>Basic Stat> 1-Proportion

Note: Consider transforming non-normal data to use more standard comparative tools. Nonparametric tests are often weaker.

2 or more Factors 2-Proportion Test

2 Samples

HO : P 1 = P 2 HA : P 1 ≠ P 2

ANOVA or DOE

Data Normal

Stat>Basic Stat> 2-Proportion

1, 2 or >2 levels?

More than 2 levels

Bartlett's Test HO : σ1 = σ2 = σ3 ... HA: σi ≠ σj for i ≠ j

Test for means

(or at least one is different)

Stat>ANOVA>Test of Equal Variance 1-Way ANOVA (assumes equality of variances) HO : µ 1 = µ 2 = µ 3 ... HA : µ i ≠ µj for i ≠ j

t = target

Test Medians

Stat>Basic Stat>Display Desc> Graphical Summary (if target sigma falls between CI, then fail to reject HO )

(or at least one is different)

Stat>Non-parametric> Mood's test

Data Not Normal

Attribute

Stat>Tables>Chi 2 Test

Test for sigmas

2 levels only Mann-Whitney Test

V

Variable or Attribute Data?

(or at least one is different)

M

U

Contingency Table

START >>>

Stat>Basic Stat>Normality Test or Stat>Basic Stat>Descriptive Statistics (graphical summary)

If Ho is rejected, then you can go no further

T

Stat>Con Chart/p-chart 2 Stat>ANOVA/Anal Means

Ho: Data is normal Ha: Data is not normal

K

S

ANOP/BNAM HO : F A Independent FB HA : F A Dependent FB

Notes: • If P > alpha (default @ 0.05), then fail to reject Ho • If P < alpha (default @ 0.05), then reject Ho • Ensure the correct sample size is taken.

1 level

1-Sample Wilcoxon or 1-Sample Sign HO : M 1 = M t H A : M 1 ≠ Mt

Test for sigmas

Test for means

Is Data Dependent?

(or at least one is different)

Stat>ANOVA>1-Way (then select stacked or unstacked data)

Test for means 1-Sample t Test HO : µ1 = µ t HA : µ 1 ≠ µ t t = target

Test for sigmas

No, Data is drawn independently from two populations

Yes, Data is Paired

Stat>Basic Stat> 1-Sample t

t = target

Stat>Non-parametric> and either 1-Sample Sign or 1-Sample Wilcoxon

Test for mean or sigma?

2 levels Test for mean or sigma?

Chi-Square Test HO : σ 1 = σ t HA : σ 1 ≠ σ t t = target

Stat>Basic Stat>Display Desc> Graphical Summary (if target sigma falls between CI, then fail to reject Ho)

F Test H O : σ1 = σ 2 HA : σ 1 ≠ σ 2 Stat>ANOVA> Homogeneity of Variance

2-Sample t Test HO : µ1 = µ2 HA: µ1 ≠ µ2 Stat>Basic Stat> 2-Sample t (if sigmas are equal, use pooled std dev to compare. If sigmas are unequal compare means using unpooled std dev)

Paired t Test HO : µ1 = µ 2 HA : µ 1 ≠ µ 2 Stat>Basic Stat> Paired t

Figure S-49: A Guide for Hypothesis Testing

Z

Correlation and Regression Analysis Regression analysis quantifies the type and shape of the relationship between two or more variables to model a population or system (either mechanical or behavioral). (See Also “Scatter Diagrams—7QC Tool” for correlation details, p. 640, and “Regression Analysis,” p. 571, for more detail.)

Statistical Tools

Design of Experiment (DOE) DOE is another important statistical tool set to demonstrate cause-andeffect relationship. It encompasses different test designs such as screening, full-factorial, fractional factorial, optimizing, response surface, robustness (Taguchi), confirming, mixture and conjoint analysis. (See Also “Design of Experiment (DOE),” and “Conjoint Analysis,” p. 250 and 207, respectively.) Probability The probability of an event is the value that ranges between 0 and 100% (or 0 to 1). If the event were called A, then the probability of that event occurring is denoted as P (A). The probability of A occurring would equate to: P (A) = (# of occurrences)/(# of total possibilities). Hence, using the preceding definition, an event that cannot occur (an impossible event) could be written P (A) = 0. Conversely, an event that were certain to occur (a certain event) could be written P (A) = 1. Typically, the probability of an event falls between impossible and certain and is expressed as 0 < P (A) < 1. The sum of the probabilities of a fixed number events (n), would be mutually exclusive and collectively exhaustive events, where the probability of n occurring is 100% or 1. When two events represent all the possible events, they are referred to as complementary events. If the total number of possibilities were binary (good or bad), then the probability of any one of those two events occurring would be complementary, in that P (good) = 1–P (bad). When writing the Null and Alternative hypothesis statements, they are expressed as mutually exclusive and collectively exhaustive of all the possible outcomes. For example, if the Null hypothesis were “mean 1 equals mean 2,” then the Alternative hypothesis must be “mean 1 does not equal mean 2.” Thus, the Null hypothesis is statistically complementary to the Alternative hypothesis, which can be expressed as P (Null) = 1–P (Alternative). (See Also “Hypothesis Testing,” p. 335)

Additive Law Risk management often applies the additive law of probability. If a risk event were independent from other events, such that the probability of its occurrence remained constant, historical data could be used to determine the likelihood of future occurrences. In probability terms, this describes a union, denoted by a ∪ symbol and the portion that covers the overlap or intersection, denoted by a ∩ symbol. If they two events are mutually exclusive, then the probability can be expressed as: P (A ∪ B) = P (A) + P (B).

695

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

696

Encyclopedia

If the two events (A and B) are not mutually exclusive, they can be added together. The equation expressing the additive law is: P (A ∪ B) = P (A) + P (B)—P (A ∩ B). (See Also “Risk Mitigation Plan,” p. 601)

A B C D E F G H I J K L M N O P Q R

Multiplicative Law Quality management uses the multiplicative law of probability to examine reliability of two components. If the components are assembled in series, the reliability of both is less than the reliability of any one component by itself, such as throughput yield. If any two events (A and B) are independent, the probability of A and B occurring can be written as: P (A ∩ B) = P (A) x P (B). (See Also “Process Capability Analysis,” p. 486, for a discussion on throughput yield.) The components also operate in parallel (Ap and Bp), where one is operational and the other is redundant, the reliability of these two components is higher than the reliability of any one component. The equation is: P (Ap) = 1–P (A); P (Bp) = 1–P (B); and P (Ap ∩ Bp) = P (Ap) x P (Bp). Sometimes Ap and Bp are denoted as A-bar and B-bar, respectively.

Graphical Tools The primary graphical tools include • Boxplot—A type of frequency plot used to summarize data (continu-

ous Y and discrete X). (See Also “Boxplots—Graphical Tool,” p. 165) • Control Chart (Super-Charged Run Chart)—Used to monitor a

process and requires continuous data in sequence of production or observation. (See Also “Control Charts—7QC Tool,” p. 217) • Dotplot—A type of frequency plot used to stratify data (continuous

S

Y and discrete X). (See Also “Dotplot,” p. 280 and “Stratification Tool—7QC Tool,” p. 445)

T

• Frequency Plot (Histogram)—Used to determine the shape, center,

U

and range of continuous or numeric data. (See Also “Histogram— 7QC Tool,” p. 445)

V W X Y Z

• Pareto Chart—Used to determine the relative contribution of differ-

ent items (such as defects or problem areas) for discrete or categorical data. (See Also “Pareto Chart—7QC Tool,” p. 445) • Run Chart (or Time Series Plot)—Used to identify data patterns

such as special causes, shifts, and trends. Continuous data in sequence of production or observation are required. (See Also “Run Chart—7QC Tool,” p. 611)

Stratification—7QC Tool

697

• Scatterplot—Used to determine if a relationship exists between two

discrete or continuous variables; correlation. Also used with Stratification. (See Also “Scatter Diagram,” p. 640 and “Stratification,” p. 697) See “Graphical Methods,” on p. 323 for a summary of the different tools.

Process A statistical tool helpful to evaluate a process is the Process Capability techniques, including Cp, Cpk, Pp, Ppk and throughput yield analysis. As previously mentioned, process tools examine the inputs and outputs of a process for variation and defects. Other statistics applied to process cycle time span the relatively simple tools such as graphical tools and variation analysis, to robust tools such as Monte Carlo simulation. (See Also “Process Capability Analysis,” p. 486, and “Monte Carlo Simulation,” p. 431) Measurement System Analysis (MSA) Because statistical thinking focuses on three core elements—process, variation, and data, it is critical that the data collected, graphed, analyzed represent what really happens in the system or process. The MSA technique ensures that the measurement system is accurate and reliable for either continuous or discrete data, and not introducing any noise into the data. It validates that whatever variation is observed is due to variation in the item of interest, and not due to a sloppy measurement system. The measurement system is comprised of the measuring device, the operator (or measurer), and the item of interest. (See Also “Measurement System Analysis (MSA),” p. 412)

Stratification—7QC Tool

A B C D E F G H I J K L M N O P Q R

What Question(s) Does the Tool or Technique Answer? Are there any patterns in the data? Is the data set homogeneous?

S

Stratification helps you to

U

T V

• Reveal patterns in the data

W

• Understand why the data set in aggregate does not make sense

X Y

When Best to Use the Tool or Technique The stratification tool should be combined with other data analysis tools; however, forethought is required to ensure that the proper stratification information is collected in case this technique proves helpful during analysis in seeing patterns otherwise hidden.

Z

698

Encyclopedia

D

Brief Description Stratification is a data analysis technique that parses data into its natural subgroups—or strata. Combine this technique with other data analysis tools to segregate the data into its different sources. Examples of different data sources include: different suppliers, shifts, departments, locations, equipment types, products, time of day, day of week, month of year. For process data, stratification cuts the data into its different sources like layers of a stratified rock. Interestingly, these different stratifying sources can be further segregated into subgroups, often linked to a special cause. Stratifying the data allows for patterns to be seen, that otherwise would be masked by the larger population.

E

Examples of different applications might involve the following situations:

A B C

F G H I J K L M N O P Q R S T U V W X Y Z

• Stratifying survey data into the demographics of the respondents

may reveal patterns that otherwise would be unseen when analyzed in aggregate. • Examining profit per employee may be inconclusive until the data is

stratified to geography to uncover economic differences and buying behaviors by region. • Evaluation of throughput yield may appear homogeneous until the

data is parsed by shift. • Understanding an organization’s revenue growth potential or weak-

nesses may seem unclear until the offerings portfolio was analyzed by offering with its respective competitive threats. In the world of quality, there is a fine distinction between stratification and segmentation. Stratification is related to segmentation but is different. Stratification defines the different sources of data, while segmentation often represents a sampling technique. Segmentation has a broader meaning describing categorization, such as classification or types of an item. Examples of segmentation might be kinds of fruit, kinds of defects, market types, customer types. If stratification appears to be a potential influence, plan accordingly prior to data collection so as to properly label the source during the data gathering activities. Moreover, the data collection tool will need to track and record the different sources to enable strata analysis later. Presuming that the proper information has been collected, software tools such as MINITAB make it easy to explore stratification. After looking at a graph, such as a Dotplot, Multi-vari plot or Scatterplot, if the output appears inconclusive, MINITAB allows the pointer to be dragged over an area of data for quick viewing of common strata. If it appears to play a factor, simply rerun the graph specifying the stratification be

SWOT (Strengths-Weakne sse s-Opportunitie s-Threats)

699

identified by other different data icons and sometimes unique regression lines. (See Also “Dotplot,” p. 280; “Multi-vari Chart” p. 439; and “Scatter Diagram—7QC Tool,” p. 640) Figure S-50 compares a pair of Scatter diagrams with the same profit against FTE (full-time equivalents) data—one with versus one without stratification. The one without stratification appears inconclusive. The stratified graph exhibits a different FTE to profit relationship in the North (represented by a circle) versus the West (represented by a square) regions, as indicated by their respective best-fit lines.

A B C D

Scatterplot of Profit vs FTE

50

50

45

45

40

40

Profit

Profit

Scatterplot of Profit vs FTE

35

30

25

20

4

5

6

7

8

F H I J

2

3

4

FTE

Without Stratification

E G

30

20

3

North West

35

25

2

Region

5

6

7

8

FTE

With Stratification

Figure S-50: Comparing Scatter Diagrams With and Without Stratification

Stratification is an original member of the 7QC Tools, (or seven Quality Control tools), attributed to Dr. Kaoru Ishikawa. The 7QC Tools sometimes are called the seven basic tools given that they were the first set of tools identified as the core quality improvement tools. Ishikawa’s original 7QC Toolset includes: Cause-and-Effect diagram; check sheet (or checklist); control charts; histogram; Pareto chart; Scatter diagram; and stratification. More recently, the 7QC Toolset has been modified by substituting the stratification technique with either a flowchart (or Process map) or a run chart (or Time Series plot).

SWOT (Strengths-Weaknesses-Opportunities-Threats)

K L M N O P Q R S T U V W X

What Question(s) Does the Tool or Technique Answer? How does the organization’s strengths and weaknesses compare with the competitive opportunities and threats? SWOT analysis helps you to • Identify the critical marketplace characteristics and organize them as

a balanced scorecard to make strategic decisions

Y Z

700

Encyclopedia • Evaluate the balance between internal and external factors for a par-

ticular organization • Identify and prioritize which market segments fit the organization

A B C

When Best to Use the Tool or Technique During the strategic planning process, use a SWOT analysis to aid in formulating strategic direction and management goals and objectives.

D E F G H I J K L M N O P Q R S T U V W X Y Z

Brief Description The Strengths-Weakness-Opportunities-Threats (SWOT) analysis focuses management on the competitive landscape and how well its organization competes. The SWOT tool identifies, organizes, and ranks internal and external factors into a quadrant matrix to analyze the balance between them. The technique examines which market segments optimally fit the organization’s capabilities and accepted levels of risk. It sets internal and external boundaries for a segment’s attractiveness and helps the portfolio management team make decisions about the offering concepts to fund and to initiate as development projects. The SWOT analysis presumes that an organization tries to fit within its external environment. This technique produces a summary plot of segments on a grid of ranked opportunity space in relation to the competitive position. It lists the organization’s strengths, weaknesses, opportunities and threats related to a particular market segment being evaluated. Conduct a unique SWOT analysis for each given region of interest, as the external environment differs from place to place. The SWOT analysis embodies two challenges. First, the organization must objectively evaluate its internal strengths and weaknesses and relative position to competition. Management tends to amplify strengths and underplay weaknesses, thereby negating the SWOT process. Collecting the perspective of objective observers may serve the organization well. Second, the analysis represents only the initial phase in a strategic plan; the next phase requires the execution and delivery of an action plan.

External Environment SWOT analysis requires an understanding of the marketplace. Benchmark data is a prerequisite to determine the competitive environment, the emergence or exit of competitive players, and how they affect your organization. Include marketplace data about the technological environment to understand the available or emerging technologies. A social environment snapshot may be important, particularly if the organization considers entering a new market. Social elements include the value system, social

SWOT (Strengths-Weakne sse s-Opportunitie s-Threats)

and demographic patterns that might impact the organization’s ability to operate in that region. For example, evaluate whether the region possesses the skills and knowledge workers required. Closely related to the social landscape is the socio-political environment. Evaluate the state (stability and influence) of and potential impact of the different governing bodies (local, regional, national), special interest groups, and/or legal and regulatory agencies. Understand the state of the economic environment for the region and marketplace, the trends, and any key events that might impact the organization. The acronym PEST acts as a reminder of the main external forces: Political, Economic, Social, and Technological. Apply this tool to each country or region in which the organization operates (or plans to operate), as these PEST forces differ by location. Variations on the PEST analysis tool include SLEPT (Social, Legal, Economic, Political, and Technological) and STEEPLE (Social/demographic, Technological, Economic, Environmental (natural), Political, Legal, and Ethical). Again, these multiple factors each may alter as the geographic location changes, so repeat the SWOT analysis for each regional situation. Opportunities are characterized by three attributes: financial, market behavior, and disruptive technology. Fulfilling market behavior needs results from a proper balance between marketing and technical innovation. An organization can benefit from taking advantage of disruptive technology opportunities by having a strong focus on technical innovation. Opportunities provide an environment for ideas to be tested, ranked, and prioritized for portfolio development, renewal, and balancing for the desired level of financial growth. Threats are usually traceable to a gap between your organization and some form of competitive force in the served markets and segments, such as actions from direct competitors or indirect competitors using a disruptive technology. Threats can come from uncontrollable sources of variation in the market environment. Currency fluctuation, price elasticity, social change, political upheaval, international affairs, and man-made or natural disasters all represent sources of external threats. Threats add risk to your ability to exploit on opportunity. Opportunities and threats represent the external factors that may create or destroy value in the marketplace. These represent uncontrollable factors to a large extent. Comprehend these external dynamics to determine how best to operate within this environment. An organization’s strategy must align with the external opportunities to best leverage its capabilities and defend against the external threats. The organization’s strategy should consider possible changes (trends) in the external environment to determine its appropriate direction and positioning.

701

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

702

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

Internal Environment The organization’s strengths and weaknesses evolve from its internal environment made up of organizational assets, resources, and skills. These internal factors create (or destroy) marketplace value relative to competition. They are measured and evaluated not only by the organization itself, but also the marketplace. An organization’s strengths usually come from its core competencies. A strength is something an organization is doing well. It is characterized by the capabilities and capacities to exploit an opportunity within or across a market. Strengths underwrite the organization’s ability to successfully design, develop, distribute, and deliver value (product and/or services) to the market. Strengths exploit the external opportunities and fend off the external threats. Internal strengths have many dimensions—anything that provides a competitive advantage. They can be depicted as new ideas, current products and services to leverage from, proprietary technologies, patents, intellectual property, skills, key resources, and expertise. Marketing capability and capacity to create brand awareness and position and a reputation for quality are all strengths. Portfolio development and balancing, commercialization and launch, channel management, distribution, sales, support, and product-line management also describe internal capabilities and capacities. Strengths include technical capability and capacity, research and development, design, manufacturing, and the service and supply chain. Moreover, an organization’s financial resources, partnerships, alliances, and joint ventures are invaluable assets. Weaknesses are areas wherein the organization lacks something or a condition that puts it at a disadvantage relative to its competitors. A weakness usually is traceable to a gap in core competencies or to poor execution within product development or post-launch processes. In addition to poor quality, weakness also may include poor cash flow, outdated technology, high overhead expenses, or unavailable skilled labor. Weaknesses represent barriers to the organization’s ability to exploit an opportunity or ward off a competitive threat.

Competitive Advantage In summary, an organization’s competitive advantage embraces the aggregate internal strengths and weaknesses to counter the competition’s opportunities and threats. Michael Porter proposes that there are two types of competitive advantage—cost leadership (low cost) and differentiation—which in combination create a third dimension—focus. Economies of scale play an important role in cost leadership. Wal-Mart is an example of a firm identified as a cost leader. If more than one firm tries to achieve cost leadership in a market segment, eventually it leads to disastrous conditions.

SWOT (Strengths-Weakne sse s-Opportunitie s-Threats)

Differentiation often serves as a counter-balance to a cost leadership position, as an organization attempts to deliver a product and/or service of unique value. To do so, it requires addressing new, unique, and difficult (NUD) requirements. At minimum, price the differentiated offering at a premium to its production costs. Apple’s iPod is a good example of a differentiated MP3 player that delivers on NUD requirements with its design and iTunes music offerings. (See Also the “Kano Model: NUD Versus ECO” section under “KJ Analysis,” p. 376) Achieving focus describes the degree with which an organization drives for results—best-in-class. The competitive strategy may call for a narrowing or widening approach to best address a particular market segment. Focus may alter a strategy to go after a different market niche. Again, Apple’s current market success may be the impetus behind its focus in launching the new products such as the cost-competitive Shuffle, the price premium video-iPod, or the iPhone that takes it into a new market segment. Ultimately, an organization’s strategy must match its internal capabilities with the external environment. The target market opportunities complement its capabilities, its defenses are suited to foil the external threats, and it’s suited for any changes.

703

A B C D E F G H I J K L M

How to Use the Tool or Technique The process to conduct a SWOT analysis, assemble a cross-functional, multi-discipline team, is as follows: Preparation: Work on flip chart paper so that everyone on the team can see the work simultaneously. You may choose to write directly on the flip charts or use sticky notes to post on the wall or a white board. Use a round-robin approach by arming each person with a pen or marker or giving each person a pad of sticky notes, which aids in capturing everybody’s input. Ensure that the proper data sources are collected and brought to the meeting.

SWOT Matrix Step 1. Construct the SWOT matrix. Create a 2x2 matrix by dividing a page into quadrants. a. Label each quadrant. Convention places strengths in the upper-left corner, weaknesses in upper-right corner, opportunities in lower-left corner, and threats in the lower-right corner.

N O P Q R S T U V W X Y Z

704

Encyclopedia

Step 2.

Create a list of Opportunities and Threats (O/T) and document them on a flip chart or in the appropriate cell of the SWOT matrix. Utilize the PEST, SLEPT, or STEEPLE techniques. Sources include a. Market and segmentation analysis

A

b. Primary and secondary market research data

B

c. Long range, over-the horizon market trend and behav-

C

ioral analysis

D E

d. External benchmarking studies

F

e. Market environmental force analysis

G

Step 3.

H I

Create a list of Strengths and Weaknesses (S/W) and document them in the appropriate cell of the SWOT matrix. Sources include a. Internal assessments, audits and analysis, capability

J

analysis of design, development, manufacturing, launch, sales, and service and support capabilities

K L M

b. Benchmark studies of internal strengths and weaknesses

N

c. Business and innovation strategy development, documen-

O

tation, and execution of the strategic plan

P

d. Product portfolio performance analysis/audit

Q

e. Training and use of tools and best practices

R

f. Project management skills and cycle-time performance

S T

g. Financial reports

U

h. Product management effectiveness over life-cycles

V W X Y Z

Step 4.

Post the completed SWOT matrix in the room for reference. This chart helps to communicate the high-level scenario, as illustrated in Figure S-51.

Candidate Market Segment Evaluation Matrix Given a couple of candidate segments, apply the SWOT matrix to begin developing a unique strategy for each. Similar factors may exist in multiple segments, but deal with each segment independently and duplicate the item of interest.

SWOT (Strengths-Weakne sse s-Opportunitie s-Threats)

Strengths (internal) • Brand recognition • Patents • Intellectual property • Good quality • Feature innovation

Weaknesses (internal) • • • •

Lack of design innovation Too many products poorly differentiated Only “industrial strength” version Limited sales channels

Opportunities (external) Threats (external) • • • •

Developing markets Mergers and acquisitions Association market segment Internet sales

705

• •

Innovative marketing and selling Fashion designs

A B C D E F G

Figure S-51: SWOT Matrix

H I

Step 5.

Create the Market Segment Evaluation Matrices to further dissect the SWOT matrix, as illustrated in Figure S-52. (See Also “Matrix Diagrams—7M Tool,” p. 399, for a discussion on different shapes.) a. Using an L-shaped structure, create two five-columned matrices, presuming three market segments to evaluate. If there are more, increase the number of columns accordingly; or if less, decrease the number accordingly. b. Label one “Opportunity Market Segment Evaluation” matrix. i. Label the column headings: Column 1 as “Opportunities/Threats (O/T)”; Column 2 as “Weight”; Columns 3, 4, 5 as “Market Segment 1, 2, 3,” respectively. ii. Subdivide Columns 3, 4, 5 into two more columns each. Label each of the sub-column pairs from left to right, as “Rating” and “Score.” c. Label the second “Competitive Position Market Segment Evaluation” matrix. i. Label the column headings: Column 1 as “Opportunities/Threats (O/T)”; Column 2 as “Weight”; Columns 3, 4, 5 as “Market Segment 1, 2, 3,” respectively. ii. Subdivide Columns 3, 4, 5 into two more columns each. Label the each of the sub-column pairs from left to right, as “Rating” and “Score.”

J K L M N O P Q R S T U V W X Y Z

706

Encyclopedia Opportunities/ Threats

Weight

Segment 1

Segment 2

Rating

Rating

Score

Segment n Score

Rating

Score

This Table will represent the Opportunities axis on the SWOT Summary Chart…

A B

Strengths/ Weaknesses

C

Weight

Segment 1

Segment 2

Rating

Rating

Score

Segment n Score

Rating

Score

D E

This Table will represent the Competitive Position axis on the SWOT Summary Chart…

F G

Figure S-52: Template for Market Segment Evaluation Matrices

H I J K L M N O P Q R S T U V W X Y Z

Step 6.

Determine the Weight of the O/T and S/W factors, as illustrated in Figure S-53. a. Create two lists, one by combining the SWOT opportunities and threats into a consolidated O/T list and the other by combining the SWOT strengths and weaknesses into a consolidated S/W list. b. Working with one list at a time, stack rank the items within the list in order of importance. Document the items in the first column of the appropriate Market Segment Evaluation matrix by placing the most important item at the top and then flow down the remaining items in descending order. c. Repeat Step 6.b. for the other list and fill the appropriate Market Segment Evaluation matrix. d. Working with one list at a time, allocate 100 points across each item within the list and record the appropriate pairing in the Weight column. i. Assigning numeric weight represents the amount of

attention or investment the issue must receive for the business to succeed in the identified market. ii. Some factors may be irrelevant to a segment and would then get a weight of 0. e. Add a row at the bottom of the matrix to hold the totals and total the Weight column to double-check that all 100 points have been dispensed. f. Repeat Step 6.d. for the other list, to weight the items.

SWOT (Strengths-Weakne sse s-Opportunitie s-Threats)

Step 7.

707

Rate each candidate market segment, as illustrated in Figure S-53. a. Market Opportunity rating. Use the following rating scale to measure the impact of each line item: i. 3 = Most Impact on the segment; very good ii. 2 = Better than most; above average impact iii. 1 = Normal impact

A B C

iv. 0 = No impact at all; not good enough

D

v. -1 = Detrimental impact; threat is high

E

b. Rate each item in the Opportunity Market Segment Evaluation matrix, using the preceding Market Opportunity rating scale. Record the score in the appropriate cell and total the Score column. i. Opportunities can exist now or appear during the

planning phases. ii. Focus on present position, trends, and future

expectations. iii. A realistic projection is the goal.

F G H I J K L M N

c. Repeat Step 7.b. for each of the market segments until all are rated and totaled for their market opportunity.

O

d. Competitive Position rating. Use the following rating scale to measure the impact of each line item:

Q

i. 3 = Our position provides a maximum effect and

impact. ii. 2 = Our position has an above average effect on the

market.

P R S T U V

iii. 1 = Our position has an expected, normal impact.

W

iv. 0 = Our position would have no competitive effect or

X

impact. v. -1 = Our position signifies poor effect—we are not

competitive. e. Rate each item in the Competitive Position Market Segment Evaluation matrix using the preceding Competitive Position rating scale. Record the score in the appropriate cell and total the Score column.

Y Z

708

Encyclopedia i. Strengths and weaknesses exist now.

f. Repeat Step 7.e. for each of the market segments until all are rated and totaled for their competitive position. Step 8. A

Calculate the SWOT Value Scores. a. Working with one Market Segment Evaluation matrix at a time, by row, multiply the weight by the rating score for each factor to calculate each factor’s T/O score and a S/W score, respectively.

B C D

b. Total each Score column, by segment, for each Market Segment Evaluation matrix.

E F G

Opportunities/ Threats

H

Weight

I J K

Totals

L

Score

45

3

135

25

2

50

20

1

20

10

-1

-10

100 Step 6

M

Strengths/ Weaknesses

N

Segment 1 Rating

Weight

Segment 1 Rating

Score

1

50

P

35

1

35

Q

15

3

45

5

-1

-5

100

Segment n Rating

Score

Step 8

50

Totals

Score

195 Step 7

O

R

Segment 2 Rating

Segment 2 Rating

Score

Segment n Rating

Score

125

S T U

Figure S-53: Market Segment Evaluation Matrices

V

Strategic Decision Grid

W

Step 9.

X

Assess the results and document the findings, as illustrated in Figure S-54.

Y

a. Create a 3x3 matrix, label it “Strategic Decision Grid.”

Z

b. Label the vertical axis “Opportunity” to represent the market opportunities found in the respective Market Segment Evaluation matrix (the strengths/weaknesses). c. Label the horizontal axis “Competitive Position” to represent the competitive positions found in the respective Market Segment Evaluation matrix (the opportunities/threats).

SWOT (Strengths-Weakne sse s-Opportunitie s-Threats)

709

d. Using the following alphabetical scale, code each cell accordingly. i. A = Good Opportunity and position. ii. B = Moderate to reasonably good opportunity/position.

A

iii. C = Less than optimal opportunity/position.

B

iv. Top Row, from left to right, label each cell B, A, and

C

A, respectively.

D

v. Middle Row, from left to right, label each cell C, B,

and A, respectively.

E F

vi. Bottom Row, from left to right, label each cell C, C,

and B, respectively.

G H

e. Plot the two summary scores for each segment on the graph.

I

f. Use the market size data (that is, potential sales volume) to scale an icon (such as a circle) that represents each market segment. Plot the appropriately sized market segment icons in the appropriate grid on the graph.

K

g. Determine which segment is providing the strongest opportunities and least resistance to entry or competition. 300

A

V

C

C

W

B

X Y

200 0 -100

0

P

U

Segment 4

-100

O

T

A

Segment 3

100

N

S

Segment 1

100

Opportunities

B

M

R

A

Segment 2

C

L

Q

Strategic Decision Grid

B

J

-100 100

200

300

Competitive Position A= Good Opportunity and position B= Moderate to reasonably good opportunity/position C= Less than optimal opportunity/position

Figure S-54: Market Segment Evaluation Matrices

Z

710

Encyclopedia

Step 10.

Identify the Best Segment-Opportunity matches. a. Document the best opportunities for each segment under analysis. Identify the issues that enhance or limit the opportunity i. The plot of SWOT data provides a map of the best

A

opportunities as they relate to market segments and the size of the segment. Segment-specific weighting enables this process and helps to focus the voice of the technology (VOT) and voice of the business (VOB) requirements and constraints.

B C D E

b. Identify the key issues for strategic improvement actions for improving the strengths and eliminating weaknesses for future enabling of growth.

F G H

c. Create tactics and identify local improvements that will help define how the opportunity can be exploited with the least risk relative to the size of the financial opportunity.

I J K L M N O P Q R S T

Supporting or Linked Tools Supporting tools that might provide input when developing a SWOT analysis include • Brainstorming (See Also “Brainstorming Technique,” p. 168) • Market Perceived Quality Profile (See Also “Market Perceived Qual-

ity Profile (MPQP),” p. 390) • Performance charts

U

• Porter’s 5 Forces (See Also “Porter’s 5 Forces,” p. 464)

V

• Real Win Worth (See Also “Real-Win-Worth (RWW) Analysis,”

W X Y Z

p. 560) • VOC/VOB gathering techniques (See Also “Voice of Customer

Gathering Techniques,” p. 737) • Won/Loss Analysis

A completed SWOT analysis provides input to tools such as • Brainstorming (See Also “Brainstorming Technique,” p. 168) • GOSPA (See Also “GOSPA (Goals, Objectives, Strategies Plans and

Actions),” p. 320)

SWOT (Strengths-Weakne sse s-Opportunitie s-Threats)

711

• Portfolio Strategic Plan • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),”

p. 287) Figure S-55 illustrates the link between the SWOT analysis and its related tools and techniques.

B

Market Perceived Quality Profile

Performance Charts

C D Brainstorming

E GOSPA

F G

Porter’s 5 Forces

H SWOT Analysis

Portfolio Strategic Plan

I J

Real Win Worth FMEA VOC / VOB

A

K L M

Won / Loss Analysis

Figure S-55: SWOT Tool Linkage

N O P Q R S T U V W X Y Z

712

Encyclopedia

T Tree Diagram—7M Tool A B C D E F G H I

What Question(s) Does the Tool or Technique Answer? What are the details behind a general topic; how does it breakdown into its component-parts? A Tree diagram helps you to • Display a topic in its hierarchical components • Organize a topic into a logical breakdown • Probe into symptoms to reveal root causes

J K L M

Alternative Names and Variations This tool is also known as

N

• Hierarchical diagram

O

• Systematic diagram

P

• Tree analysis, analytical Tree

Q R S T

Variations on the tool include • Work Breakdown Structure (WBS) (See Also “Work Breakdown

Structure (WBS),” p. 753)

U V W X Y

When Best to Use the Tool or Technique When there’s a need to parse a topic into supporting details or component parts, use a Tree diagram to systematically analyze and illustrate the hierarchical relationship.

Z

Brief Description The Tree diagram is a flexible tool that displays the hierarchical structure of a high-level topic broken down into its subcomponent details or related topics. It is a generic tool that illustrates the logical, sequential flow from a general topic down its multiple, related branches or paths. A completed diagram resembles a multi-branched Tree, stemming either

Tree Diagram—7M Tool

713

from the top and branching down or from the side and branching horizontally (conventionally from left-to-right.) Using the Tree-structure systematically dissects a topic into greater detail and communicates an often complex and multi-faceted hierarchy with a single picture. Its flexibility allows it to apply to several different scenarios. For example, a Tree diagram helps to analyze a process of flow into its logical paths and map out the next steps from a specific starting point. The tool also drills down into symptoms to reveal the root cause(s), as in a Cause-and-Effect diagram or a 5-Whys Tree—or analyzes a general customer requirement into specific critical-to-quality (CTQ) metrics, as in a CTQ Tree. It can display a systematic flow of an action plan, starting with the requirements and logically stepping through its related deliverablestasks-tool linkage, similar to a Work Breakdown Structure (WBS). It can organize a response plan to a potential failure, as in a Cause and Prevention Diagram or similar to a Fault Tree analysis. (See Also “Cause-andEffect Diagram—7QC Tool,” p. 173; “Cause and Prevention Diagram,” p. 198; “Critical-To-Quality (CTQ) Matrix,” p. 242; “Fault Tree Analysis (FTA),” p. 309; and “Work Breakdown Structure (WBS),” p. 753) Tree diagrams are a member of the 7M Tools, attributed in part to Dr. Shewhart, as seven “management” tools, or sometimes referred to as the 7MP or seven management and planning tools. These 7M Tools make up the set of traditional quality tools used to analyze quantitative data. The 7M Toolset includes: Activity network diagrams (AND) or Arrow diagrams; Affinity diagrams; Interrelationship digraphs or Relations diagrams; Matrix diagrams; Prioritization Matrices, often replacing the more complex Matrix data analysis; Process decision program charts (PDPC); and Tree diagrams. The Quality Toolbox, by Nancy Tague, presents the 7M Tools ranked from those used for abstract analysis to detailed planning: Affinity diagram, Relations diagram, Tree diagram, Matrix diagram, Matrix data analysis (commonly replaced by a more simple Prioritization Matrix), Arrow diagram, and Process Decision Program Chart (PDPC).

A B C D E F G H I J K L M N O P Q R S T U V

How to Use the Tool or Technique A Tree diagram can be built by an individual or with a team. The procedure to build a Tree diagram, as illustrated in Figure T-1, is as follows:

W

Step 1.

Y

Step 2.

Document the topic of interest at the top of the page (presuming a vertically displayed Tree). If diagramming the Tree horizontally, start to the left and work toward the right. Identify the subcomponents of that topic and align them by similar hierarchical levels and within natural affinity groupings.

X Z

714

Encyclopedia

a. Work either down one path at a time or across one hierarchical level at a time. Exhaust ideas and then proceed to either the next branch of the Tree or the next level. b. Use brainstorming techniques, documentation, illustrations, and Process maps to assist with identifying the different detailed topics. (See Also “Brainstorming Technique,” p. 168)

A B

c. Ask probing questions to help identify subcomponents, such as the 5-Whys technique if exploring potential root causes. (See Also “5-Whys,” p. 305)

C D E

d. Continue until the ideas are exhausted.

F G H

Figure T-1 illustrates a simple horizontally-displayed Tree diagram of what is critical-to-quality for car fuel mileage.

I

Car Fuel Mileage

J K L

Aerodynamics

Horsepower

Weight

M N

Body Profile

Body

O P Q

Drag Coefficient

Engine

# of Cylinders

Horsepower Rating

R Antenna

S T U V

Spoiler

Figure T-1: Car Fuel Mileage CTQ Tree Diagram Example

W X Y

Hints and Tips

Z

Complex topics mapped into a Tree diagram often include a number system (similar to that used in an outline format) to identify the different levels and cells within the structure.

TRIZ

715

Language should reflect the purpose of the Tree diagram such that if it is a general topic worked as a noun, describe the subcomponents as nouns or noun-verb (in past-tense). If the Tree diagram represents an action plan, define the detail using verb-noun. Diagram the Tree down to as much detail as necessary, depending on the topic or purpose of the analysis; however, as a general rule when considering the diagram as a communication vehicle, the simpler the better.

Variations Work Breakdown Structure (WBS) (See Also “Work Breakdown Structure,” p. 753)

A B C D E F G H I

TRIZ

J K

What Question(s) Does the Tool or Technique Answer? What is the best solution to address this problem and create a competitive advantage? TRIZ helps you to

L M N O P

• Invent ideas that solve technical or business issues

Q

• Analyze concepts to bolster their design, such as patents, from com-

R

petitive alternatives

S T

When Best to Use the Tool or Technique Use the TRIZ method to rapidly generate solution concepts to problems, either technical or business in nature.

U V W X

Brief Description The TRIZ method (pronounced “TREEZ”) is a Russian acronym representing a theory of inventive problem-solving. It was developed by Genrich S. Altshuller, a Russian scientist and engineer, in the 1940s to develop new “inventive” ideas by considering possible solutions from other disciplines and minimizing trial-and-error. TRIZ is a structured approach (algorithm) used to rapidly generate plausible concepts that solve technical problems. Altshuller believed that inventiveness can be taught and that creativity can be learned.

Y Z

716

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

Altshuller claimed that traditional inventing was by trial and error, which produced much wasted time, effort, and resources. He developed a theory that one solves problems through a set of approaches that have certain patterns. He found that people apply common sense, logic, and some physics to problem-solving. After studying and analyzing over 1.5 million patents from around the world, he discovered three groups of methods to solve technical problems: 1) using various tricks (a reference to a technique), 2) methods based on utilizing physical effects and phenomena to change the state of the physical properties, and 3) complex methods (combining the tricks technique and physics). Informed by this pattern discovery, Altshuller developed this rigorous inventive problem-solving process. While serving in the Soviet Navy as a patent expert, he found that 95% of the knowledge to solve most technical problems already existed elsewhere. He discovered that two problems from different technologies often had the same model and similar solutions. Altshuller was able to condense technical problems into a generic set of standards. He categorized the patents and respective solutions into five levels that depict the patterns. These patterns were governed by objective laws to develop a system along its path of technical evolution, which are known as the Laws of Technical Systems Evolution. These five levels of determining and implementing innovation include descriptors and the percent contribution of the given category to the total solutions: • Level one—Routine designs of well known methods, with few vari-

ants, wherein no invention is needed (about 32% of the total). • Level two—Minor improvements or conflict partially eliminated by

known industry methods usually involving some compromise (about 45% of the total). The means of solution are found from within the same field of technology as the system. • Level three—Fundamental improvements, or complete conflict

elimination, drawing primarily upon knowledge from another industry, wherein contradictions resolved the problem (about 18% of the total). • Level four—A new principle introduced as a primary function, usu-

ally from science rather than technology (about 4% of the total). • Level five—A rare scientific discovery or invention fundamentally

creating a new system (about 1% of the total).

TRIZ

717

The TRIZ philosophy builds on five core elements: • Ideality—A system evolving to increasing good and decreasing bad. • Resources—Maximizing the effectiveness of things inside and

around system. • Space and Time—Viewing a system in the context of space and

time. • Functionality—The importance of function when thinking about

systems; wherein solutions change, but functionality remains constant. • Contradictions—All systems contain contradictions; look to elimi-

nate them as a primary evolution driver.

40 Inventive Principles This inventive problem-solving process initially involved 27 TRIZ tools, which was expanded to 40 innovated tools in the late 1990s. The TRIZ 40 inventive principles solve contradictions (problems) without compromising quality. These principles, originally intended for technological innovation, have been generalized for practical application in business, education, and everyday life. While the core principle language (the principle name) remains constant, depending on the application, the principles are adapted to a problem’s context and intent of the solution. For example, the 40 Inventive Principles for marketing, sales and advertising are adapted as follows:

A B C D E F G H I J K L M N O P Q R

• Segmentation—Stratify or divide into groups or independent parts

or sections such as market segmentation by customer profile, demographics, geography, and buying habits. • Extraction or taking out—Separate two things, extract or remove

something such as differentiate competitive products, outsourcing, and even separating data (that is, qualitative from quantitative). • Local quality—To alter an object or system from uniform (homoge-

neous) to non-uniform (heterogeneous), to make something better suited for its operation, or to fulfill a different and useful function. For example, personalized advertising or customized cuisine offerings to include regional recipes. • Asymmetry—Change the shape or function of an object or system

from symmetrical to asymmetrical or increase its degree of asymmetry. For example, offer right- and left-oriented products or move to offering male- and female-oriented services.

S T U V W X Y Z

718

Encyclopedia • Merging or combining—Bring closer together (or merge) identical

or similar objects or make them contiguous or parallel (in operations), such as business partnership/alliance, cross-selling, and jointcoupons (that is, milk and cereal) • Universality—Make an object or system perform multiple functions A B C D

and standardize on them to eliminate the need for other parts. For example, new standardized forms to collect customer profile data, or cell-phones with built-in camera or email capabilities. • Nesting (that is, nested doll)—Putting a product (or system) inside

F

of another or allow it to pass through another, such as a grocery stores selling greeting cards or liquor or a coffee shop selling music CDs.

G

• Counterweight or anti- weight—To compensate for weight (down-

H

ward tendency) to provide uplift, such as advertising or endorsements for a movie about a difficult topic.

E

I J K L M N O P Q R S T U V W X Y Z

• Prior counteraction or preliminary anti-action—To perform a

counter-action in advance; to control harmful effects in the future. Take action to create beforehand stressors that will counter or oppose them. For example, gathering competitive information in anticipation of them releasing a new product, settling a lawsuit out of court, or using patents or licenses to protect intellectual capital, Poka-yoke (mistake-proofing). (See Also “Poka-Yoke,” p. 462) • Preliminary action or prior action—Prearranged action or change

of an object or system before it is needed, such as preliminary market research of an offering concept (before it is designed), pre-paying for something, or an introduction to a document in anticipation of questions. Or conversely, Afterward action—The opposite (or inverted) of preliminary action, such as post-paying (payment in arrears), after-sales services, Q&As after reading a document, or a coupon for purchasing an item for the next purchase. • Beforehand cushioning or cushion in advance—Prepare emergency

means prior to an event or compensate for the relatively low reliability of an object by countermeasures taken in advance. For example, utilize Poka-yoke (mistake proofing) strategy, excess inventory, contingency planning or clauses in a contract and split commissions paid upon order-taking and customer acceptance of an offering. (See Also “Poka-Yoke,” p. 462) • Equi-potentiality—Change the working conditions so that an object

need not be raised or lowered, or in a potential field, limit position changes to make something balanced or equal. For example, leveling

TRIZ

719

of a relationship by translating language or terminology, seamless transition from a trial period to permanent ownership, and toll-free customer service phone numbers. Or conversely, Potentiality gap—Such as building up market entrance barriers against competition. • Inversion or “the other way around’”—Invert the action used to

solve an issue or do the opposite or turn something upside down or view from the opposite perspective. For example, instead of speeding up, slow down; push versus pull; and a rebate for the return of a competitive product. • Spheroidality or the use of curves—Use spherically-shaped objects,

curvature, or using a rotary motion or centrifugal forces. For example, use rollers or balls, round prices to the nearest dollar, implement revolving credit, or make rolling forecasts of sales volumes. • Dynamics—Design or allow an object or system to be flexible and

adaptive and change to achieve optimal operations. For example, implement seasonal pricing, flexible pricing to account for quantity purchases, moving billboard, and a library van to bring books to the reader/neighborhood, or ice cream truck. • Partial or excessive actions—If something is difficult to achieve, use

slightly more or less of that same method to try to make it easier or simplified. For example, pricing at $1.99, rather than $2, compromising in a conflict resolution, and under-promising but over-delivering on a project completion date or customer service. Or conversely, All or nothing—”We pay for your meal if we forget to give you a receipt,” or “We pay for your personal-pan pizza if it is delivered to your table late.” • Another dimension or moving to a new dimension—Take some-

thing to another plane, re-orient something, use another side of something. For example, use multiple market research sources (that is, passive, primary, and secondary), or customer and non-customer feedback. Use multiple sales distribution channels, use a multidimensional tool or analysis (that is, QFD or multiple regression analysis). (See Also “Regression Analysis,” p. 571 and “Quality Function Deployment (QFD),” p. 543) • Mechanical vibration—Cause an object to oscillate or vibrate;

increase its frequency. For example, video gaming controllers that shake when a shooting sequence happens. Also frequently communicating with customers using multiple modes such as phone, surface mail, and email.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

720

Encyclopedia • Periodic action—Use infrequent or pulsating rather than continuous

actions. For example, instead of calling on a customer on a regular basis, stagger or call infrequently, or air a television commercial with just music or silence versus one with talking. • Continuity of useful action—Carry out work continuously with all A B C D E

portions of the system fully operating or eliminate any idleness or intermittent activity. For example, leverage long-term business relationships or alliances, focus on customer retention or loyalty, and continue successful traditions. • Rushing through or skipping—Perform harmful or hazardous

F

operations at a very high speed, such as getting through a liquidation sale as quickly as possible.

G

Or conversely, Lagging—Delay in action.

H I J K L M N O P Q R S T U V W X Y Z

• Convert harm into a benefit, or “blessing in disguise”—Utilize

harmful factors or environmental effects to obtain a positive effect or until it ceases to be harmful. For example, using customer complaints as opportunities for improvement or minimizing the fear of switching (IT) systems by introducing fear of competition. Or conversely, Cursing in disguise—Lack of customer complaints indicating lack of caring or candor. • Feedback—Introduce cross-checking or refer back to improve a

process, or if it already exits, reverse it (or change its magnitude or influence). For example, use Voice of the Customer input and customer complaints to help design a solution or including engineers with marketing to gather customer input. Or conversely, Feed-forward—Marketing forecasting to anticipate future needs. • Mediator or Intermediary—Use an intermediary object or transfer

or carry out an action; a temporary connection. For example, utilize a two-tiered sales distribution model using wholesalers or retailers or sales agents or using a third-party logistics company (that is, UPS or FedEx) or a consultant. Or conversely, Intermediary removal—Conduct an interview without a professional interviewer or conduct online Internet sales without a sales force. • Self-service—Make an object service itself or make use of wasted

materials or energy. For example, conduct self-benchmarking or encouraging customers to use word-of-mouth about your services/products.

TRIZ

721

• Copying—Use a simple and inexpensive replica instead of the origi-

nal, which could be complex, expensive, fragile, unavailable, or inconvenient to operate. For example, utilize video-conferencing to replace a face-to-face meeting, franchise to replace full-fledged ownership. Or conversely, Anti-copying—Avoid negative associations in advertising or political campaigning. • Inexpensive short-lived objects—Replacing expensive, durable

A B C

objects by collecting inexpensive ones. For example, using inferential statistics or sampling techniques, using temporary staff or student interns rather than full-time employees, and selling second-hand goods cheaper than new items.

D

• Mechanics substitution—Replace a system by an optical, acoustical,

G

or olfactory (odor) system; use an electrical, magnetic or electromagnetic field for interaction with an object, such as electronic communication, electronic banking, e-cards, faxing, or scanning.

H

• Pneumatics and hydraulics—Replace solid parts of an object with a

liquid or gas, utilizing air or water for inflation, or use air or hydrostatic cushions. For marketing or business, it is a bit of a stretch, but an example could be introducing “breathing spaces” into contracts or sampling expansion during a survey. • Flexible shells and thin films—Replace traditional constructions or

isolate an object, using flexible membranes or thin film. For marketing or business, it is a bit of a stretch, but an example could be walking a fine line during a tough negotiation. • Porous materials—Make an object porous or add porous elements

(that is, inserts, covers); or if it is already porous, fill the pores in advance with something. For marketing or business, it is a bit of a stretch, but an example could be empowering the sales people to negotiate pricing or empowering customer-facing support people to resolve customer issues. • Color changes—Change the color of something or its surroundings

to improve visibility (or mask flaws). For example, change the company colors and logo to indicate a new brand image or implement different colors (as color coding) to assist with mistake-proofing. (See Also “Poka-Yoke,” p. 462) • Homogeneity—Make objects that interact with a primary object out

of the same material or characteristics, such as a family of products. • Discarding and recovering—After something has completed its

function or become useless, reject, discard, dissolve, or evaporate it, such as a manufacturer’s warranty. Or immediately restore, modify, or regenerate something if exhausted or depleted.

E F

I J K L M N O P Q R S T U V W X Y Z

722

Encyclopedia • Parameter changes—Transform the physical and chemical states of

an object—its density, flexibility, or temperature (that is, liquid, gas, solid), such as virtual shopping or virtual banking. • Phase transitions—Implement an effect developed during the phase A

transition, for instance, during the change of volume or lifecycle stages of an offering.

B

• Thermal expansion—Use a material that expands or contracts with

C D E F G H I J K L M N O P Q R S T

heat. For marketing or business, it is a bit of a stretch, but expanding or contracting marketing efforts depending on the success of an offering’s revenues or profitability. • Use of strong (boosted) interactions—Replace common air with

oxygen-enriched air or atmosphere. For marketing or business, it is a bit of a stretch, but hiring very creative marketing professionals or using an industry award to bolster a marketing (advertising) campaign. • Inert atmosphere—Replace the Norman environment with an inert

one; carry out the process in a vacuum by adding neutral parts or inert additives to a system. For example, utilizing an anonymous survey or interview, neutral, indifferent line of questioning, or using a neutral third party during a difficult negotiation. • Composite materials—Replace or change a homogeneous material

with a composite or multiple structures. For example, utilize a crossfunctional representation on a process improvement team or use a “good cop/bad cop” combination on a negotiating team.

TRIZ Toolset The TRIZ toolset shares many tools with the Six Sigma discipline. Some of the common tools and techniques include

U

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

V

• 5-Whys technique (See Also “5-Whys Technique,” p. 305)

W X Y Z

• FMEA (See Also “Failure Modes and Effects Analysis (FMEA),”

p. 287) • Pugh Concept Selection (See Also “Pugh Concept Evaluation,”

p. 534) • QFD (See Also “Quality Function Deployment (QFD),” p. 543)

TRIZ also introduces a whole host of unique tools. In addition to the 40 inventive principles, some of the flagship TRIZ tools and techniques include: Contradiction matrix, Conflicts (both technical and physical),

TRIZ

Ideal Final Result (IFR), Trends of Evolution, Size Time Cost Operators, Anticipatory Failure Determination, Knowledge/Effects, Resources, Trimming/Functional Analysis, Measurement Problems, Subversion Analysis, Omega Life Views, Optimization methods, Re-focus/Re-frame, Paradigm Paralysis, ARIZ and Psychological Inertia (PI) tools. Note that this is a partial list of tools and techniques. The TRIZ Ideal Final Result (IFR) is a powerful tool. It leads to a highlevel solution that establishes the boundaries of a system at its extreme result of ideality. Typically it is the place one starts identifying the solution concepts. It focuses on both customer needs and functional requirements. The benefits have been fully delivered, costs have been reduced to zero, and harmful effects have been eliminated. It is independent of the currently used equipment or processes. The use of the IFR technique encourages out-of-the-box thinking to help the innovation team reach breakthrough solutions. For example, consider cutting the grass. The system would be a lawn mower, the product is grass, and the purpose is to control the height of the grass so that the lawn looks nice. The IFR for this system might be that the grass grows to the desired height and stops growing; no lawn mower is needed!

723

A B C D E F G H I J K L

The Principle of Ideality states that systems evolve in the direction of increased ideality. It is the sum of all the benefits divided by the total sum of its costs and harmful effects. Evolution increases ideality by increasing benefits and decreasing both costs and harmful effects or eliminating harmful effects all together. Real solution concepts move the system in the direction of ideality.

M

The TRIZ Functional Analysis and Trimming is a technique that first identifies the important tangible components of a system and how they interact with each other and the environment. Second, the trimming identifies candidates and removes a component from the system that presents problems. The overall technique assists in determining the “right” problem to work on, which typically is one that also moves the system in the direction of ideality.

R

The Zone of Conflict helps to distinguish what kind of conflict and identifies where and when a conflict arises. Several questions help to identify the conflict: Who has the problem? What is the problem? When does the problem occur? Where does the problem occur? Why does the problem occur? How does the problem occur? The 5-Whys technique works as well. Note that there are two types of conflicts—technical and physical. Technical contradictions use Inventive Principles, Standard Features, and Contradiction Matrices; while physical contradictions use Separation Principles. (See Also “5-Whys Technique,” p. 305)

N O P Q S T U V W X Y Z

724

Encyclopedia

The ARIZ technique stands for an Algorithm for Inventive Problem Solving. It is a structured technique that evolves a complex problem into a simple problem to solve. It is a complicated multi-step process of questions that integrates different elements of TRIZ and reformulates the problem to get a fresh look at it. A B C

Typically an improvement involves at least one of three approaches: • Elimination of a harmful effect

D

• Modifying a useful effect

E

• Increasing ideality

F G H I J K L

The TRIZ standard solutions are used to eliminate or modify an effect. Altshuller identified 76 standard solutions. Trends of Evolution can bring a system to the next evolution step and help accelerate the move to ideality. The 76 standard solutions include • Improving the system with no or little change (13) • Improving the system by changing the system (23)

M

• System transitions (6)

N

• Detection and measurement (17)

O

• Strategies for simplification and improvement (17)

P Q R S T U V W X Y Z

Conclusion It takes time to appreciate and utilize all of the TRIZ concepts. The TRIZ toolset is plentiful, with many variations available for different circumstances. TRIZ is a flexible technique that can used by everyone—the technical community, business, marketing, sales, education, and in everyday living. Common tools applied to a large variety of scenarios include the Contradiction Matrix or Trends of Evolution. Altshuller developed tools that could be used to break down technical systems and describe their underlying characteristics. For each generic problem set, Altshuller identified solution techniques that were used to solve these problems. This information forms the TRIZ toolset that initially was applied to technical innovation but has been expanded to apply to business and education. TRIZ formulates the problem in an abstract (non-technology-specific) manner. It helps to identify the key technical and physical contradictions. It provides a detailed list of inventive principles to ensure that the inventor is able to find the solutions needed to solve a problem even if it is out of his or her field of expertise. TRIZ uses known general trends in technology development to predict system evolution.

TRIZ

The technique involves typically a small team of two to six people. After clearly defining the problem, 10-30 concepts are generated in a few hours. The concepts include strong candidates from previously run brainstorming sessions plus additional, more novel concepts. It leverages other best practice tools such as Quality Function Deployment (QFD), robust design, design for manufacturing, and the Pugh process. (See Also “Quality Function Deployment,” (QFD), p. 543; “Design for Six Sigma,” in Part I, p. 45, “The Anatomy of Quality Loss in a Product,” in Part III, p. 763, and “Lean for Six Sigma for Fast Track Commercialization,” in Part III, p. 835 for a brief discussion on Taguchi’s robust design; and “Pugh Concept Evaluation,” p. 534) Solution concepts can be identified at any point in the TRIZ process. Be prepared to record any concept solutions that are generated as soon as they get generated, to capture the idea and not loose it. Everyone can use the TRIZ technique and receive some benefit from it. Continue to explore the TRIZ literature or consult with an expert to identify the best translation for your specific application or purpose.

725

A B C D E F G H I J K

How to Use the Tool or Technique At a high-level, the TRIZ process involves the following four steps: Step 1.

L M

Identify and analyze the problem. Identify the system being studied, its environment, resource requirements, primary function, problem effect, and ideal result.

O

a. Identify the important components of a system at the fundamental level.

Q

b. Define the Ideal Final Result (IFR) using the correct TRIZ tool for solution concept generation.

S

i. Write a statement that describes the function the system performs on the product. ii. Consider that this result must preserve the original function; eliminate the system deficiencies; does not make the system more complicated (using free or available resources); and does not introduce new disadvantages. c. Determine how they interact with each other and the environment—functional analysis and trimming. i. A product or process can be described by identifying the components and defining the functions that exist. (Components are tangible objects.)

N P R T U V W X Y Z

726

Encyclopedia

ii. Functions describe the interaction between two components, such as a tool-action-object. In a function, an object is acted on using a tool, and the action typically involves a parameter change for the object. Example: An auto transports a passenger; a chair supports a person; and an oven heats food.

A B C D

d. Find the zones of conflict. Step 2.

E F G H

a. Propose a solution.

I

b. Identify the new problem created by the solution.

J K

Step 3.

L N

b. State the standard conflict or contradiction to eliminate the conflicts; find the contradicting principle that might need to be changed.

O P

c. Identify the principle representing any undesirable secondary effects.

Q R

d. Identify improving and worsening standard features and determine the best tools needed and apply.

S T

e. Identify inventive principles to consider using in the contradiction matrix.

U V X

Select the appropriate TRIZ principles and tool(s). a. Identify the standard solutions; research if the problem or similar problem has been solved before.

M

W

Select the right problem to solve. Restate the problem in terms of contradictions to formulate it. Identify problems that could occur. Understand the interrelationships between fixing one characteristic and its downstream (or upstream) impact. Identify any conflicts or trade-offs, similar to the roof of a QFD. (See Also “Quality Function Deployment (QFD),” p. 534)

Step 4.

Generate solution concepts. a. Look for the analogous solutions and adapt to the solution.

Y Z

Additional Resources or References • www.triz-journal.com • www.aitriz.org • www.aia-consulting.com

Value Stream Analysis

727

V Value Stream Analysis A

What Question(s) Does the Tool or Technique Answer? Where is value being added in the process? Conversely, where does waste exist in the process? A Value Stream analysis helps you to • Identify the value-add and non-value-add activities in a process and

discover improvement and/or redesign opportunities • Analyze where waste occurs in a process and develop a strategy to

improve the process flow by eliminating the waste • Make the processes work for you, instead of you working for your

processes

B C D E F G H I J K L

Alternative Names and Variations Variations on the tool include • Process map (See Also “Process Map (or Flowchart)—7QC Tool,”

p. 522) • Value Stream map

M N O P Q R S

When Best to Use the Tool or Technique When analyzing a current process, use the Value Stream Analysis to identify waste to eliminate any pockets of excellence (where the process flows smoothly and work is done correctly the first time, every time).

T U V W X

Brief Description Value stream analysis is a critical Lean tool that examines a process and identifies improvement opportunities. This tool analyzes which activities contain waste and which truly create value, with the aim being to enhance those value-adds and eliminate the waste. The end-user, paying-customer defines what is or is not of value. This is one of the few times in Six Sigma that internal customers cannot define pure value-add activities, but they can determine what items fall into the non-value-add business category.

Y Z

728

A

Encyclopedia

A value stream is defined as a set of activities required to deliver a product to the customer from raw material input. The total value stream may encompass multiple companies and organizations. It is a big-picture perspective required to avoid selective implementation resulting in isolated islands of Lean within a overall non-lean process. The Value Stream analysis output typically entails all or some of the following calculations:

B

• Total number of steps

C

• Total process cycle time

D E F G H I J K L M N O P Q R S T U V W X Y Z

• Percent and number of value-add steps • Percent value-add time

• Percent and number non-value-add activities (often split into non-value-add business and non-value-add) • Percent non-value-add time

As a rule of thumb, a typical process includes only 0–17% value add steps and 0–5% value-add time because it is defined by what the end-user customer is wiling to buy. Value stream analysis embraces two perspectives when examining a process—the people working in the process and the item that travels through the process as it is transformed to its final commercial-ready state. Figure V-1 illustrates the two-pronged perspective and displays a value stream of 35 total steps from the perspective of the tangible item and only five value-added activities (denoted by the white, bolded squares). The tool examines each process step and groups the activity in one of three categories—Value Add, Non-Value Add Business, or Non-Value Add. Value added work is defined as the activity that physically changes the a product or adds important information. The activity must not be reworked and the customer must be willing to pay for it. Non-Value Add Business (NVAB) represents an activity required by the business but not the customer. These NVAB activities help the business operate efficiently, such as legal requirements, regulatory requirements, recording financials, and maintaining an intellectual capital management system. Examples of NVAB include completing an expense report, filing for a patent, and entering status and lessons learned information into the project database. Non-Value Add is pure waste, also called “muda” in Japanese.

Value Stream Analysis People Perspective

Item Perspective

Open Mail

1 hr.

2 hrs.

4 hrs.

30 min.

2 hrs.

4 hrs.

4 hrs.

10 min.

Process Order

Check Credit

Clarify Order

Enter Order

En ter Data

Verify Data

Receive A pproval

Release to Mfg.

Done

2 days

2 min.

10 min.

75 min.

5 min.

10 min.

5 min.

120 min.

10 min.

Arrive in truck

Wait

Open Mail

Mo ve

Wait

Sort

Check

Number

Wait

Move

90 min.

90 min.

1 min.

30 min.

12 min.

30 min.

2 min.

2 min.

Wait

Send

Wait

Clarify

Mo ve

Wait

So rt

Wait

Code

Move

30 min.

10 min.

1 min.

240 min.

5 min.

15 min.

2 min.

2 min.

1 min.

3 min.

Wait

Verify

Copy

Wait

Mo ve

Check

Sen d

Release

Prin t

Move

60 min.

30 min.

Wait

Match

180 min. 10 min.

729

A B C D

10 min. 1500 min. Group

Wait

2 min. Move to Mfg.

Done

value-added

Figure V-1: Value Stream Perspectives

E F G H I J

Waste Categories Originally the Toyota version of Lean Manufacturing listed seven different categories of waste. Currently, given the western culture, an eighth waste category was added for under-utilized (or non-utilized) people (often classified as a management waste). An acronym to help with recalling the different wastes is DOWNTIME, which stands for: • Defects—Repairs and rework needed to get something to function

properly, which slows the process flow and reduces first-pass throughput yield. • Overproduction—Producing parts ahead of schedule (before a cus-

tomer requests or needs it), while other items in the process wait; working on unneeded parts (the wrong parts at the wrong time); evidenced by producing too much or too early of parts or finished goods. • Waiting—People unnecessarily waiting, stalled in the workflow due

to either shared equipment, unbalanced work activities (operations), decisions, approvals or inspections. • Non-utilized people (a grammar stretch)—Under-utilized

resources, lack of empowerment, or too many workers in the process causing inefficient operations.

K L M N O P Q R S T U V W X Y Z

730

Encyclopedia • Transportation—Moving parts or objects unnecessarily (including

papers or files); excess travel distance, which also can waste time and equipment. • Inventory—Excess space consumed by shelving, floor space, excesA B C D E F G H I J K L M N O P Q R S

sively wide aisles, bins, filing cabinets or files that house accumulated in-process or finished goods, including parts waiting for rework or scrap storage. • Motion—Excess, unnecessary, or non-valued people activities such

as searching, walking, sitting, choosing, copying, stapling, sorting, climbing, bending over, and lying down. • Excess-processing—Unnecessary operations, steps (including

inspection or approvals), and complexity (at times excessive documentation). (See Also “Lean and Lean Six Sigma,” in Part I, p. 29 for more detail.)

Value Stream Analysis Templates The output of a Value Stream analysis can be displayed in multiple formats. Some of the more common templates include a matrix or a bar chart. Summary Value Add Matrix A summary Value Add matrix typically provides a high-level picture as to the number of process steps containing value and non-value activities and a categorization of what type of waste exists. Figure V-2 provides an illustration of a summary matrix that features three parts: • Checkmarks to indicate the categorization (Value Add and the dif-

ferent types of waste)

T

• Total number of checkmarks per category

U

• Percent of the total for each checkmark

V W X Y Z

Detailed Value Add Matrix The detailed Value Add matrix typically contains two parts—one for the current state and the other for the action plan to reach a lean state. Figure V-3 shows both parts as two separate sections for illustrative purposes; ideally, these two sections are part of a single matrix. If this template is created as a spreadsheet (such as Excel), it can easily sum the value-add and non-value-add times and calculate the analysis data.

Value Stream Analysis Process Step

1

2

3

4

5

6

7

8

9

10

Time (Hours)

12

10

1

10

20

6

10

1

10

20

731

Total %Total 100

100%

2

2%

10

10%

Duplication

6

6%

Movement

52

52%

Defects (Customer)

30

30%

Value-Added Non-value-added Inventory (WIP, Backlog) Delay (Operator)

A B C D

Rework (internal)

E

Checking

F G

Lost Opportunities

Total

100

100%

I

Figure V-2: Summary Value Add Matrix

J K

Current State Value Contribution (in Time) Process Step

H

Value Add

Non VA Bus.

Non-VA

Category of NonVA

L

(Type of Waste)

M

1

N

2 3

O

4

P Q R

Action Plan & Leaned Process Improvement Action Taken

NEW Process Step

New Value Contribution (Time) Value Add

Non-VA Bus.

Non-VA

1 2

Comments

S T U V

Tally of Value Stream Analysis

Summary Total Steps Total Cycle Time # Value-Add Steps: % Value-Add Steps:

Current State

Improvement

30 25

12 16

3 10%

4 33%

Value-Add Time: % Value-Add Time: Total NVA Time:

6

7

24%

44%

19

9

Figure V-3: Detailed Value Add Matrix Template

W X Y Z

732

A B C

Encyclopedia

Time-based Value Add Analysis The time-based Value Add analysis uses a bar chart structure to depict a high-level view of the time allocation between Value Add and Non-Value Add activities. This is a good communication tool to focus the team and management on leaning the process to be more efficient. It provides the current state snapshot, the targeted lean process, and the generally unattainable process containing 100% value add activities. Figure V-4 shows an example of a time-base Value Add analysis for receiving a customer order.

D

Receive Customer Order

E F

Log In

G

Request Credit Check

H

Review Status

I J

1 hr. 1 hr.

Determine Discount

2 hrs.

Confirm Pricing

8 hrs.

M

Ideal Process

N

Target Process 0

O P Q

Value-Added Time Nonvalue-Added Time

2 hrs.

Calculate Payment Terms

K L

1 hr.

< 3 hrs. Maximum 4 hours idle - one day turnaround

2

4

6

8

10

12

14

16

18

Time / Hours

Figure V-4: Time-Based Value Add Analysis for a Customer Order

R S T U V W X Y Z

How to Use the Tool or Technique A Value Stream analysis can be conducted by an individual or with a team, using the following procedure. Step 1.

Create or collect a detailed Process map of the process of interest. a. Modify or ensure that the diagram maps both what happens to the item and what the process players do. b. Calculate and record the cycle time (average and range— minimum and maximum) for each step. (See Also “Process Map (or Flowchart)—7QC Tool,” p. 522)

Value Stream Analysis

Step 2.

733

Develop a Value Add template. a. Select which type of matrix or bar chart best suits the analysis (reference Figures V-2, V-3 and V-4). b. Record the process steps in the Value Add document. Consider numbering the steps if the process contains more than ten.

Step 3.

Categorize the activities.

B C

a. Examining each step one at a time, decide if the activity is necessary to meet the end-user customer requirements and produce the final output.

D

b. Classify necessary activities as value add and unnecessary activities as non-value add.

G

c. Document the value add activities on the Value Add matrix or bar chart. d. Analyzing each of the non-value add activities, separate them into those that meet business requirements from those representing pure waste. Document them in the respective columns on the Value Stream matrix. Step 4.

A

Identify the ideal flow. a. Analyze each step that contains waste, prioritize the waste, and develop an action plan to eliminate it. Potential strategies include i. Eliminate loop-backs or rework ii. Reduce batch size

E F H I J K L M N O P Q R S T

iii. Re-sequence steps

U

iv. Standardize work

V

v. Eliminate work in process (WIP) and inventory

W

vi. Balance workloads vii. Segregate work by complexity, re-sequence, and

rebalance viii. Establish visual workplace

X Y Z

734

Encyclopedia

b. Record the action plan on the Value Add document. c. Define the improved, lean process after the implementation of the action plan. Step 5. A

Execute the action plan and apply the appropriate tools to achieve the ideal flow.

B C D

Hints and Tips • Value Stream analysis is a good starting point to work at the

E

“door-to-door” level within your organization. This keeps the

F

project at manageable level of complexity but retains a suffi-

G

cient “big-picture” perspective.

H

• Waste is really a symptom rather than a root cause of the

I

problem. Waste points to problems within the system at both

J

the process and value stream levels.

K

• Working faster is not the answer; working smarter and adding

L

value is.

M

• Work is to be done right the first time, every time; hence,

N

activities such as inspection and rework are non-value adds.

O P Q R S

Supporting or Linked Tools Supporting tools that might provide input when developing a Value Stream analysis include

T

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

U

• Process Map (See Also “Process Map (or Flowchart)—7QC Tool,”

V W X Y Z

p. 522) • Standard Operating Procedure (SOP)

A completed Value Stream analysis provides input to tools such as • Standard Operating Procedure (SOP) • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram—

7QC Tool,” p. 173)

Value Stream Analysis

735

• Control plan (See Also “Matrix Diagrams—7M Tool,” p. 399, for a

brief discussion on control plans.) • FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” p. 287) • Training Plan (See Also “Matrix Diagrams—7M Tool,” p. 399, for a

brief discussion on training plans.)

A

Figure V-5 illustrates the link between a Value Stream analysis and its related tools and techniques.

B C D

SOP

Cause and Effect

E F G

Brainstorming Technique

Control Plan Value Stream Analysis

Process Map

H I

FMEA

J K L

Training Plan

Figure V-5: Value Stream Analysis Tool Linkage

M N O P

Variations Value Stream Mapping (VSM) is a tool used understand and visualize the overall flow of material and information as a product or service makes its way through an end-to-end process. It focuses on the big picture in order to improve the whole system, not just sub-optimizing individual processes (holistic). It identifies sources of waste as opposed to just identifying waste, similar to the Value Stream analysis, but depicts it in a flowchart-like format. VSM provides a common language to describe processes and forms the basis of an implementation plan by helping you design how the whole door-to-door flow should operate. As part of VSM, process areas should assess their effective use of direct labor and overhead. Similar to Process mapping, the VSM technique usually includes diagramming both the current state and improved state. It typically records the metrics including cycle times, down times, in-process inventory, number of workers, working time (less breaks), work-in-process (WIP), and

Q R S T U V W X Y Z

736

Encyclopedia

scrap rate to mention a few. The Value Stream map format displays the process flow a bit differently than a traditional Process map. It tends to include more icons to highlight the potential areas of waste. Figure V-6 provides some of the common VSM icons, and Figure V-7 illustrates a simple Value Stream map. A B C

Finished Goods

D

Manual Information Flow

E F

Operator

Electronic Flow

G

Go See

Kaizen Bursts

I Inventory

H I

Truck Shipment

Source Buffer Stock

Supermarket

Push Arrow

J K

Kanban Signal

L M

Kanban Withdrawal

Kanban Batches Kanban Post

Pull Arrow

N

XOXO

O

Load Leveling

P

Pull Circle

Process Box

Schedule Box

Q R S T U V W X Y Z

Figure V-6: Common Value Stream Map Icons

FIFO

Voice of Customer Gathering Te chnique s

737

Production Control

A Supplier

Customer

B C D E

Sub Assm

Assembly

Warehouse

F G

I

I

I I

Figure V-7: Simple Value Stream Map

H I J K L M

Voice of Customer Gathering Techniques

N O P

What Question(s) Does the Tool or Technique Answer? What is the best way to capture customer requirements?

Q

VOC gathering helps you to

S

• Understand what is important to customers and non-customers

R T U V

When Best to Use the Tool or Technique Before developing a strategy or action plan or before starting any work, first understand what the customer requirements are.

W X Y Z

Brief Description The Voice of the Customer (VOC) gathering techniques represent a suite of different approaches to capture their requirements. It also applies to gathering Voice of the Business (VOB) requirements. This technique aims to collect what is important to both the internal and external customers and

738

Encyclopedia

non-customers, as these requirements guide all the go-forward work. For if your organization fails to satisfy your customers, your competition will.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

One of the first critical questions to ask is, “Who are the specific customers?” Why that seems like an obvious question, it sometimes is a difficult one to answer. If a business engages in a two-tiered distribution model, it may not have direct contact with the end-user of their offerings. Some businesses only keep direct contact with those with issues or compliments, which may represent only about 20% of its user base. Just as important are those potential customers, for it is the prospects that represent growth opportunity. After identifying the target audience, the second phase focuses on what matters to them. Who takes precedence, the VOC or the VOB? The answer is that it depends. A customer request may not align with strategic goals of the organization. Or a customer need may exceed the competency and capabilities of the business. A customer may demand pricing that is not market driven and/or financially feasible. Thus, the VOB provides a reasonableness check on the VOC to keep the guiding requirements within context of the organization’s business model. The VOC and VOB may represent contrasting views of a similar topic. For example, a customer may define quality according to an items’ fitness for use, while the VOB may focus on the conformance to a specification. The end user may describe cost as the purchase price, operating costs, maintenance costs, downtime, and depreciation—the total cost of ownership (TCO). Alternatively, the VOB views cost through production/manufacturing lenses, taking into account the cost of raw materials, labor, overhead and operating costs. Responsibility for service from a customer’s perspective lasts the entire useful life of the product, whereas the VOB may focus primarily on the warranty period. Spare parts are viewed as a necessary evil in a customer’s eyes, while the VOB sees a profitable business opportunity.

Gathering Techniques There are a range of gathering techniques that represent a varying degree of cost, reliability, and detail. Key considerations when selecting the optimal VOC gathering technique includes the available budget for the initiative, the required level of granularity and opportunity for follow-up, and the customers’ profile and demographics. With respect to the budget required to gather input, it needs to cover three dimensions—the costs associated with developing the approach, collecting and analyzing the data. The development cost may require the development of a survey tool or interview guide that could range in level of specificity and complexity. Consider whether the scenario requires customers to experience a prototype of a new offering, or if the required data already existed in a database, development work could be minimal.

Voice of Customer Gathering Te chnique s

Collecting the needed information ties to the level of specificity. Determine if generic input suffices, or if “what if” scenarios best capture the depth of thinking required. Passive data collection provides the lowest level of reliability and may fail to represent the entire cross-section of the customer base. Passive data typically is stored in a variety of functional databases including customer service and support, customer administration and financing, and customer relationship management (CRM)/sales. Other sources of useful data may be buried in other the company records including customer letters, complaints, and customer phone logs. Market research falls into two categories—secondary and primary. Secondary customer research sources data from customer surrogates such as industry experts, partners, competitors, benchmarks, and the customerfacing value chain resources. This gathering technique can be passive or active, using either trade journals, reviews and editorials, or direct conversations with the opinion leaders. The data collected is more reliable than passive techniques but is less than primary customer research. Similar to the game of telephone, which tests the accuracy of a whispered message passed from one person to the next as compared with the original phrase, secondary research operates on similar principles. The more removed from the desired source of information, the less accurate the message will be. Sometimes talking directly with customers presents obstacles (expense and accessibility), but little else can substitute for it. Primary customer research goes directly to the source. This approach covers existing customers, lost customers, and non-customers. Loyal customers kindly take the time to give you feedback, both positive and negative. Retaining current customers is crucial, given that it costs more to win a new customer than it does to keep those active customers. However, lost customers can be a gold-mine for improvement opportunities ideas; the challenges is to secure some of their time. Primary research techniques include face-to-face interviews or focus groups, observation, and surveys conducted in different media. Typical market research spans many objectives, including a planning guide to determine the basic economic trends and how they may impact the offering. It seeks purchase patterns and key influencers that drive change—change in real and disposable income, in consumers’ tastes and values, and in distribution patterns. Market research tries to identify new emerging market segments and other potential opportunities matching current offerings, as well as unfulfilled new, unique, and difficult (NUD) requirements. It also may attempt to solve current offerings’ issues such as sluggish sales volume, weighing the price-to-value ratio relative to competitive offerings. (See Also the “NUD versus ECO,” p. 376, section under “KJ Analysis,” p. 375)

739

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

740

Encyclopedia

Note Think Before Doing—Sampling requires planning and forethought. Take care to envelope a representative mix of the customer-base. A B C D E F

Take time to understand the profile and demographics of a market to identify the various segments. The respondent mix should reflect that of the market and cover the target audience identified in the project charter or SIPOC. (See Also “Sampling,” p. 618; “SIPOC, (SupplierInput-Process-Output-Customer)” p. 663; and “SMART Problem and Goal Statements for a Project Charter,” p. 665)

G H I J K L M N O P Q R

Trade-offs Each gathering technique poses a set of pros and cons. Existing data such as industry studies or government publications (census data, for example) have the advantage of being relatively inexpensive and easy to find. In contrast, their disadvantages may be that they lack sufficient enough specificity or detail to satisfy the research objectives. Surveys (in various forms) are used to collect original data. They have the advantage of being designed to directly address the research objectives. Their disadvantages are not only do they cost more, but also they are more time consuming to design, collect, and analyze data. Evaluate each during the planning process to make the appropriate trade-offs. Figure V-8 provides a quick summary of the major VOC gathering techniques and respective strengths and weaknesses.

S T U

+

Method Increasing data detail & effort required

V W

Z Decreasing data detail & effort required

Considerations

Focus groups

Interaction among participants

Influential participants may sway other’s opinions

Skilled facilitator required

Interview

Direct & immediate feedback

Time consuming

Need to align schedules

Phone survey

Direct feedback

Selective response rate (customer can always hang up)

Bias inherent in data from selective response

Web survey

Ease of use for customer

If voluntary need to limit number of questions

Need to focus number and quantity of questions

Customer service data

Typically easily queried from a database

If voluntary may not capture all customer opinions

Data integrity

X Y

-

Figure V-8: VOC Selection Matrix

Voice of Customer Gathering Te chnique s

741

The active Focus Groups represent the most ergs of effort and gather the detail, while at the other extreme passive gathering technique of mining a customer service database is relatively easy to execute but yields less detail. The budget and time required to develop and collect a VOC/VOB plan tend to enjoy much focus and scrutiny. Often the planning, time, resources, and budget to properly analyze the collected data is often overlooked or rushed in the process. However, this final step is when all the benefits are reaped. Taking the time upfront to plan this phase helps keep the end in mind and may inform how to design an interview guide and develop the overall implementation plan. A rule of thumb is that the time it takes to collect the data, usually requires double the amount of time to analyze the verbatim data.

Interview Guide The Interview Guide should be built to aid in collecting input from either secondary or primary research. A data collection aid for passive sources includes a data collection sheet. An interview guide can take on many forms—a tool to guide someone through a one-on-one or phone interview or a questionnaire deployed electronically or hard copy. (See Also “Data Collection Matrix,” p. 248)

A B C D E F G H I J K L

Arriving at the optimal number of questions entails a mix of both art and science. The tool should comprise enough questions to reveal the answers to a few critical questions, perhaps allowing for multiple ways of posing the question to validate understanding and ensure consistency of related answers. In addition, judicious consideration of the respondents’ time should inform as to the appropriate number of questions.

M

The interview guide is a formal set of questions, statements, and stimulus materials designed to elicit responses that will accomplish the objectives of the research project. The questionnaire responses measure individual attitudes, behavior, and feelings toward topic(s). It provides consistent framework for a collection of responses to support analysis and comparison of quantitative and qualitative data. Specific research objectives and questionnaire design should be driven from the business objectives and the marketing questions that fulfill them.

R

The guide can pose two basic types of questions—open-ended and closed.

Open-ended Questions Open-ended questions accept any type of responses provided by the respondent. It allows them to answer in their own words. Within this category, free responses can be selected. Their advantages include receiving unanticipated responses. The input represents the real views of the respondent. This approach is useful when the list of alternatives under

N O P Q S T U V W X Y Z

742

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

consideration is too long to present each option to the respondents. Their disadvantages include the potential difficulty interpreting what the respondents meant or intended by their answers. It can create interviewer/recorder bias and is less suited for self-administered questionnaires. More weight may be given to more articulate respondents. It can be costly to tabulate, code, and analyze responses. Probing questions are another form of open-ended questions. Their advantages include the fact that this form elicits additional information. They often obtain more complete responses. However, their disadvantages are that they are expensive and can be difficult to interpret what the respondent meant. These questions may create interviewer/recorder bias. They are less suited for self administered questionnaires and may cause more weight to be applied to more articulate and verbose respondents. Three broad categories describe this type of questioning—association, construction, and completion techniques. Association techniques present respondents with a series of words, sounds, images, or other stimuli and are asked to respond to each with the first word that comes to mind. Construction techniques ask respondents to view a stimulus and create a story or draw a picture to explain the stimulus. Completion techniques provide respondents with an incomplete sentence and asks them to complete it in any manner they choose. At times a projective-technique is used, wherein the respondent projects into the future. This technique primarily focuses more on potential solutions, rather than simply requirements gathering. The advantages of this approach again is that it solicits more informative, probative information than closed-ended questions. It provides useful information in exploratory stages of a research process in which insights, ideas, and hypotheses are sought. The disadvantages of projecting is that it requires highly trained interviewers and analysts to interpret the data. Typically it involves a small sample size, which often cannot be projected onto the entire population.

Closed-ended Questions Closed–ended questions limit the responses to those provided to the respondent. The respondent makes a selection among a set of options. Technically, the term “closed-end” characterizes both the questions and the responses for this technique. There are seven main subtypes for this approach, which depend on the type or responses collected. They are • Dichotomous—Bipolar and unipolar. • Ranking—Involving critical, ratio, and interval. • Checklist—Both categorical and hierarchical.

Voice of Customer Gathering Te chnique s

743

• Multiple choice—Respondent chooses one response from a list of

options. • Multiple response—Respondent may give more than one response

from a list of options. • Numerical.

A

• Rating Scales (Likert or Semantic Differential).

B

Dichotomous questions allow for only two possible answers, such as, “Do you own or lease your car?” The most common variety are yes/no responses. Thus the name, dichotomous, meaning opposite pairs of descriptive words or phrases representing the two extremes. They are easy to ask and easy to answer, but they do not provide much information. Use this type of questioning format to collect classification data about the respondent and consider following it with an open-ended question to obtain explanatory comments if necessary. The advantages are that they can be good lead-in to questions that require more detail and that they are quick and easy to administer and answer. There is less chance of interviewer bias, and the responses are easy to analyze. The disadvantages are that they provide no detailed information and they can be difficult to word to obtain appropriate responses. Ranking questions compare within a list of items, typically in order of preference or importance. The order of items relative to one another is the key, not the differences between values. An example is, “Please rank the following reasons from least important (1) to most important (7) to you.” There are two kind of ranking systems—interval and ratio. Interval ranking uses a constant scale that lacks a natural zero value. The differences make sense, but ratios do not (that is, 40–30 and 30–20 decibels, but 40/20 decibels is not twice as loud. This covers topics such as loudness, brightness, saltiness, and temperature. Ratio rankings are ordered on a constant scale with a natural zero, for example height, weight, age, and length. The advantages are that they yield information quickly, and they are easy to ask, tabulate, and analyze. It usually is a familiar activity for the respondent, or it is easy to explain to them. The disadvantages include the number of responses should be limited to five or seven, and it assumes the respondent has information about all items in the question. The question format using checklists asks the respondent to check one (or more) response categories. This approach is useful for factual answers (that is, demographics). Its advantages cover easy to ask and answer, easy to tabulate and analyze, and provides specific answers for the respondents. The disadvantages are that it assumes the researcher knows all relevant alternative responses and often requires a long list that can be boring for the respondent.

C D E F G H I J K L M N O P Q R S T U V W X Y Z

744

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

The multiple choice question format lists a number of answers and requires respondents to select the one answer that best approximates their own. The advantages include the fact that it overcomes many of the problems of open-ended questions, assures respondents will answer on the same dimension, and is less demanding than open-ended to administer. Moreover, this format is less expensive to administer and process than open-ended and is easy to edit, tabulate, and analyze. However, the disadvantages are that the responses must be mutually exclusive and all inclusive, and the order of responses can create response bias. The multiple response format is similar to the multiple choice, except the respondent is allowed to make no choice or more than one choice (instead of just one). Its advantages are that it overcomes limitations of open-ended and multiple choice questions and is useful when research seeks responses; support induction data to theories. It is less demanding and expensive than open-ended to administer and process, and the analysis of structure in response clusters can offer insights. The disadvantages are that the responses must be inclusive of all possible responses, the order of responses can create response bias, and it requires advance multivariate analysis to yield insight into new theories and models. When using rating scales, the respondents are given a range of categories, a continuum, or scale in which to express their opinions. There are a wide variety of scaling techniques. The most common are comparative, non-comparative, uni-polar, and bipolar. Comparative scales involve the direct comparison of stimulus objects. The comparative scale data must be interpreted in relative terms and employ paired comparison, rank order, or constant sum scales. Paired Comparison is a technique wherein the respondents are asked to choose which item rates higher, according to a predetermined criterion. The Rank Ordered technique asks the respondents to rank products according to some predetermined criterion. The Constant Sum technique request the respondents to allocate a predetermined number of rating points among several items, according to some criterion. The Q-Sort technique involves the respondents ranking a group of items into sets according to some criterion, and it discriminates among a large group of items in a relatively short time. The advantages of these comparative techniques include that they are easily understood and applied, only small differences can be detected between stimulus objects, and they use the same known reference points for all respondents. These techniques tend to reduce bias introduced by one judgment on another. Conversely, the disadvantage is their inability to generalize beyond the stimulus objects scaled. Non-comparative scales place each object in its own independent scale from the others in the stimulus set. The respondents evaluate only one object at a time, and the data are generally assumed to be interval or

Voice of Customer Gathering Te chnique s

745

ratio-scaled. The non-comparative scale categories are 1) continuous rating scales, 2) graphic rating scale, and an 3) itemized rating scales, using a finite number of choices with numbers or descriptors for each choice. The Likert rating scale is an ordinal measurement scale. It is named for Rensis Likert, who invented the scale in 1932, using a series of statements followed by five response alternatives reflecting a person’s attitude—typically strongly agree, agree, no opinion, disagree or strongly disagree. Each category encompasses five equally weighted, bi-directional choices that represent both positive and negative options. The positive responses yield a higher score than the negative ones. A similar, but more specific scale is the Thurstone scale, which uses equal-appearing intervals and so provides interval level data and thus more powerful statistical procedures, such as 7, 9, 11, or 13 intervals reflecting unfavourable, neutral, and favourable (respectively) to determine where the respondent’s attitude falls. Another option is the Semantic Differential scale using several criteria each with its own a bipolar scale to capture the respondent’s attitude toward something, usually used to evaluate the strengths and weaknesses of a product or company. The bipolar scale involves opposite adjectives such as excellent-poor, good-bad, complex-simple, and typically have a numeric scale within these two extremes (-3, -2, -1, 0, 1, 2, 3). The Stapel Scale is a uni-polar similar to Semantic Differential Scale, except only one word or phrase is used instead of using two dichotomous descriptive words or phrases as choices, and each choice is numbered. There are three types of scales—ordinal, interval and ratio. The ordinal uses numbers to indicate the response relative to some characteristic. The scale establishes relative magnitude (if a response has more or less of a characteristic), but not how much more or less. In addition to the counting operation allowable for nominal scale data, ordinal scales permit the use of statistics based on centiles, (that is, percentile, quartile, median). The interval scale employs numerically equal distances on the scale represent equal values in the characteristic being measured. It permits comparison of the differences between objects. The location of the zero point is not fixed, and both the zero point and the units of measurement are arbitrary. Any positive linear transformation of the form y = a + bx will preserve the properties of the scale. Statistical techniques that may be used include all of those that can be applied to nominal and ordinal data and in addition the arithmetic mean, standard deviation, and other statistics commonly used in marketing research. The ratio scale possesses all the properties of the nominal, ordinal, and interval scales. It has an absolute zero point. It is meaningful to compute ratios of scale values. Only proportionate transformations of the form y = bx, where b is a positive constant, are allowed. All statistical techniques can be applied to ratio data. (See Also “Statistical Tools,” p. 684)

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

746

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Encyclopedia

The advantages of using a rating scale are that they can measure intensity of feelings toward issues, and the numerical values can be assigned to each point on the continuum, and statistical routines can be run. They are easy and efficient to ask, answer, and tabulate. The disadvantages are that the scale intervals may not provide clear distinction for respondents. The respondents may have difficulty interpreting questions, or may both agree and disagree depending on their interpretation, thereby loosing interest or getting frustrated by the question.

Format Considerations The questionnaire design involves several decisions. Itemized scales can be balanced or unbalanced. In general, the scale should be balanced and bipolar to obtain objective data and minimize compliance bias. Although there is no single, optimal number of scale categories, the rule of thumb is to employ between five and seven (sometimes as many as nine) categories. Using and odd versus even number of categories needs to be decided. If a neutral or indifferent response is possible from at least some of the respondents, an odd number of categories should be used. In situations where the respondents can be expected not to have an opinion, the accuracy of the data may be improved by a non-forced scale and/or the inclusion of a “don’t know/not applicable” response. Place the description of each scale response category near the response categories. The wording of questions is critical. The questionnaire is the interface between researcher and respondent. There are no hard and fast rules in determining question wording, but the guidelines include simple language, familiar vocabulary, the shorter the questions the better, and the more specific and precise the wording the better. For example, What kind of car do you own? versus What brand and model car do you own? Newspapers aim to write to a fourth-grade level of reader to avoid misunderstanding and misinterpretation, so should a questionnaire. If heeded, the respondent can process the intent of the question faster, with less ambiguity, therein creating a less disruptive, more pleasant experience. The strength of using closed-ended questions is that they are easy to complete for the respondents, simple to code and analyze, have reduced interpreter bias, are free of grading bias, and are more valid and reliable responses. Conversely, the development of such a tool should take time and effort to properly design a closed-ended questionnaire. Researchers need to create carefully worded questions and exhaustive lists of possible responses to ensure that the responses are representative. Care should be given that the questions are not leading, that the respondents answer thoughtfully, not mechanically, and that the number of questions are kept to the necessary minimum. The number of questions must be managed to the necessary few, and it takes time to word them succinctly and capture the exact essence of the question’s intent.

Voice of Customer Gathering Te chnique s

747

Note Attribute Measurement System Analysis—Prior to executing the questionnaire, ensure that appropriate rigor is present in conducting a measurement system analysis. Validate that the purpose of each question is clearly defined, but also that the wording and phrasing is understood by third parties (surrogate or sample respondent set) as intended. Moreover, ensure that each question’s response is actionable—have the end in mind. Heed caution if asking questions that probe into uncontrollable items, as they may set improper expectations and clutter the field of actionable responses. Whatever time you can spend with prospective customers (whether blind or direct interaction) is precious, so use the time wisely.(See Also “Measurement System Analysis (MSA),” p. 412)

How to Use the Tool or Technique VOC Data Collection Process The collecting of VOC and VOB requirements generally involves a team of people. The following guidelines cover a six-step process: Step 1.

Identify the target audience and determine the objectives— what information is required?

Step 2.

Develop a data collection plan. a. Consider starting with a passive approach first, collect data from existing information, and then fill the information gaps afterward using either secondary or primary research. b. Identify the various sources of data and best methods to collect it.

A B C D E F G H I J K L M N O P Q R S T U V W

c. Define the profile and demographics of the target audience.

X

d. Document the data gather plan (who, what, when, where, why, how).

Z

e. Develop a budget and timeline that incorporates developing the collection tools (if necessary), gathering, and analyzing the data.

Y

748

Encyclopedia

Step 3.

Develop the data gathering tool(s). a. Develop the appropriate collection tool depending on the approach—passive, secondary, or primary research. b. Refer to the following Interview Guide procedure, if appropriate.

A B

Step 4.

Execute the data collection plan.

C

Step 5.

Analyze the data to generate a key list of customer requirements using their language.

D E

a. Consolidate and organize findings into a useable format.

F

b. Consider using a KJ Analysis technique and an Affinity diagram if appropriate. (See Also “KJ Analysis,” p. 375 and “Affinity Diagram—7M Tool,” p. 436)

G H I

Step 6.

J K

a. Develop the appropriate level of detail, such that the necessary number of CTQs per requirement are developed and delegated to the accountable process players.

L M N

b. Set specifications (or metrics) for CTQs.

O P Q R S T U V W X Y Z

Translate the customer language into the actionable, measurable critical-to-quality (CTQ) items for the organization. (See Also “Critical to Quality (CTQ), p. 242)

Interview Guide Development Process An interview guide is the primary data collection vehicle for primary and some forms of secondary research. It can be built individually or with a team. However, ensure that the appropriate subject matter experts are consulted if they’re not an active part of the development process. The procedure to develop an interview guide is as follows: Step 1.

Build the analysis plan. a. Refine research objectives within the context of the larger VOC Data Gathering Plan Review business and research objectives and determine the specific data needs. b. Brainstorm measures to accomplish objectives, to avoid saying at the end of the research initiative, “If only we had asked…” c. Reduce the list to its essential items. d. Develop a draft analysis plan to answer the question, “What will be done with the data once it is collected?”

Voice of Customer Gathering Te chnique s

Step 2.

749

Determine the primary research process. a. Decide on the best research for the specific business situation. Consider the following options: i. Face-to-face interviews ( Monad or Dyad) ii. Face-to-face Focus Groups iii. Direct or Indirect Observation Studies

iv. Web-based Focus Groups and Bulletin Board Discussions v. Web-based Surveys vi. E-mail Surveys vii. Telephone Interviews viii. Telephone Surveys ix. Mail Surveys

B C D E F G H I J K

x. Delphi Studies

L

xi. Omnibus Polls

M

xii. Lab or Field Study Step 3.

A

Evaluate question content.

N O P

a. Are they aligned with business and research objectives?

Q

b. Evaluate potential stimulus materials and research questions to determine whether or not they will elicit meaningful response data.

R

c. Consider the following criteria:

U

i. Does the respondent understand the question? ii. Does the respondent have necessary information to respond? iii. Will respondent provide the right information? iv. Will stimulus or questions bias the responses?

S T V W X Y Z

750

Encyclopedia

Step 4.

Determine the question format and structure. a. Decide to use open-ended and/or closed questions. b. Determine the structure—format and sequence to facilitate the interviewer or the respondent (if self-administered) through the various topics. Consider the following structure sequence:

A B

i. Introduction/Purpose of Survey

C

ii. Opening/Lead-in Question—Should be easy to answer, not too personal, and may not contribute any useful data.

D E F

iii. Qualifying Questions—To ensure that the respondent qualifies as part of the target audience.

G H

iv. Warm-up Questions—To break the ice and get the respondent comfortable.

I J

v. Specific Questions—To capture the main objective of the research.

K L

vii. Demographics—The data to categorize the respondent into groups or segments, which may be used later when applying stratification to the data to reveal certain patterns. (See Also “Stratification,” p. 697)

M N O P

c. Design the layout. The appearance should be clear and uncluttered. If using open-ended questions, leave enough space for recording answers. If using “skip patterns” for potentially inapplicable questions for some respondents, the instructions should be crystal clear.

Q R S T U

Step 5.

V

a. Pre-test or pilot the questions to ensure that what is intended to be communicated is what is understood by the respondent. Test both the question wording and the analysis technique to ensure the respondent’s answers are actionable.

W X Y Z

Pre-test and revise questions to arrive at the final draft.

Step 6.

Execute.

Voice of Customer Gathering Te chnique s

751

Hints and Tips • Consciously try to use precise, unambiguous language in an interview guide. • When developing an interview guide, consider several cautions in crafting the actual wording: • Avoid double-barreled (two) questions, such as, “Do you consider Bran cereal sweet and tasty?” • Avoid leading questions, for example, “Don’t you agree that…?” or “Wouldn’t you say that…?” • Avoid loaded questions including, “How do you like the performance of the car?” with the option of selecting Extremely well ( ), Very Well ( ), Pretty well ( ), Not too Well ( ). • Avoid words with an emotional tone, such as oil monopolies and luxury items. • Avoid using labels for people (that is, right- or left-wing, conservative, liberal). They often are ambiguous (different meanings for different people). • Avoid questions that assume knowledge or rely on memory. • Avoid asking too many positive or negative questions.

A B C D E F G H I J K L M N O P Q

Supporting or Linked Tools Supporting tools that might provide input when gathering VOC include • Benchmarking (See Also “Benchmarking,” p. 160)

R S T U

• Written reports

V

• Brainstorming (See Also “Brainstorming Technique,” p. 168)

W

A completed VOC data provides input to tools such as • Brainstorming (See Also “Brainstorming Technique,” p. 168) • Affinity diagrams (See Also “Affinity Diagram—7M Tool,” p. 136) • KJ Analysis (See Also “KJ Analysis,” p. 375) • CTQs (See Also “Critical To Quality (CTQ),” p. 242)

X Y Z

752

Encyclopedia • Cause-and-Effect diagram (See Also “Cause-and-Effect Diagram,”

p. 173) • QFD (See Also “Quality Function Deployment (QFD),” p. 543) • Solution Generation (See Also “Brainstorming Technique,” p. 168) A B

Figure V-9 illustrates the link between VOC data and its related tools and techniques.

C D

Affinity Diagrams

E F

KJ Analysis

G H I

VOC Gathering Techniques

J K L

CTQs

Benchmarking

Written Reports

Cause-Effect Diagram

M N

QFD Brainstorming

O

Solution Generation

P Q R S T U V W X Y Z

Figure V-9: VOC Gathering Techniques Tool Linkage

Work Breakdown Structure (WBS)

753

W Work Breakdown Structure (WBS) A

What Question(s) Does the Tool or Technique Answer? What are the project deliverables, and what activities will produce them?

B

A WBS helps you to

D

• Communicate the project deliverables and define the respective

activities • Ensure the deliverables’ subcomponents are identified as necessary

and sufficient for project completion • Provide a baseline by which to measure project performance and

control • Begin to estimate required time for every activity by understanding

the work required

C E F G H I J K L M N

When Best to Use the Tool or Technique At the beginning of the project, during its planning phase, use a Work Breakdown Structure to illustrate the detail project scope. It takes the milestones and deliverables from the project charter and begins to break them down and organize them into their associated activities.

O P Q R S

Brief Description The Work Breakdown Structure is an important project management tool that describes the project deliverables and their descriptions and is structured as a hierarchical Tree diagram. It defines the total scope of a project, its milestones, deliverables, and activities, wherein nouns depict the milestones and deliverables, and verb-noun combination define the activities and/or tasks. The tool further develops the deliverables from the project charter, maps them to the key milestones, and defines the activities and tasks needed to produce those project outputs. Hence, the tool helps to visualize the project deliverable-task linkage, and can be expanded to include assigning of the individual accountable to produce that deliverable. (See Also “Tree Diagram,” p. 712)

T U V W X Y Z

754

A B C D E F G H I J K L M N O P Q R S

Encyclopedia

A WBS can identify a project’s key milestones, the critical reference points in a project that delineate a major event. Milestones are used to monitor the project’s progress. A milestone has zero duration on a project schedule and represents zero effort or work. However, preparing for a milestone can involve significant work, as they serve as a checkpoint to help to manage and control a project. For example, if a project improvement team were following the Lean Six Sigma DMAIC methodology (Define-Measure-Analyze-Improve-Control), each of those five steps or phase-gates represent a milestone. The basic building blocks of a WBS are referred to as work packages. These work packages may be sub-projects and contain one or more tasks. They list the most important tasks in the project, as tasks require time and resources (people, equipment, facilities, materials, and sometimes project funding). A WBS can help to ascertain which tasks depend on the completion of others, and find the relationships between tasks. A WBS also can identify points of completion that can be seen as milestones. The tool also can define the duration for each activity. However, the WBS does not show sequence of these activities and does not show dependencies between activities. (See Also “Activity Network Diagram (AND)—7M Tool,” p. 127, “Critical Path Method (CPM),” p. 242, “Gantt Chart,” p. 317, and “PERT (Program Evaluation and Review Technique) Chart,” p. 453, for a discussion on project activity sequencing and interdependencies.) Drill down into as many levels of detail as necessary, however the convention is a minimum of four levels. For example, each supporting task can be broken down further into either any in-process deliverables produced by a given task indicating any component parts to the higher-level project deliverable or the tool needed to support that activity. If the latter approach is utilized, it documents the project’s tool-task-deliverable combination needed to manage the project performance. Figure W-1 illustrates the basic hierarchical structure of a WBS.

T U

Project Title

V W Milestone 1 (noun)

X

Milestone 2 (noun)

Milestone 3 (noun)

Y Z

Deliverable 1a (noun or noun-past tense verb)

Work Package Level:

Activity / Task (verb-noun)

Activity / Task (verb-noun)

Deliverable 1b (noun or noun-past tense verb)

Activity / Task (verb-noun)

Figure W-1: WBS Basic Hierarchical Structure

Deliverable 2a (noun or noun-past tense verb)

Activity / Task (verb-noun)

Deliverable 3a (noun or noun-past tense verb)

Activity / Task (verb-noun)

Activity / Task (verb-noun)

Deliverable 3b (noun or noun-past tense verb)

Activity / Task (verb-noun)

Work Breakdown Structure (WBS)

755

Complex projects may require that the manageable components of the project be coded for easy referencing. In addition, the WBS can apply to organizational units, aligning the work packages to them, and being referred to as Organizational Breakdown Structures (OBS).

How to Use the Tool or Technique A WBS can be developed individually or with a team. If created individually, the project team should be invited to edit and modify the WBS to ensure accuracy and gain ownership from the key individuals accountable to produce the project deliverables. Develop a WBS by executing the following procedure: Step 1.

Identify the project milestones and deliverables defined in the project charter.

Step 2.

Using a hierarchical Tree diagram structure, record the project title as the tip of the Tree and record the milestones as the first level of the Tree diagram. a. Decide whether to build the WBS diagram from top-down (vertically) or left-to-right (horizontally). b. Draw a connecting line between the milestones and the project title to indicate the relationship.

Step 3.

Align each deliverable with the appropriate milestone and document as the next detail layer in the hierarchy. This represents the second level. a. Draw a connecting line between the deliverables and the milestone to indicate the linkage.

Step 4.

Determine the activities needed to produce each output and document as the third level of branches in the Tree. a. Draw a connecting line between the activities and the deliverables to indicate the relationship.

Step 5.

Determine if any subordinate tasks are required to support the activities and document them as the fourth level in the Tree hierarchy. a. Draw a connecting line between the tasks and the activities to indicate the relationship. b. Note: this fourth level of the Tree represents the work packages.

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

756

Encyclopedia

Step 6.

Optional: Identify the one person accountable for each of the deliverables, activities, and tasks, and record his/her name in each corresponding box on the Tree diagram.

Step 7.

Optional: Ask the one person accountable for a given activity or task to determine the amount of time required to complete that work and record it in the corresponding box on the Tree diagram.

Step 8.

Optional: Ask the one person accountable for a given activity or task to document either any in-process deliverables produced by a given task indicating any component parts to the higherlevel project deliverable or required tool to complete the task.

A B C D E F G H I

Supporting or Linked Tools Supporting tools that might provide input when developing a WBS include

J K L M N

• Project Charter(See Also “SMART Problem nd Goal Statements for a

Project Charter,” p. 665) • SIPOC

A completed WBS provides input to tools such as

O P Q R S

• Activity Network Diagram (See Also “Activity Network Diagram

(AND)—7M Tool,” p. 127) • Critical Path Method (CPM) (See Also “Critical Path Method

(CPM),” p. 242)

T

• Gantt chart (See Also “Gantt Chart,” p. 317)

U

• PERT chart (See Also “PERT (Program Evaluation and Review

V W X Y Z

Tecnique) Chart,” p. 453) • RACI (See Also “RACI (Responsible, Accountable, Consulted,

Informed) Matrix,” p. 554) Figure W-2 illustrates the link between a WBS and its related tools and techniques.

Work Breakdown Structure (WBS)

757

Activity Network Diagram

Project Charter

Critical Path WBS

SIPOC

A Gantt Chart

B C D

PERT Chart

E F

RACI

Figure W-2: WBS Tool Linkage

G H I J

Variations Cause-and-Effect matrix (See Also “Cause and Effect Diagram,” p. 173)

K L M N O P Q R S T U V W X Y Z

758

Encyclopedia

Y Y = f (X) A B C

What Question(s) Does the Tool or Technique Answer? What are the critical parameters in the process?

D

Y=f (X) helps you to

E F

• Identify and summarize the significant inputs

G

• Process elements that drive the output

H I J K L M

When Best to Use the Tool or Technique Hypothesize the components of the Y = f (X) equation early in an improvement project, starting with the desired outputs, and then validate the critical variables that impact the output throughout the remainder of the project.

N O P Q R S T U V W X Y Z

Brief Description The equation, Y = f (X), reads “Y is a function of X,” where Y is the output or final product of your process, and the Xs describe the critical inputs and the process elements that influence the output. In other words, it defines that the output depends on its key process and inputs variables, given that the process transforms the inputs to produce the output. The Y = f (X) concept focuses improvement (or problem-solving) teams on identifying, optimizing and controlling the vital few inputs and process variables to successfully achieve the desired output level. The equation could involve several subordinate outputs, perhaps as leading indicators of the overall “big Y.” For example, if growth were identified as the big Y, the improvement team may examine leading indicators, such as customer loyalty, customer satisfaction, and customer complaints (as little Ys). Each subordinate Y may flow down into its own Y= f (X) relationship, wherein some of the critical variables for one also may affect another little Y, as illustrated in Figure Y-1. Another example is for Days Sales Outstanding, wherein it could be determined as a function of invoice accuracy, payment terms, and the customers’ processes. This equation summarizes and the Process map illustrates the important input-process-output (IPO) model. Hence, Y=f(X) links to the Process map tool, whereby the big Y represents the overall process outputs, the

Y=f(x)

759

little Ys represent the in-process outputs, and the Xs represent the process steps and the overall and in-process inputs.

Ygrowth = Grow Sales Income = f (y1, y 2, … y5) y1 = Customer Loyalty y2 = Customer Satisfaction y3 = Customer Complaints y4 = Number of Sales Calls per month y5 = Number of New Products per year

Subordinate Ys may flow down into their own Y= f (X) relationships:

A

y1= f (x1, x2, x3)

B

y2= f (x1, x4)

C

y3= f (x2)

D

y4= f (x1, x4, x7 , x8) y5= f (x4, x7, x8)

Figure Y-1: Y=f(: X) Example

Early in an improvement project, the team speculates as to the critical variables affecting the output. They begin to develop hypotheses as to the potential causes currently dampening the output—a theory of cause-andeffect. Next the team tries to validate its hypothesis. Depending on the type of data available, validation may occur through statistical means— either Hypothesis testing, regression analysis, and/or Design of Experiment. Regression analysis statistically verifies that this equation adequately models the results. Pilots testing potential improvement solutions and simulations (using historical data to model toe process) also may assist in verifying cause-and-effect, but the results yield less confidence. (See Also “Hypothesis Testing,” p. 335, “Regression Analysis,” p. 571, and “Design of Experiment (DOE),” p. 250)

E F G H I J K L M N O P Q R S T U V W X Y Z

This page intentionally left blank

Pa r t I I I Best Practices Articles The Anatomy of Quality Loss in a Product The Anatomy of Variations in Product Performance Benchmarking—Avoid Arrogance and Lethargy Building Strength via Communities of Practice and Project Management Complex Organizational Change Through Discovery-based Learning Lean Six Sigma for Fast Track Commercialization High Risk-High Reward, Rapid Commercialization: PROCEED WITH CAUTION! Listening to the Customer First-Hand; Engineers Too The Practice of Designing Relationships A Process for Product Development Selecting Project Portfolios using Monte Carlo Simulation and Optimization

761

This page intentionally left blank

The Anatomy of Quality Loss in a Product By Bill Jewett For some time companies have prioritized problems with the quality of their products by using the “cost of poor quality” (COPQ) as the metric. This might be measured as the cost of repair or replacement. The concept of “quality loss” reaches beyond the COPQ to include additional costs suffered by customers, producers, and potentially by communities, when the performance of products deviates from its intended requirements, particularly by large amounts. These broad economic consequences serve to place more emphasis on stabilizing performance. The mathematical relationship, termed the “Quality Loss Function,” provides analytical criteria for decisions about the funding of improvements, either through the additional development of designs or by the tightening of tolerances for manufacturing and service. This article describes the Quality Loss Function and relates its influence to engineering methods that reduce the variability in performance due to stressful conditions beyond the control of a product’s design. The Quality Loss Function places emphasis on increasing the robustness of product designs and of their manufacturing and service processes, beyond just conforming to the tolerances of specifications. When developed for product components and subsystems and optimized at the system level, increased robustness can be a major contributor to increased product reliability and to decreased severity of customer-observable problems. When applied to manufacturing and service processes, increased robustness serves to increase yields and reduce cycle times and costs. During product development and often during production support, engineers and managers have to decide which problems to correct and how much investment in specific corrective actions is justified. Are the common measures of product reliability the most useful tools? For example, do you prioritize problems just by their frequency of occurrence? Engineers view the role of a product to deliver outputs for which there are performance requirements and some tolerance for deviations from those requirements. Analytically, they think of the actions of a product as a collection of engineered functions influenced by factors both within and outside of the control of the design. The amount of variation in the output of a specific function is a measure of the quality of that function. When an output deviates from its requirement by more than a tolerable degree, customers judge it to be a failure. Failures in this sense are noticeable and unacceptable deteriorations. In extreme situations, deviations can cause 763

764

The Anatomy of Quality Loss in a Product

parts to break, materials to become useless, systems to no longer operate, or environments to become contaminated. It is a mission of the design of a product, implemented through its manufacturing and service processes, to achieve its requirements and to minimize its variations due to stressful factors that are beyond anyone’s control. Notice that we have not said, “Comply with the tolerances for its specifications.”

Why Not “In-Spec?” Consider the illustration in Figure 1, the familiar scenario of kicking field goals in American football. Getting the ball through the uprights is one of the measures of quality for the kicking function. Traditionally, its requirement is viewed as a “Go/No Go” criterion. For example, if you meet the specification, you get three points; if not, you get no points and give the ball to the other team. This would be one view of the cost of poor quality. If you are a spectator who is only worried about the score, the “in-spec” view may be acceptable. However, if you 0 pts 3 pts 0 pts are the team’s coach “Out-of-Spec” “In-Spec” “Out-of-Spec” and worried about the predictability of the field goal kicking function, highly variable kicking patterns would indicate an important process for improvement and even a reluctance to depend on field goals for scoring. The concept of a “product” can have a wide range of applications. For example, Figure 1: Football Goal Posts Illustrate “In-Spec” and something static like a parking lot is a product. “Out-of-Spec” Acceptance Criteria Its design has a set of requirements, derived from customers’ needs for a particular set of applications. It incorporates specific architectural elements and technologies such as for road building, drainage, and marking. It is produced by a manufacturing process that itself was designed to replicate the product with a high degree of predictability from one parking lot to another. The performance of a parking lot is subjected to a wide range of stressors, from the many elements of weather to the abrasive effects of traffic and snow plowing. Although customers may have a wide range of tolerance

Why Not “In-Spe c?”

for deterioration, they remain sensitive to small variations in quality. For example, faded markings may be annoying, while worn off markings are not acceptable and may cause an accident. A cracked surface is annoying, while a pothole is a failure in need of repair. Rain water puddles are annoying if small, although they may not be acceptable if deep and along walkways. The costs associated with these examples of poor quality are reasonably easy to imagine, particularly for the drivers of the cars. If warranty repairs are needed to the surface, the installer will also suffer costs. In the extreme, the installer may suffer a loss of future business. This parking lot example illustrates that deterioration in performance is often noticeable to customers and may be a cause of dissatisfaction prior to an actual failure event. Small variations can be leading indicators of customer dissatisfaction and product failures. For example, a cracked road surface is a predictor of a future pothole. Another problem is that the criteria for acceptability may depend on the application. A loss of appearance for a VIP parking lot may deteriorate the corporate image. A drainage puddle may damage a visitor’s shoes. A pothole may damage a car’s steering mechanism if it’s in the traffic pattern, but it would not necessarily be a failure if along the perimeter. Certain customers may have reduced expectations due to more stressful applications or operating conditions, such as heavier vehicles around loading docks or in more severe weather environments.

Product Reliability How do people perceive product reliability? Engineers and managers may view “reliability” simply in terms of the frequency of repairs or of performance exceeding specification limits. Customers may view reliability in terms of the usage interval between service calls for whatever reasons that cause the product not to be usable. More critical customers may be concerned with situations when just barely noticeable deteriorations are evident. The replacement of consumable materials may be excluded from their assessment, unless the downtime and inconvenience are unacceptable. The stressfulness of the applications may influence the judgment of whether or not a product is acceptable and superior to alternatives. In our parking lot example, how long does the road surface last between the needs for pothole repair? For how many seasons does the parking lot remain useful before it has to be ripped up and rebuilt?

Decision Criteria The range of customers and their applications doesn’t make it easy for engineers to have unambiguous requirements and specification limits.

765

766

The Anatomy of Quality Loss in a Product

• Do all customers have the same applications? • Do all customers have the same tolerance for deviations in performance? • Are the consequences of performance deviations the same for all customers? Similarly, decisions about investments in problem correction need useful criteria. • Do you correct those problems that contribute the most to system reliability? • Do you correct those with the most significant consequences for customers? • Do you correct those with the highest repair costs? The reliance on specification limits has the difficulty that the willingness to tolerate a deteriorated condition may not be universal, and the consequences of a failure will vary quite a bit among customers. This emphasizes that there are some fundamental difficulties in judging the priorities of problems. 1. Just counting the frequency of occurrence of problems is insufficient for judging a design or making decisions about its improvement. 2. Describing the consequences of problems in terms of the impact on customers’ use can contribute substantially to understanding the severity of problems, particularly when frequency of occurrence is included. However, the lack of a common metric does not help the comparison of problems. 3. Justifications for the application of resources to correct problems are much easier if the benefits are in the same terms as the investments. 4. The criteria must include the consequences of deteriorated performance both to customers and to the producer, particularly the economic consequences.

What is Quality Loss in a Product? The concept of Quality Loss was introduced in the 1980s as a part of the methods of Quality Engineering developed by Dr. Genichi Taguchi. Quality Loss changed the measurement of the consequences of non-compliance from technical to economic terms. It emphasized the view that

What is Quality Loss in a Product?

variation in a product’s performance is a more important concept to embrace than is a failure. This is the premise of robustness development. Variation is a measure of poor quality, that is, the loss of quality. However, with not all variations being the same and their units of measure depending on what is being measured, how can we compare them and use this information? The premise is that all quality losses should be viewed as unnecessary costs. What are these costs? The consequences of a failure may be the costs to replace parts, make adjustments, replace consumable materials, or clean the system. In some cases the failure may end the useful life of the product, generating costs to replace the product and to scrap the failed one. However, that’s a little too easy and misses an important concern. Customers also suffer losses due to performance variations and during the time they must wait while their product is being repaired. Is there a cost when the loss of quality is seen as a variation in an important quality characteristic but is not yet a failure? Our parking lot story illustrated several examples. Stability in performance may be critical to the applications. In certain cases, no customer-observable variation may be acceptable, as might be the case for the printed image in high-quality advertising or corporate publications. Customers may notice that the variation in one product is greater than that in a second example of the same product, such as with two adjacent printers whose printed output would be merged into a single document. Customers may view an observable performance variation as a leading indicator of a failure, such as you might experience if the cylinders in your car’s engine begin to misfire. While a repair may not yet be justified, the performance variation may make customers uneasy or even change their product usage behavior. As a customer, you value a product by comparing its benefits to its costs of ownership. You suffer economic loss when those benefits are not available because performance does not comply with its requirements. The consequences in these examples are unnecessary wastes of money for customers that contribute to higher costs of ownership. Table 1 gives examples to illustrate the point. Even more costly are examples for which the capabilities of the product do not exist to meet the expectations of the customer. Table 1: Examples of Quality Losses Affecting Customers

a. An unexpected vulnerability to stresses that comes from the routine operation of the product, such as premature tire wear due to extensive high speed driving b. Observable differences among products due to root causes in the manufacturing process

continues

767

768

The Anatomy of Quality Loss in a Product Table 1: Continued

c. Lost production for a company due to the loss of use of equipment, such as when a centralized printer is being repaired d. The need to provide a backup capability to protect against downtime, such as a battery-powered auxiliary for a sump pump in your basement that would be vulnerable to power outages e. The cost to train operators to replace consumables and perform routine maintenance, reducing the downtime in waiting for a service person, often the case with office copiers and printers f. Excessive replenishment of consumable materials, such as unexpectedly high fuel consumption or costs of repairs, as is the case for non-warranty failures to your vehicles

As a producer of a product, the correction of a problem in manufacturing or service causes economic losses that add unnecessary costs to manufacturing and service. Product redesigns to correct problems not only add unnecessary engineering costs, but also distract resources from working on new products. Depending on their nature and implementation, late design changes to mechanical or electrical components, for example, can have severe costs for inventories and tooling and delay market entry. These costs are unnecessary because good designs and product development processes should not have allowed them to occur. Table 2 gives some additional examples. Table 2: Examples of Quality Losses Affecting Producers

a. Loss of customer acceptance due to poor product quality and reliability or higher than expected costs with the resulting loss of revenues and market share b. Excessive costs of repair and maintenance such as for failures under warranty c. Product recalls or field modifications of products when high severity problems are found late d. Product replacements and disposals in extreme cases of product failures e. Cleanup of environmental contamination caused by product malfunctions f. Lost corporate respect in the marketplace due to customers learning by experience to have lower expectations for your products

These costs hurt profit margins, jeopardize operations and product development, and put upward pressure on prices that can reduce the

What Is the Quality Loss Function?

product’s attractiveness to customers. With quality losses being seen as economic losses both to customers and to producers, and in extreme cases to the community, the total quality loss is viewed as a “loss to society.” The examples described earlier made it clear that customers and producers suffer economic losses due to performance variations, not just due to performance failures. Intuitively, as deviations from requirements increase, more and more contributors to Quality Loss become factors. As illustrated in the next section by the Quality Loss Function, the relationship between performance variation and costs is not zero until there is a broken part, and it is not at all linear.

What Is the Quality Loss Function? In the Taguchi methodology, the relationship between performance variation and costs is simplified to a quadratic function in which the costs are proportional to the square of the deviation from the requirement. Reality checks confirm that this is a reasonable approximation. It is characterized by losses that are small when performance deviations are close to the requirement and that increase rapidly as deviations from the requirement increase. In fact, it’s often called the quadratic Quality Loss Function. The mathematics based on this assumption can be useful for engineering and management decisions because a range of performance variations is converted into the universal economic metric—money. There are three basic forms of the Quality Loss Function. If variation from the target value of a performance response can be in either direction, then the loss function is centered on the target value. This scenario is referred to as Nominal is Best. If the performance response can only be greater than the target value, then it’s referred to as Smaller is Better, while the opposite situation would be Larger is Better. In our parking lot example, the variations were one-directional, depending on the functional response chosen. Figure 2 illustrates the three forms of the Quality Loss Function graphically. The mathematics of the Quality Loss Function1 represents those situations easily, given that the function is simply that of a parabola: Quality Loss = k (Y - T)2 where Y = measured value, T = target value, and k = economic proportionality constant. The economic sensitivity is expressed through the constant, k, which would be different for each type of functional response and the economic consequences of deviations from its performance requirements. It characterizes the local slope of the parabola. 1.

Phadke, Madhav S., Quality Engineering Using Robust Design, Englewood Cliffs, N.J., Prentice Hall, 1989, ISBN 0-13-745167-9

769

The Anatomy of Quality Loss in a Product

T

Y

Tex t

Quality Loss ($)

Functional Response

“Smaller is Better” Quality Loss ($)

“Nominal is Best”

Quality Loss ($)

“Larger is Better”

Functional Response

T

Y

Functional Response

T

Y

Requirement Target for Functional Response

Figure 2: Quality Loss Is Illustrated by a Quadratic Function

Figure 3 illustrates that performance deviations that are more costly to repair (quality loss) would have a higher value of the constant, k. Also seen is that more sensitive customers who would call for service at smaller performance deviations would force a higher value of the constant.

“Nominal is Best” Quality Loss Function

T

e

xt

Quality Loss ($)

770

Cost of the consequences of the problem

Functional Response (Y) T-

Response Requirement (T)

T+

Average of customers’ tolerance for deviation from target.

Figure 3: The “Nominal is Best” Quality Loss Function Illustrates the Cost Due to a Deviation from the Requirement

When considering decisions about the level of investment to correct a design problem or to improve a manufacturing process, the total quality loss across the population of customers must be considered. Across that

What Is the Quality Loss Function?

population, the total quality loss would be the sum of the quality losses for individual products over time. This is one way to determine the constant, (k). If for the population the total cost to repair a specific problem is known from the service database, then that service cost could be assumed to be the quality loss at the average customer tolerance. The average deviation from target (∆ = Y - T) that caused the failure could be determined from engineering experiments. By using the average data, the range of customers’ applications and tolerances and the range of product vulnerabilities and repair costs will be accommodated.

Total Quality Loss (QL) = k Σ(∆)2 Solving this equation for (k) will give you the shape of the quadratic loss function for that particular problem and market segment. Customers in different markets may have different tolerances for deviations. For the quality of a printed image, you could expect that the printing of corporate literature would have less tolerance for image quality artifacts than would desktop printing for personal use. You can use the loss function for subsequent decisions, based on the cost to correct the problems and expected benefits. The benefits would be the Quality Loss saved. Let’s try a simple, although contrived, example. One simplification is to assume that customers request service before the performance deviation causes them any other costs. The Quality Loss can then be just the repair cost. Certainly that’s not always the case, as discussed earlier. Consider a product that suffers performance losses that are correctable by onsite service by the producer of the product. Suppose that when a characteristic performance response (Y) deviates from its required value (T) by the average amount (∆) = 10 millimeters, a service call is generated, at an average cost of $500 for each repair. Then, k = $500/10 2 = $5 per millimeter squared. When the response deviates by a larger amount, say 15 millimeters (a 50% increase), that increases the cost to a QL = 5 x 152 = $1,125 (a 225% increase). So you can see that the quadratic form adds more and more costs at larger performance variations.

Decision Criteria The Quality Loss Function can be used for decisions about product costs. Suppose the problem just described causes the repair five times per year, for the average deviation of 10 millimeters. That problem would cost $2,500 per year per product. How much increase in manufacturing cost would you accept to save $2,500 per unit per year in service costs? If you were considering that decision during product development, the frequency and cost of a repair would be forecasts based on factory testing and expected customer use.

771

The Anatomy of Quality Loss in a Product

How else can the Quality Loss Function be used? An example would be a decision about how much money to invest to change the product design as a way to eliminate the problem. Suppose that there are 5,000 products in the market and that they have the same problem occurring an average of five times per year. The total Quality Loss due to the problem in question would then be 5,000 x $2,500 = $12,500,000 per year. This would be a very big problem! Depending on the complexity of your design, the consequences for its manufacturing, the particular implementation plan for the design change, and other factors, your business case could justify an improvement project costing up to $12.5 million with a one year breakeven time. It would be beneficial for you to invest in product improvement as long as the investment is less than the total savings in quality loss. Here are some illustrations that may be helpful. Table 3: Examples of Reduced Costs of Variation and Increased Tolerance for Variation Action

Effect

Change the design to reduce the cost of repair, that is, reduction in the consequences of the variation or actual failure.

Reduces the slope of the Quality Loss Function by reducing the constant, (k).

Change the design to reduce the consequences of the variation, such as compensation or a graceful failure mode. Select market segments with customers that have a greater tolerance for performance deviation.

Result Quality Loss ($)

Current

Improved

QL1 QL2 T-

T+

T

i

772

Figure 4: Reduced Costs of Variation Lowers the Quality Loss

Reduces the Quality Loss ($) slope of the Quality Loss Current Improved Function by increasing the QL deviation, ∆, for which the TTT+ T+ T Quality Loss is suffered. Figure 5: Increased Tolerance for Variation Reduces the Reduces the Occurrence of Quality Loss frequency of occurrence. 2

1

1

In the two examples described in Table 3, investing in both actions would reduce the Quality Loss even further. Other investments can be

2

What Are the Implications for Product Development?

expected to reduce the frequency of occurrence of the Quality Loss, such as a specification for higher quality materials, which would increase manufacturing costs, or a project to improve the design, which would require increased engineering costs. Quality Loss will be alleviated by reducing the consequences of the deviations in performance or its failure. Quality Loss will also be saved by reducing the frequency, or probability, of occurrence—that is, by increasing the product’s reliability. The Quality Loss Function enables the investment decisions to be justified by the expected savings in Quality Loss.

What Are the Implications for Product Development? Recording types of failures and their frequency of occurrence, with priorities based on severity levels, are useful activities for judging the progress of design improvements and for forecasting expected reliability after market entry. However, just counting failures and their frequency is not very useful to engineering. When failures happen, they tell you almost nothing about their root causes or how to improve the design. Product development is more successful and efficient when factors that contribute to or anticipate the failures are measured. Performance variation is the enemy of customers. However, it can be the friend of engineers who aim to improve the product design. Variations in a performance characteristic can be measured and considered as leading indicators of a failure. For example, in an office printer, the stoppage of paper that requires clearance by an operator would be an observable failure. Just counting it would tell you something about the customers’ view of the printer’s reliability but nothing about the root cause of the stoppage or how to reduce its probability of occurrence. However, the variation in the timing of the paper’s arrival at a paper path sensor would not only provide information about the variability of the paper’s movement, but also provide a prediction of the paper’s timing exceeding a control limit and causing the stoppage. This variation would not be observable to a customer but would be important for engineering to measure and control. The absence of performance variation, particularly under stressful usage conditions, would then be a competitive advantage! Certain best practices for product development focus on reducing the variations in performance, relative to the specific requirements. During engineering phases, test plans impose stressful conditions known to be effective at forcing variations. By improving those design parameters that are controllable and most influential, the vulnerability to stresses can be reduced. A design that has less vulnerability under factory-imposed stresses is expected to be less vulnerable to whatever stresses are experienced in customers’ use. These are methods of robustness development.

773

774

The Anatomy of Quality Loss in a Product

Product Robustness The concept of something being “robust” is familiar. It implies adjectives such as tough, rugged, durable, long-lived, or stable, for example. For the purposes of the development of reliable products, “robustness” in a product has a specific definition. It is the ability for a design to achieve its intended function with output responses that are aligned with their requirements (“on-target”) and are stable under stressful conditions (“minimum variability”). Robustness can be measured and normalized. The value proposition advertised by Timex (“Takes a licking and keeps on ticking”) specifically speaks to robustness that we expect in their products. Being less vulnerable to stress, that is being “more robust,” provides a competitive advantage. The development of robustness is achieved initially without depending on tightened tolerances, higher grade materials, or other tactics that drive up the costs of manufacturing or service. If developed experimentally, under stressful laboratory conditions, problems will be found faster and better solutions will be developed. If done earlier, more costly design or tooling changes will be avoided. This is very important to achieving shorter cycle times, for example, “lean” product development. Making the product design more robust is the way to achieve a product that is more reliable. So, focusing on minimizing performance variation is more important for reliability development than is a focus on failure events and their rate of occurrence. The development of robustness may be thought of as over-design in that it may be more costly than an “in-spec” design strategy. How can the benefits of this approach be understood? One answer is by viewing any deviation of a performance response in terms of the costs of the loss of quality. The Quality Loss Function provides an economic measure of the consequences of variation. An investment in robustness development will reduce the Quality Loss. The value of the project can then be judged by comparing the savings in Quality Loss to the investment necessary to achieve that savings. Quality Loss, as a measure of the costs of variations from target performance, adds substantial value to decisions about those investments.

Summary In this article we’ve seen that the losses in product quality are unnecessary costs for both customers and producers. Our focus has been on viewing Quality Losses as being due to performance variations from requirements, not just non-conformance to specification limits. The Quality Loss Function extends beyond the concept of the cost of repair and is more valuable for decisions. It has several advantages for engineering and management:

About the Author

1. It focuses the discussion on economic terms instead of on the frequency of occurrence, on the consequences of deterioration, or on the population statistics. 2. It provides a mathematical relationship that can be useful in decision-making by engineering and management. 3. It broadens the scope of costs that are included, beyond just those of repair or replacement. 4. It focuses attention on variations in performance that are closer to the requirement of the functional response. 5. It changes the objective for the design of products and manufacturing processes from an “in-spec” model to one that controls performance to be “on target, with minimum variability,” particularly under stressful conditions. For product development, a focus on variation is more useful to engineering than is a focus on failures. Improvements to a product design or manufacturing process would be justified as long as the investment is less than the savings in Quality Loss.

Additional Readings 1. Don Clausing, Total Quality Development: A Step-by-Step Guide to World Class Concurrent Engineering, New York: ASME Press, 1994, ISBN 0-7918-0035-0 2. Fowlkes, William Y., Creveling, Clyde M., Engineering Methods for Robust Product Design, Using Taguchi Methods in Technology and Product Development, Reading, Mass.: Addison-Wesley Publishing Company, 1995, ISBN 0-201-63367-1

About the Author Bill Jewett is a consultant to businesses engaged in the development of new technologies and multi-disciplined products. With insights into important paradigms and advancements in practices, he assists improvement teams in upgrading their engineering and management processes, project management practices, cross-functional teamwork, and the governance of development programs. For many years Bill worked for Eastman Kodak Company and Heidelberg Druckmaschinen, his focus on the development of high-volume electrophotographic copiers and printers. Among his division-level

775

776

The Anatomy of Quality Loss in a Product

responsibilities were the management of product development programs and of competency centers for mechanical and systems engineering. At the corporate level, he was one of the authors of the processes for advanced product planning, technology development, product commercialization, and their related governance. For over a decade he taught the processes and coached teams in their adaptation and implementation. As the process steward, he evolved the models to incorporate lessons learned from internal practice and external benchmarking. Currently Bill and an associate are writing a book to integrate strategies and methods for the development of increased robustness and improved reliability in products. They expect their book to be available in the last half of 2007. Bill can be reached by telephone at 585-705-3100 or by email at [email protected].

The Anatomy of Variations in Product Performance By Bill Jewett This article characterizes basic elements that contribute to variations in the performance of products. The functional model is a foundation for analytical and empirical methods that improve product reliability and durability. Also, it’s a useful way to understand the valuable role played by stressful conditions in the development of products. With a focus on performance variability, root causes can be determined, leading to the development of improved robustness, with stressful laboratory conditions enabling problems to be found faster and solutions proven to be superior.

Performance Variations in Product Development An objective of product development is to control performance variations. The concern is not only for deviations from requirements that are beyond the “specification limits,” but also for small variations, those that are leading indicators of failure events. Even “in-spec” variations may be observable to customers and cause concerns for long-term reliability and durability. Large variations in the functioning of components and materials can lead to permanent deterioration or breakage. These variations are the essence of failures and their frequency of occurrence. In whatever way customers observe variations and judge them to be performance failures, the occurrences of variations are the elements of failure rate and usage life. In other words, they are the contributors to reliability and durability. Variation is the enemy of reliability and durability. During product development, however, variations can also be the friend of engineering teams who can do the following: a. Measure variations in output responses during experiments and anticipate failures b. Use experimental stresses to force variations to occur deliberately under laboratory conditions c. Identify the controllable design parameters that reduce the vulnerability to stresses, as measured by reduced variability By forcing variations deliberately, problems can be found more quickly. To the extent that product designs can be improved to reduce the 777

778

The Anatomy of Variations in Product Performance

variations under deliberately stressful laboratory conditions, more robust products can be developed faster and can be superior. To the extent that controllable design parameters can be successful without the tightening of their manufacturing tolerances, or depending on higher quality materials or components, increased manufacturing costs can be avoided. The resulting designs will be less vulnerable to degrading forces they cannot control in real customer use and thereby will have higher reliability and durability. A product that is more “robust” is one whose performance is aligned with its requirements and is less vulnerable to stressful conditions that are beyond control. The development of robustness is an objective of product design, to make the product less vulnerable to stressful conditions in its use by customers and thereby more reliable and more durable. As an analogy to product performance, field goal kicking in American football illustrates variability relative to requirements. The coaching of improvements in kicking demonstrates steps that are analogous to those for the development of 0 pts 3 pts 0 pts robustness in product “Out-of-Spec” “In-Spec” “Out-of-Spec” designs or manufacturing processes. As shown in Figure 1, the goal posts define the tolerances for the kicking game. An “in-spec” kick earns three points. An “out-of-spec” kick earns no points and gives the ball to the opposing team—a cost of poor quality. However, is the concern just about staying within the specification limits? Figure 1: Football Goal Posts Illustrate “In-Spec” and Do we care about vari“Out-of-Spec” Acceptance Criteria ability in the flight of the ball? If you were a spectator and worried only about the team scoring, you might just think with an “in-spec” view of the requirement. However, if you were the coach, you would worry about how predictable the kicking game is, particularly under poor weather or field conditions—or better yet, in the last minutes of a close game, the variability of the kicking function will be of great concern. When a kicker places the ball just inside the goal post on one side and then misses widely to the other side on the next attempt, this wide variation reduces the predictability of future attempts.

Cause-and-Effe ct Analyse s

Improvements to reduce the variability of the kicking process and to place the ball in the middle of the goal posts, regardless of the game conditions or position on the field, are analogous to the development of robustness in product designs. I’ll talk more about this later. First I’d like you to think about factors that cause performance variations.

Cause-and-Effect Analyses For the purposes of robustness development, it is necessary to identify those parameters, either controllable or not controllable, that have the most significant influence over variations in the more critical performance responses. The scientific and engineering understanding of the function is the starting point, enriched by the history of trouble/remedy experience with prototypes or products. With the range of potential root causes identified, analyses and experiments can be planned to define the most influential parameters that can be controlled to optimize their set points. The collective wisdom of knowledgeable people on the development team contributes greatly to the planning of these experiments. Additional experiments may need to be designed to explore the possibility that parameters not yet understood are actually more effective. The development of product robustness depends on this. Record what you know as part of the design documentation. The cause-and-effect diagram is one good way to do this. Figure 2 is an example “cause-and-effect” diagram for our friend, the football place kicker. It illustrates that the mean and variability of field goal kicking depends on both controllable and non-controllable factors. Non-Controllable Factors (Stresses)

Family Problems

Charging Defense

History of Missing

Jumping Defense

Ridicule from Coach

Yelling Defense

Lack of Confidence in Holder

Blinding Snow

Fans Waving Flags

Misaligned Snap from Center

Over Confidence

Cross Wind

Insults from Fans

Wet Ball

Fear of Losing

Lose Turf Muddy Field

Sun Glare Ball Direction Too Wide

Shoe Laces Facing Ball Inability to Block Distractions

Wrong Cleat Length Inaccurate Approach to Ball

Ball Laces Facing Shoe Clumsy Ball Holder Ball Not Held Steady Inaccurate Centering of Ball

Inability to Focus on the Ball Inconsistent Follow-through Head rising too early

Controllable Design Parameters

Figure 2: A Football Example Illustrates the Value of a Cause-and-Effect Diagram to Coaching

779

780

The Anatomy of Variations in Product Performance

With that understandable example, you can develop diagrams for the particular engineering function that you are developing. Be certain to include the adjectives. Figure 3 illustrates the more technical example of fuel economy. With technical systems, as with the football example, there can be many potential root causes and controllable parameters. What else can you contribute to the fuel economy example? Are those factors controllable parameters or not? How would they be controlled? Are they controllable by the operator, by the service mechanic, or by the car designers? Speed too High Impatient Driver Late Leaving Use of Wrong Gear Can’t Hear Engine RPM’s

Low Octane Gas

Inaccurate Air Pressure Gage Many Stop Lights

Clogged Fuel Injectors Fuel Mixture Too Rich

Many Hills

Dirty Fuel Filter

Rough Roads

Dirty Air Filter

Under-inflated Tires Low Gas Mileage

Fast Accelerations

Poorly Trained Mechanics Vehicle Too Heavy

Stop and Go Driving

Heavy Viscosity Oil Improper Tires

Heavy Loads Towed Lack of Driver Awareness

High Idle Setting Obsolete Repair Specifications

Engine Over-sized for the Application Lack of Aerodynamic Vehicle Shape

What other causes can be identified? Which are factors are controllable?

Figure 3: This Cause-and-Effect Diagram for Fuel Economy Illustrates its Value to Engineering

The question to think about for the moment is how to determine which stress factors are most influential over the variations and which controllable parameters are most beneficial to reducing the variability and aligning the performance with its requirement. These are then designed into experiments to determine the magnitude of their influence and to establish the optimal set points for those parameters that are controllable. This is the strategy of the methods of robustness development.

Development of Robustness in Products How can we improve the reliability of field goal kicking? There are three fundamental steps: 1. Select superior design concepts.

Development of Robustne ss in Products

2. Reduce the variability in the distribution of performance. 3. Shift the mean of performance to the required target. The strategy of robustness development has us aggravate the variability by using stressful conditions. Figure 4 uses the football analogy to illustrate the steps in the robustness development process. The cause-anddiagram developed earlier helps us to make the experiments stressful and identify those parameters to adjust in order to improve the design. The imagery shown in the example makes it more intuitive that variability has to be reduced first before the mean of the distribution can be shifted. Ideally those parameters that reduce variability will be independent of those that shift the mean. In real life that may not necessarily be true, but the principle is sound. If the design concept does not have controllable parameters that are effective at achieving the requirement and minimizing the vulnerability to stresses, then it is a flawed concept and needs to be replaced by one that does have these properties. In the football example, replacing the kicker would be analogous to changing to a better design concept.

The initial has too much variability and is not predictable

First, reduce the variability.

Then, adjust the mean.

Figure 4: The Steps to Improve the Kicking Process Are Analogous to Those for Developing Robustness in Products

Before we consider how to make the design more robust, let’s first consider how to make the performance worse, deliberately.

Take Advantage of Stress Factors (“Noises”) What are those stressful factors that are unavoidable and not controlled by the design? In the engineering world there are three basic types of stress:

781

782

The Anatomy of Variations in Product Performance

1. External—These forces originate exterior to the product during its operation, storage, shipping, or handling. Examples of these can include extreme variations in ambient temperature, humidity, pressure, vibration, and shock, as well as user abuse, misbehaviors of off-brand consumable materials, and other factors that may be imposed by the customer. 2. Internal—These are forces internal to the product that cause deterioration with the use of the product. Examples of these can include subsystem interactions, mechanical wear, contaminations such as leaks of liquids or dust, oxidation, thermal hot spots, electromagnetic interference, end-of-life misbehaviors of consumable materials. 3. Unit-to-Unit—These factors are within the processes for manufacturing, storage, shipping, handling, or service of the product that cause variations within the population of products. Examples of these include uncontrolled internal or supply chain processes, assembly or service human errors, loss of calibration in instrumentation, and variations in ambient environments (temperature, humidity and atmospheric pressure). In our example of field goal kicking, what are the analogies to the descriptions just listed? Here are some ideas: External→rain, mud, loose turf, sun glare, wind, charging defense, wet ball Internal→wrong cleat length, inaccurate centering of the ball, clumsy ball holder Unit-to-Unit→lack of confidence, shoe laces facing the ball, changes in field position The categories can get a little vague. However, the real question is which of these factors can you control and at what cost? If the answer is not to control them, then they are stresses. How important they are is a second question. With the stresses identified, which of the controllable parameters are more important for reducing the vulnerability to the stresses? For field goal kicking, are they equipment features that can be improved with a little money? Are they just human movements that can be improved with practice? Are they mental abilities that can block out the many distractions? Those of you who have played this sport will have good ideas. Those who make their living playing the game on Sundays have already demonstrated that they have superior solutions. If you were coaching the place kickers, what would you suggest for them to practice? How would you set up stressful practice sessions? That would be the time for creative experiments on the practice field. I can

Development of Robustne ss in Products

envision rain from a sprinkler, wind from a large fan, mud in a naturally deteriorated field, glare flood lights, and abuse from hecklers. How else would you make life miserable for a kicker trying to improve under nasty conditions?

Select the Best Design Concept The best design concepts aren’t just bright ideas people have during product development. They usually evolve from the development of new technologies and design architectures or from the development of solutions to problems with current products. Given that there may be several concepts available for the solution, there is a need to use a structured selection process to choose the best concept. It is important to have more than one solution concept. Otherwise you’ll be entirely dependent on making that one work, regardless of its inherent capabilities. Selection criteria must then reflect characteristics critical to achieving the requirements and optimizing robustness. Figure 5 illustrates a selection matrix attributed to Stuart Pugh1. It facilitates the unbiased comparison of solution concepts based on criteria chosen by the development teams to be most important.

Selection Criteria Relevant to Robustness Effective at minimizing variability in performance Do not require high costs Effective at achieving the requirement Controllable parameters are Independent Controllable parameters are additive Able to be commercialized with low risk and development costs

A

B

C

D

Reference Concept for Comparisons

Available Alternative Solution Concepts

S

+

–

+

–

S

Conclusions

+

+

S

+

–

+

S

S

S

+

–

–

+ =2 – =3 S =1

+ =0 – =2 S =3

+ =3 – =0 –S = 2

Choose the solution concept equal to or better then the alternatives for all of the selection criteria

Figure 5: The Best Solution Concept Must Be Selected from Available Alternatives

What selection criteria would be most important to achieving robustness? First, the design concept must be available to the team as a solution within the timeframe of the project. Otherwise it’s just a dream. Possibly it’s not yet ready to be commercialized, or its intellectual property is owned by another company. Can you license it or buy it? If not, then it’s 1.

Pugh, Stuart, Total Design, Integrated Methods for Successful Product Engineering, Addison Wesley, New York, 1991, ISBN 0-201-41639-5

783

784

The Anatomy of Variations in Product Performance

not an option. What are other selection criteria that are important to consider? This can get a little technical, but the field goal kicking analogy might help. Criteria would include the ready access to controllable design parameters that do the following: a. Are effective at minimizing the vulnerability to stresses (reduce the statistical variance) b. Do not rely on tight manufacturing tolerances or on high cost components or materials c. Are effective at achieving the requirement of the function (shift the statistical mean) d. Are independent, that is, do not interact with each other e. Are additive, that is, their effects add to each other, rather than one causing an effect opposite to another f. Are able to be commercialized with low risk and development costs g. Are able to achieve robustness that is superior to available alternatives under the same stressful conditions There may be more, but I want you to see that these criteria are deep into the functioning of the concept. Comparing alternative solutions is not a matter of gut feel, but rather one of comparing engineering details and data from analyses and experiments. The concept that is most able to satisfy these criteria and be available within the timeframe of the project would be the winner.

Optimize the Controllable Design Parameters Now what can be controlled? A product is an integrated system of subsystems, components, materials, software and the like, performing functions that provide value to customers. A system-level function can be decomposed into its contributing sub-functions, which themselves can be decomposed down to the level of the basic science and engineering of the technology involved. Figure 6 illustrates the concept of mapping the functions of a product system. With that level of understanding, what then are the requirements for each sub-function and the parameters that control the ability to satisfy those requirements? Figure 7 illustrates the use of this tool for our harassed field goal kicker. You may have more insight into the nuances of this functional decomposition, but I just want to show the thought method in an understandable way. The analysis will help the coach decompose the process and look for ways to improve those elements that contribute the most to improving the trajectory of the ball.

Development of Robustne ss in Products f7

f2 Input Signal

f3

f4

f1

f6

Output Response

f5 f8

Figure 6: Functional Modeling Enables Product Systems To Be Decomposed into Their Basic Elements Call the Signal

Center the Ball

Catch the Ball

Orient the Ball

Move Forward in Alignment

Put on Kicking Shoes

Align Body with the Goal Posts

Place the Ball

Swing Leg in Alignment

Block Out Distractions

Align Foot Angle

Keep Eyes on Ball

Contact the Ball

Ball Trajectory

Keep Head Down

Figure 7: Modeling of the Field Goal Kicking Process Enables the Basic Elements to be Understood, Measured, and Improved

Figure 8 shows the input signal, controllable parameters, stress factors, and output response mapped along with the definition of the basic function. Think of it as the core element of a functional flow diagram but with its forces described in a way that helps us to understand robustness development. As a functional parameter diagram2, somewhat akin to a free body diagram in engineering mechanics, it can be applied to the design of either a product or a process. The design concept that is selected and developed has in its architecture those controllable parameters that are found to be effective at controlling the mean and variability of the output response. If those controllable parameters are not effective, particularly under stressful 2.

Phadke, Madhav S., Quality Engineering Using Robust Design, Englewood Cliffs, N.J., Prentice Hall, 1989, ISBN 0-13-745167-9

785

786

The Anatomy of Variations in Product Performance

conditions, then the architecture of the design concept is flawed and must be improved or replaced. Remember that the best selection of set points for design parameters cannot make a bad concept into one with superior performance and robustness.

Stress Factors (“Noises”)

Input Signal

Technical Function

Output Response

Controllable Design Parameters

The second group of Figure 8: A Parameter Diagram Assists in the parameters that influence the Understanding of an Engineered Function function includes stressful factors that are unavoidable, that is, beyond control of the design. Alternatively, you may choose not to control them due to concerns for their economics or practicality. Jargon based on Taguchi Methods calls these stress inducing parameters Noises, a term derived from communications technologies for which undesirable variations in audio output are clearly noise. It may be more understandable to think of these factors as stresses or deteriorating forces beyond control. In our football example, wind, rain, and mud are stress factors beyond the control of the team unless they chose to invest in a roof over the stadium. In many cases, that would be judged to be offensive to the purity of the game or too expensive. However, there are controllable parameters of the kicking function that reduce the vulnerability of the process. These may be in the equipment design, such as the characteristics of the shoes and their cleats, in the human movements that are practiced, or in the ability of the kicker to focus his attention. Proper shoe features, the practiced centering and alignment of the ball, and the kicker’s mental ability to disregard distractions may have great effects on the variability of the ball’s trajectory. The mean of the distribution of trajectories may be more controlled by the kicker’s alignment with the goal posts and the consistent movement of the kicker’s legs and feet. What really works will depend on the particular skills of the individuals. If nothing works well enough, it’s time for a new kicker. So the parameter diagram illustrates details most relevant to robustness development. Figure 9 shows the model of the parameter diagram with more details about the output response of a function and those parameters that affect the mean and variation of that response. It’s a good reference, illustrating the following: a. The objective of the function is to transform energy to achieve the output response. b. The input signal comes from a function upstream in the system.

Summar y

c. The output response has requirements decomposed from the higherlevel system requirements. d. Design parameters that are controllable can be either fixed or adjustable; their roles are to reduce the variability of the output response and to position its mean at its requirement. e. The stresses that influence the output response are not controllable, and can be found in any of three broad categories, depending on the actual applications. Stress Factors (“Noises”) • Degrading forces beyond control • Cause variations in output response • Sources: ° Customer’s use; Environments ° Internal interactions; Wear ° Manufacturing process variations in assembly or supply chain Input Signal • Output from an upstream function

Technical Function decomposed from full system static or dynamic energy transformation

Controllable Design Parameters • Fixed parameters (CTF) or adjustments (CAP) • Fundamental to the science and engineering of the function • Found to be most efficient at reducing the variations due to stresses and adjusting the mean to the target • Chosen to be controlled economically

Output Response (CFR) • Requirement deployed from customer expectations for quality • Characterized by performance mean and standard deviation

Figure 9: A Detailed Parameter Diagram Facilitates the Development of Robustness

The benefits of laboratory experiments can be greatly improved by employing stresses known to be effective at forcing variations in performance. Certain stresses, such a temperature and humidity variations, may act as surrogates for other stresses not so easily implemented, such as dimension changes due to manufacturing process variability. By deliberately introducing stressful conditions into experiments, problems can be found more quickly. Their solutions can be developed earlier in the process and proven to be superior to alternatives.

Summary We have seen that the fundamental contributors to the loss of product quality, reliability, and durability are the variations in the functional responses of designed subsystems, components, and materials. Performance variations are aggravated by stresses but are controlled by supe-

787

788

The Anatomy of Variations in Product Performance

rior design concepts, with controllable parameters set to optimize robustness. Stress is the enemy of performance. However, stresses are the friends of engineering. They enable problems to be discovered more quickly and generate superior solutions.

Additional Readings 1. Brassard, Michael, The Memory Jogger Plus+: Featuring the Seven Management and Planning Tools, Methuen, Mass.: ASQC Quality Press, 1989, ISBN 1-879364-02-6 2. Don Clausing, Total Quality Development: A Step-by-Step Guide to World Class Concurrent Engineering, New York: ASME Press, 1994, ISBN 0-7918-0035-0 3. Fowlkes, William Y., Creveling, Clyde M., Engineering Methods for Robust Product Design, Using Taguchi Methods in Technology and Product Development, Reading, Mass.: Addison-Wesley Publishing Company, 1995, ISBN 0-201-63367-1

About the Author Bill Jewett is a consultant to businesses engaged in the development of new technologies and multi-disciplined products. With insights into important paradigms and advancements in practices, he assists improvement teams in upgrading their engineering and management processes, project management practices, cross-functional teamwork, and the governance of development programs. For many years Bill worked for Eastman Kodak Company and Heidelberg Druckmaschinen, with focus on the development of high volume electrophotographic copiers and printers. Among his division-level responsibilities were the management of product development programs and of competency centers for mechanical and systems engineering. At the corporate level he was one of the authors of the processes for advanced product planning, technology development, product commercialization and their related governance. For over a decade he taught the processes and coached teams in their adaptation and implementation. As the process steward, he evolved the models to incorporate lessons learned from internal practice and external benchmarking. Currently Bill and an associate are writing a book to integrate strategies and methods for the development of increased robustness and higher reliability in products. They expect their book to be available in the last half of 2007. Bill can be reached by telephone at 585-705-3100, or by email at [email protected].

Benchmarking—Avoid Arrogance and Lethargy By Sy Zivan This article describes how functional and overall enterprise performance may be improved through the use of Benchmarking, a process that utilizes performance indices and Benchmarking partners to uncover leading practices. Executives typically defend their familiar and current methods and practices, especially within an industry that lacks competitive pressures. Such arrogance and lethargy propagates limited vision and stagnancy. Benchmarking compares areas of an enterprise to similar work performed by other, hopefully better-performing, practitioners. It is designed to help executives overcome managerial arrogance and lethargy. Leading practices are uncovered through researching how other business models perform given tasks. Those in search of solutions ask, “How are others accomplishing what we are attempting to accomplish?” Performance indices measure the effectiveness of a target area. Other companies, commonly called Benchmarking partners, are chosen for the actual comparison of target area performance and practices. Motivated to reach higher levels of performance, the executive conducting the benchmark comparisons looks for potential changes that will improve performance when implemented. These changes make up a set of leading practices, which will be included in the revised business plan. After locating these leading practices, commitments are made to adopt methods and practices that maximize performance. Executives attempt to achieve the highest levels of performance by finding new ways to do business. When targets such as unit manufacturing cost, time-to-market for new products, order-to-delivery cycle time, or inventory turns are simply assigned as part of planning, executives are then left to find the practices that will achieve targets. Using Benchmarking, the search is first for the most effective practices that should lead to the highest levels of performance. After leading practices are identified and assessed, commitments are made within the context of the current business and operating plan.

What Is Benchmarking? Benchmarking compares a company’s performance against that of recognized industry leaders. Its goal is to assist an organization in finding and establishing practices that significantly bring about improvements when implemented. Benchmarking is a systematic way of looking outside the enterprise for improvements as opposed to staying inside and defending the status quo. Frankly, the practices sought were first labeled “best practices” in the early years of Benchmarking. Lacking any proof that the “best” were indeed the best, the search turned to a more defensible target, leading practices. 789

790

Benchmarking—Avoid Arrogance and Lethargy

In truth, we should constantly be in search of better practices and not be paralyzed into inaction arguing over whether a practice is leading or best. Leading practices must be evaluated and selected by the particular executive managing the area under study. In this respect, Benchmarking does not replace functional expertise but helps the executive seek and recognize the better if not the best way to perform certain tasks. Benchmarking must be imbedded within the business planning process to be effective. The Business Plan lays out the future of the enterprise and how each function will contribute most effectively to this future. The role of Benchmarking is to compare results but, most importantly, to uncover the practices that can generate the best results. Corporations, divisions, functions and individual workers must then change the way they work by adopting the practices that will permit addressing competitive challenges. It is the association of Benchmarking as an integral part of planning that puts teeth and memory into the process. Without this, Benchmarking takes the form of another interesting study whose results are filed but never implemented to benefit the enterprise. Figure 1 illustrates how the Benchmarking process links to and supports the overall business planning process. For additional information, see “Common Problems and Pitfalls of Benchmarking and How to Overcome Them,” The Journal of Applied Manufacturing Systems; University of St. Thomas Technology Press; Winter, 1992. Corporate Priorities

Business and Functional Priorities

Where Benchmarking Adds Value

Search for Opportunity Areas Development of Business and Divisional Strategies to Capture Opportunities

Commitment to Improve Performance

Work Process Changes

Performance Measurement Comparing Actual Results to Commitments

Rewards Systems

Figure 1: Business Planning Process

What Is Benchmarking?

One further distinction must be made. Because this process was conceived from the realization that new competitiveness challenged the complacent achievers of old, the initial effort was to seek out or benchmark the practices of these new competitors. In fact, the process was labeled Competitive Benchmarking. Further refinements of this process revealed that all leading practices were not owned by competitors. In fact, as we shall see, uncovering leading practices requires stepping through the various major functions of a business and determining where leading practices truly exist. There should be no assumption that leading practices may only be found within competitors’ operations. One very practical consideration in the search for leading practices is the fact that competitors are rarely willing to share anything with their competition. To limit your search to competitors may be short-changing your efforts. Xerox Corporation, for instance, found leading Distribution Center practices at L.L. Bean—hardly a competitor. In the past, internal brainstorming and “squeezing” had been the way improvements were typically sought. Seeking greater effectiveness is the role that Benchmarking must fill. It looks outside the firm, within all industries, for ways to improve. How should products be developed to reach the marketplace in the shortest period of time? How should these products be sold? How should they be distributed and serviced to meet the needs of customers? Benchmarking was conceived from a need to overcome lethargy and arrogance. Corporations in several industries in the United States had experienced dominance and believed this dominance to be theirs forever. In the 1970s, many of these corporations were faced with falling market share, revenues, and profits as a result of new competition, mainly from new entrants from the Pacific Rim. Their existing ways of doing business had served them well and allowed them to gain dominance. So why change? Executives and managers had been richly rewarded for doing what they had been doing—these new reversals had to be someone else’s problem. Xerox began using Benchmarking in 1979 to analyze unit production costs in manufacturing operations. Pricing of Japanese plain-paper copiers was extremely low and a major challenge to Xerox. What were the practices of the Japanese that allowed them to manufacture high quality copiers and sell them at attractively low prices? Having created the copier industry, Xerox had not faced Pacific Rim competition before. Its business grew, and its people were rewarded. It was difficult to imagine someone coming out of seemingly nowhere and out performing Xerox in any area. Xerox was guilty of and had started paying the price of arrogance and lethargy. A business planning process involving Benchmarking that forced each process owner to look outside Xerox for leading practices was the only way to prove to those who thought they were the best that there indeed might be better practices yet.

791

792

Benchmarking—Avoid Arrogance and Lethargy

Integral to an emphasis on quality improvements, forward-looking industry leaders rejected complacency and sought a way to motivate their organizations and to develop cultures where change was a way of life and competitiveness ruled. One way to do this was to compare the performance of their company to that of others. Right off the bat, the slippage of market share and profitability signaled a performance problem relative to the new competition. But why? To get to the heart of this slippage, business practices had to be examined and a way had to be found to compare essential functional practices to those of the best in industry, capture and adopt these leading practices, and accept as business goals the results that leading practices would produce.

The Benchmarking Process There are three major elements in Benchmarking a business function: 1. Determine the performance indices of the function. How do you wish to measure its effectiveness? 2. Identify Best Performers in industry for this function. Who do you wish to use as sources of leading practices? Who will be your Benchmarking Partners? 3. Uncover the leading practices from your Benchmarking Partners.

Select Performance Indices for all Participants

Identify Best Performers

Determine the Practices of “Best Performers” for Implementation

Figure 2: Benchmarking Process Flow

Figure 2 illustrates the Benchmarking process flow.

Performance Indices Benchmarking is often confused with numerical measurements. In other words, as opposed to uncovering leading practices, the process is reduced to graphical or tabular presentations that indicate how an array of competitors’ performance compare. These indices are often referred to as benchmarks and are presented as the culmination of a Benchmarking study. A Benchmarking study with these limitations is not of great help. Consider measuring the blood pressure of people gathered at a meeting place. Telling someone of their relative high blood pressure without indicating how he or she can attain greater health is of questionable value. Benchmarking seeks leading practices and is of little use if it stops at numbers alone.

The Benchmarking Proce ss

Having established this, the Benchmarking process must start with the identification of the indices that will describe the function’s performance but must not end here. These are the indices that must be improved to garner success. These are the indices that adoption of truly leading practices will move in the desired direction. Examples may include: inventory turns, orders per sales rep, order to delivery cycle time, manufacturing unit cost, and bad debt as a percentage of revenue. Frankly, these measurements should not have to be developed for the Benchmarking process but should exist as part of the existing business planning process. Assuming there is a business planning process, you would think that these are the same indices used to measure the effectiveness of the function, whether Benchmarking is at play or not. A leading practice, when adopted, will move measurements dramatically in the correct direction.

Benchmarking Partners An ideal Benchmarking partner is a company whose business demands excellence in the area you wish to improve. If you are trying to improve your safety program, you should seek partners whose business is by nature dangerous and where there may be a higher probability of accidents or, in turn, a need for an aggressive safety program. In this case you might choose a mining or chemical company given that they will probably have addressed safety more than others. When Benchmarking Customer Service, you might look to overnight package delivery or a mail order house because Customer Service is their business. A turnaround success story would probably be a good benchmark for Employee Communications. Team brainstorming meetings are good venues for the exploration of potential partners using this logic. Research journal articles concerning the work of potential partners in select fields to help confirm their candidacy. In summary, overall business success might not be the criteria for choosing a Benchmarking Partner for specific areas. The Benchmarking enterprise needs to seek companies that focus on and excel in the same or similar functions that are being benchmarked.

Leading Practices One company’s leading practice may not work for all companies due to changes in customer needs. A leading practice takes advantage of current technology to meet customer needs at the lowest possible cost and capital investment. It can only work if it is located and adopted by the people responsible for performance. Three examples of the use of Benchmarking to uncover leading practices involve Supply Chain Management, Human Resources and Community Relations, and Strategic Planning.

793

794

Benchmarking—Avoid Arrogance and Lethargy

Supply Chain Management A common leading practice utilized by several clients is the adoption of Supply Chain Management by a firm that hitherto had not properly recognized the continuum of material flow from its vendors through inhouse processing, on to out-bound distribution, and final delivery to its customers. Improvements in the predictability of customer demand significantly reduces the investment needed to support inventory in the “pipeline.” Additionally, inventory requirements were significantly reduced through the elimination of unneeded product stocking locations. In this scenario, customer needs were met while focusing on reduced capital and expenses. Benchmarking Partners attempting to meet comparable customer needs and exhibit both high return on assets and inventory turns provide good practice comparisons. Human Resources and Community Relations Another less common client scenario involved Human Resources and Community Relations. The rural location of its plant and research facilities required a company to satisfy the community needs of technical employees in order to both attract and keep these employees. A Community Relations Department sought leading practices for its involvement in local school boards to assure these schools met the high requirements of employees with advanced degrees and their families. Methods used to assure adequate shopping locations in rural settings also became a leading activity. Recruitment and retention targets measured practice effectiveness. Strategic Planning Process Through Benchmarking A third example of a leading practice in the area of strategic and business planning is the requirement that Benchmarking be made part of the strategic planning process. This assures that all future plans are developed using leading practices in all areas of the enterprise. The requirement that Benchmarking be made integral to the planning process assures that the findings of the Benchmarking process are locked as plan commitments by process owners. Additionally, the projected and measured outcomes become the committed numerical targets for process owners. Uncovering leading practices first involves generating the questions about functional performance that you would like to address with your Benchmarking partners. Here again, so many willing participants in this endeavor go wrong by concentrating all their efforts in asking questions that relate to the levels of performance of the partner and not how the partner is achieving their performance. It is important to realize up front that Benchmarking Partners rarely share performance indices with others, even when they are more than willing to share practices. The practices Benchmarking is designed to uncover are those that will improve your measured functional or overall enterprise performance. You cannot transplant the

Tips for Succe ss

results of others—you can only transplant the processes used by others in a manner that has the potential to improve your own performance. To generate a suitable set of questions, Cause-and-Effect diagrams are often used to identify factors that are blocking the function from achieving better levels of measured performance. This listing of factors is then turned into a list of questions to be addressed with the chosen partner. For example, high turnover in the sales force is a sure block of sales performance. A question could be raised that has to do with the tactics employed to maintain low levels of turnover. Another example may involve the need to maintain high inventory levels to meet order cycle time. A question may be crafted relative to the design of the partner’s supply chain that balances customer needs, inventory costs, and logistics expense. Questions comprising the information desired about a Benchmarking partner’s functional practices are brought together into a guide that can then be forwarded to potential Benchmarking partners when they are approached to participate in the study. Perhaps the first thing that a potential partner may say when approached will involve the questions that will be asked of them. Be prepared and share these questions as early in the process as possible. Benchmarking studies may be carried out in several ways. The most customized and perhaps costliest way is for an individual company or organization to sponsor its own study of a particular function. In this situation, the sponsoring company designs the entire study and executes it, sometimes with the help of consultants. Companies who agree to participate will often do so under the condition that they will receive, at minimum, an executive summary of the leading practices that were uncovered. A second way to perform a Benchmarking study would be through the formation of a consortium of companies that are all interested in the same information. Through everyone’s cooperation and with the help of a third party, Benchmarking practice questions are developed. A major advantage of this technique is that all the members of the consortium start off as Benchmarking partners. Also study costs are shared. A version of the consortium of companies is the industry association that recruits its membership to participate or offers these studies as benefits of membership. Regardless how the study is organized, the administrator must recruit participants, ask all the formulated questions, and report findings for action back to all participants. The benefits of using Benchmarking can now be realized by reflecting in the uncovered leading practices that were judged to benefit performance.

Tips for Success The following is a set of tips or bits of advice that should assure success in performing Benchmarking studies.

795

796

Benchmarking—Avoid Arrogance and Lethargy

1. Benchmarking is too often thought of as a comparison of numbers or performance indices as opposed to a search to uncover leading practices. Benchmarking is a search for leading practices that can be translated into projected performance measurements. You cannot use another company’s numbers, but you can adopt another company’s practices to make your numbers better. Having said that, it is urgent that the performance of each benchmarked function be measured so that improvements can be targeted and captured once the leading practices have been identified. 2. Benchmarking practitioners often do not realize that this process should be made an integral part of the strategic planning process. It is all too often used outside of planning where its performance benefits can be lost. It is viewed as an activity of interest but not a way of capturing permanent improvements. Benchmarking within the planning process allows for improvements to be tracked and captured. 3. All too often, when Benchmarking is adopted by a company, it is mandated for use throughout all areas of the company. Given the cost of using this process, areas should be selected for Benchmarking that will have a substantive effect on overall organizational performance. There is no need to benchmark everything in sight, but there is a great need to seek leading practices within critical areas of the business. 4. Those who benchmark a function should be the same people who manage the function and will be responsible for implementing change. The only way to gain commitments for improvements from change in a functional process is to have the managers of the function responsible for seeking the changes. 5. The questions that are to be posed to partners must be fully documented. Often those seeking leading practices cut corners and assume documentation is not needed. First, Benchmarking Partners will not agree to participate without knowing the detailed subjects to be covered, and any attempt to benchmark without such documentation is a waste of time for all involved. 6. Be able to answer all the questions that you will pose to partners. There is a need to compare how all participants perform. The originator of the study is part of the study. 7. Assure that you have selected qualified Benchmarking Partners. Remember that you must select partners whose business requires excellence in the function under study. In the last analysis, success in Benchmarking depends on the willingness and desire of executives and managers to change and become more

About the Author

competitive. Lacking this, the best planned Benchmarking effort will not capture needed and available performance improvements.

About the Author Sy Zivan currently assists numerous corporations in developing strategies, plans, and programs to ensure that their products and services meet or exceed customer needs. He is best known for his expertise in using Benchmarking to make clients aware of opportunities to meet customer needs with minimized expense and investment in assets. Zivan retired from Xerox where he held finance, systems, control, and administrative positions. He was Vice President of Logistics and Distribution, responsible for all equipment, consumables, and spare parts logistics in the United States when the Benchmarking process was first designed and applied within this function. He was also influential in establishing strategies for a world-wide logistics network. His publications include • “A Xerox Cost Center Imitates A Profit Center,” Harvard Business Review, coauthored with Frances Gaither Tucker and Robert C. Camp; January 1, 1987. • “How to Measure Yourself Against The Best,” Harvard Business Review, coauthored with Frances Gaither Tucker; May 1, 1985. • “Common Problems and Pitfalls Of Benchmarking and How to Overcome Them,” The Journal of Applied Manufacturing Systems; University of St. Thomas Technology Press; Winter, 1992. Mr. Zivan can be contacted at (585) 586-3105 or at [email protected].

797

This page intentionally left blank

Building Strength via Communities of Practice and Project Management By Lynne Hambleton

Successful implementation of introducing a Six Sigma method to an organization works best when coupled with two other tactics: 1) strengthening the approach with solid project management (PM) fundamentals and competencies and 2) leveraging communities of practice to embed the concepts into daily work. Project management improves a Six Sigma deployment by expanding the toolset to better manage a project. Nine knowledge areas and project-lifecycle concepts reinforce and fortify 1) the balanced-focus on the customer requirements (the process or item of interest) and 2) the project perspective of how to deliver the requirements. The project management discipline adds robustness to any type of Six Sigma approach. While Six Sigma fundamentals and its tool-task-deliverables linkages contain some project management elements, the full deployment of the PM discipline takes the integration a step further. Regardless of what type of Six Sigma an enterprise implements, whether it selects one or all of the Six Sigma disciplines (for example, Lean Six Sigma, the problem-solving DMAIC, Design For Six Sigma (DFSS), or Six Sigma for Marketing (SSFM)), the resulting benefits are enhanced by incorporating project management. Irrespective of the organization’s size, scale, or scope (whether in a Fortune 100 firm, a non-profit, or government agency), the addition of PM improves a Six Sigma implementation. The operative question is, “how can the implementation architects entrench both Six Sigma methodology and project management into the workforce’s daily operations so it becomes ubiquitous?” The ultimate challenge is to sow a new way of thinking and operating into the work culture—a working environment that empowers the community of employees to embrace and adopt the new philosophy and new way of work with the least resistance. Building and strengthening a new discipline is best achieved through communities of practice. The purpose of this article is to provide chartered architects the methods to implement Six Sigma in their organization using a blueprint that goes beyond the topic at hand and strengthens the deployment and adoption rate. This blueprint includes a strategic direction, key enablers, and a high-level set of required deliverables that create and maintain a community of practice across an enterprise. It provides guidance to help build and strengthen organizational discipline for both Six Sigma and project management.

799

800

Building Strength via Communitie s of Practice and Proje ct Management

Community of Practice—What Is It? A community of practice (CoP) is a defined network of people who come together voluntarily to share a particular interest, discipline, profession, or area of interest. Members gain satisfaction through mutual benefits. The individual contributes to the knowledge and betterment of the working community and in return, receives professional recognition for his/her contributions. The CoP, as an entity, serves as a subject matter expert (or knowledge resource) to its members. Thus, its members can leverage one another’s experiences, skills, and thinking when exploring possible approach modifications or adaptations that better address situational changes. The CoP creates a safe environment wherein a member seeks confirmation and collaboration in practicing their trade or discipline. A CoP’s roots originate from the master-apprentice relationship, where a trade or skill often is handed down through role modeling and word-ofmouth. At first, the master mentors the (typically younger) apprentices. As a master’s apprentices grow in number, so expands the connectedness of a group whereby new ideas are debated or explored. Eventually, the creativity and adaptive nature of the apprentices begin to enlighten the master’s thinking. Knowledge eventually propagates among the group fluidly. Over time, the original master may no longer be the hub of the wheel, and the leadership role begins to shift naturally to another wise person; perhaps one with more time to nurture the group’s connectivity or interrelationship. A community of practice benefits an organization in several ways. It provides extended reach, where the utilization and knowledge of a particular practice or topic reaches beyond a team or departmental boundaries into other parts of the organization. A community of practice also helps to establish knowledge and skills into the day-to-day work practices better than a traditional training program. A CoP extends beyond what an event-based classroom can deliver and embeds the learning within ongoing operations at the workplace. If a community of practice emerges within an organization, it builds the organizational knowledge through best practices sharing. Best practices sharing can occur organically—intra-business and even interdepartmental sharing. This form of in-house knowledge exchange is not only economical, but also powerful in that the best practices already accommodate the internal culture. Organizations can foster the establishment and infrastructure of a CoP to promote organizational learning. To harness this invaluable intellectual capital, an enterprise may offer it resources as aids to knowledge sharing, such as ongoing training, availability of related books and journals, an electronic database or repository, conferencing technology (rooms, telephony, and web technologies), additional collaboration technology, and most important, company-sanctioned time to participate in the CoP.

Proje ct Management—What Is It?

Caution The organization walks a fine line between establishing a formal CoP and allowing one to emerge naturally. If the CoP becomes too institutionalized, the members may lose their passion and desire to actively participate. The most successful CoPs evolve from a passion for the topic, a sense of ownership toward the community, and a sense of value received from participating (as a giver or receiver of information, or both). A CoP should not be a formal department found on an organization chart. Its tentacles reach beyond a classic organizational team or department, and it seeps into the fabric of the organization through networking, role modeling, workshops, and simply practicing its trade. For the most part, CoP membership is incremental to employment in the department that funds the paycheck. Hence, each CoP member has a personal investment in an area to take on these additional responsibilities. If the institution controls the CoP operations such that participation seems mechanical, administrative, or sterile, then the active and innovative nature of membership will wane. An organization can provide some of the infrastructure and administrative underpinnings to a CoP, as an enabler, not a controlling entity. CoPs thrive in a nurturing environment. How to create such a setting is described below.

Communities of practice have a finite life, and organizations should avoid promoting a CoP beyond its natural life. Interest will naturally subside as shared knowledge or level of expertise becomes relatively uniform amongst its members. Communities of practice form when a concept or new application emerges, then people gravitate toward one another to share thoughts, ideas, and theories. Once a concept becomes relatively ubiquitous, the natural attraction to unite and explore possibilities for that particular subject declines, and gets replaced by a newer challenge.

Project Management—What Is It? Project management, as a concept, enjoys greater awareness than communities of practice. Similar to Six Sigma, project management takes on several dimensions—as a profession, as a discipline, as a structured method or way of work, and even as a company philosophy (as some organizations organize and operate using a PM structure). The work by the

801

802

Building Strength via Communitie s of Practice and Proje ct Management

non-profit Project Management Institute (PMI) to promote, educate, and certify practitioners on its set of standards has gone a long way to advance awareness and professionalism of the discipline. PMI reports the existence of over 200,000 project management professionals in over 125 countries. A well-executed project management discipline benefits an organization in several ways. It provides a professional foundation for employees to respond intelligently to change, think before acting, and deliver the requirements on time and within budget. The PM discipline adds a structure to thinking and working that applies to any situation, and complements and strengthens any Six Sigma approach. By adding the PM discipline to Six Sigma projects, clients have solved the issue of run-away projects, in that they seem to have no end. PM provides additional rigor and tools to better manage a project, extends beyond the product, service, or process of interest, and extends into the leadership and management dimensions. According to PMI, “Project management is the application of knowledge, skills, tools, and techniques to a broad range of activities in order to meet the requirements of a particular project. Project management is comprised of five project management process groups—Initiating Processes, Planning Processes, Executing Processes, Monitoring and Controlling Processes, and Closing Processes—as well as nine knowledge areas. These nine knowledge areas center on management expertise in Project Integration Management, Project Scope Management, Project Time Management, Project Cost Management, Project Quality Management, Project Human Resources Management, Project Communications Management, Project Risk Management, and Project Procurement Management.” [Resource: A Guide to the Project Management Body of Knowledge (PMBOK® Guide), Third Edition.] Basic PM terminology includes the following: • Project management is the management of discrete projects that have a definite beginning and end. • Program manager has two definitions. The first being a super-project manager, whereby several projects (or sub-projects) and project managers all supporting a common and higher objective roll up to large program. A program often involves a large scale and/or complex scope, which requires its objective(s) to be parsed into smaller subsets or projects. Again, programs have a discrete lifecycle, with a defined beginning and end. In contrast, a second definition of a program manager is similar to a customer account manager, whereby the role focuses on the ongoing

Proje ct Management—What Is It?

process of managing respectively a program or a client/customer relationship. In this case, work is ongoing, and never ceases even if objectives are redefined. The interrelationship between program and project management is when project activities of planning, designing, and implementing a given solution are initiated to support a particular client/customer.

Project Management Industry—Best Practices The Project Management Institute (PMI) periodically conducts studies to understand how professional project managers fare in the marketplace. A 1995 PMI study of 22 companies identified key themes needed to achieve optimal and pervasive project management implementation. Some of those critical areas include the following: • Entrepreneurial leaders as project managers, such that the person serves as an inter-functional link that continually communicates the organization’s vision, mission, and business strategy. • Project management as a functional expertise, to span multiple functions within an organization. • Establishment of a program office at a corporate or division level. • Integration of project management as a core organizational competency. PMI reports that “functionally autonomous towers become political centers.” They often are “slow to accomplish tasks, particularly tasks requiring cross-functional cooperation. Hence, the overall effect caused the organization to become sluggish in responding to challenges, opportunities, and goals.” Project managers who serve as the project leader keep the team focused on the project goals within the defined time and budget performance measurements. These project teams are efficient and quick to respond to areas of risk such as poor planning, immeasurable objectives, scope creep, political changes, third-party cooperation, and an under-resourced team. According to PMI, the strong forms of project management—where the project manager has multifunctional authority—experience better success. PMI reports that the optimum structure to maximize business results is to have the project management function report to an executive or executive committee, whose authority extends across business units. A program office should promote strategic communications, mentor, cultivate technical competence of the professionals, and conduct (formal and informal) briefings.

803

804

Building Strength via Communitie s of Practice and Proje ct Management

Achieve Operational Excellence—Putting It All Together The combination of these three concepts (Six Sigma, project management, and communities of practice) solves many operational implementation issues associated with people resorting to old habits after training, or after the initial excitement of a new program wears off. The three concepts reinforce and complement one another. United, they direct people’s work and processes toward operational excellence so that the organization better achieves its growth goals. An organization can gain significant benefit from implementing just a Six Sigma method. However, it receives greater rewards when integrating both Six Sigma and project management concepts into the fabric of its ongoing operations. Depending on its starting point, the potential advantages of strengthening its operations via a standard approach supported by a CoP add exponential gains. The triad of concepts helps cement changes in how work gets accomplished, and brings an organization closer to achieving operational excellence. Some of the benefits clients have realized include: • Reaching and sustaining growth goals. • Improved market recognition for sound processes and professional resources. • Better leveraging of current intellectual capital resources through organizational learning. • Gaining a competitive advantage.

Achieve Growth To obtain the growth goals of an organization, the disciplines of both Six Sigma and project management support the planning, designing, implementing, and sustaining of the activities to achieve its strategic goals. An organization’s growth goals cascade down to several of its processes, all of which will respond by kicking off a program or project. Resultant projects could emerge in the strategic portfolio renewal process, the more tactical offering development process, and also in the post-launch field operations arena. Regardless of the organization, the combination of a data-driven Six Sigma approach, supported by PM and CoP, will better answer and deliver those targets. These projects will be managed more effectively and efficiently.

Excel in the Customer Marketplace Regardless of the industry, customers are more discriminating shoppers of products and services. They place higher demands on companies to

Achieve Operational excellence—Putting It All Together

provide faster, better, cheaper… offerings. Through both direct and indirect experience, customers “touch” a company’s processes, and gain an impression as to the professional nature of its employees and processes. If the customer has a poor experience, the company is lucky if the customer takes the time to complain; otherwise, it may be unaware as to the reason for the lost customer. An organization’s people and processes are an extension of its brand, and customers increasingly expect excellence, or they will shop elsewhere. Thanks to globalization, fueled by the internet, customers have a larger marketplace in which to shop. The flattening world understands the value of partnering and the risk of depending on an external organization. Hence, customers tend to select those suppliers who demonstrate professionalism, beyond technological expertise and success. Customers view a firm’s people and processes as two important and distinguishing factors. Market excellence will be granted to those companies that demonstrate efficient and effective operations and project leadership. The combination of Six Sigma, PM, and CoP provides an organization with the extra insurance policy that its people and process meet benchmark requirements, and exude operational excellence that wins over the customer.

Gain Competitive Advantage Professional implementation of both Lean Six Sigma and project management enhances an organization’s probability of success, and differentiates it in several ways. Regardless of where they are deployed, the benefits ultimately make the firm more competitive. Both internal and customerfacing processes become not only more efficient and effective, but also more proactive to anticipating and responding to market changes. The benefits of integrating the two disciplines include: • Consistent, predictable performance that balances the fulfillment of both customer specifications and business needs. • Adapting to marketplace and environmental changes in a proactive, timely fashion. • Enhanced work environment achieved through more purposeful, better guided, and integrated work organization-wide, coupled with improved professionalism.

Leverage Resources The Six Sigma and Project Management disciplines and processes cannot work without the people. The human resources coupled with the

805

806

Building Strength via Communitie s of Practice and Proje ct Management

processes are a key success ingredient. An organization that nurtures and leverages the intellectual capital amongst its employees possesses a competitive advantage over its peers. Nurturing involves investment in work environment, knowledge, and skills of the collective whole, not simply the individual. Work environment speaks to a host of traditional human resources topics (such as career path, compensation, and training), plus the infrastructure and support to nurture collaboration. The old adage that two brains are better than one applies. Collaborative work taps into the minds of many. To draw upon the many and leverage their resources, organizations invest in cross-functional teams and intellectual capital databases to capture and share intellectual property. However, communities of practice help to set the tone and culture wherein people draw upon one another to strengthen work— the idea generation and execution. CoPs thrive in a collaborative work environment. As a result, an organization’s human capital becomes that of the collective many, rather than that of a few key individuals. An enterprise leverages its intellectual capital to better fulfill the required activities and produce the desired deliverables. The use of CoPs encourages and accelerates the leveraging resources, which, in turn, reinforce the disciplines of Six Sigma and project management to promote an integrated team approach. A CoP often reaches beyond organizational boundaries. The resources needed to achieve a certain goal may exceed that of an organization’s current workforce. An organization needs to ensure that the appropriate complement of knowledge and skills participate in its processes. Hence, ensuring that appropriate resources are available may include sourcing from partnering, sub-contracting, hiring, and/or developing current employees. Blending these multiple resource types (all with varying backgrounds, experiences, and talents) and focusing them toward a common goal (for example, delivering a new solution) can be achieved through standardization in approach, methodology, and terminology— namely both Lean Six Sigma and project management. A CoP network cultivates this common language and approach across multiple groups. It connects them quicker. It embeds the common practice quicker in their work, and upholds a professional standard.

Meet the Challenge of Diversity Leveraging resources can present a challenge of balance—balance across a diverse workforce. This diversity comes from different work experiences, different professional disciplines, and different cultures. An organization often encompasses several pockets of expertise, characterized by similar disciplines, backgrounds, and experiences. Ideally, these pockets get knitted together by a common method of working—that is Six Sigma

Strategic Dire ction—How to Get Started

and project management structure. If no standard approach exists, then the organization loses an opportunity to easily leverage not only the individual pockets of expertise, but also the exponential synergies that can link them. Worse yet, if the organization hosts multiple methods (one per pocket of expertise), it lacks one common unifying language. This results in miscommunication, or little to no communication, across the pockets. A multi-approach existence fosters a silo-centric view and often-unhealthy inner-company competition across the operational arms. Adding another dimension of working with partners and sub-contractors often can exasperate the communications and linkage challenges further. The combination of a common language and approach (offered by Six Sigma and PM), and the inherent networking and best practices sharing nature of a CoP address the challenges of diversity, and converts them to a benefit. The integration of these three concepts reinforces the power of leveraging the knowledge and skills of the many. United, these three concepts help to celebrate each individual’s talents and work for the betterment of the organization and its mission.

Strategic Direction—How to Get Started Select the Methodology A methodology provides common speak, an approach across a diverse population (regardless if the diversity is experiential, geographical, organizational, or functional, and so on). The implementation of an integrated Six Sigma and project management discipline will provide the foundation, the structure for a common approach and a common language. Both Six Sigma and PM utilize a phase-gate review structure to define how a project progresses through to completion. Moreover, both disciplines emphasize the importance of documentation, and communication of lessons learned at the end of each phase. This reflection of what worked well and what could be improved provides the substance about which a CoP likes to debate. A CoP flourishes in an environment that promotes a common approach—in this case, an integrated approach. The CoP supports the methodology with sharing of knowledge (best practices/lessons learned). Mentoring and role modeling can occur across several operational entities. Managing of the business (at all levels) becomes easier. However, all this starts from a foundation of a common methodology. Ideally, the organization integrates the Six Sigma and project management into one hybrid model. Customize a Common Methodology and Toolset Given that the organization starts by establishing a common methodology and toolset, it needs to determine whether to embrace a generic

807

808

Building Strength via Communitie s of Practice and Proje ct Management

industry model or customize it to its business model and culture. Either approach has been successfully deployed; however, most that assimilate the discipline enterprise-wide eventually adopt the discipline as part of its philosophy. As a result, the leadership actively promotes the methodology and toolset. Such an organization may begin to utilize an industry standard model or customize it to better suit its unique needs. Regardless, organizations that successfully customize their methodology also maintain an active role in the industry discipline standards and best practices. A custom methodology should be a blend of best of breed approaches that meet an organization’s business model and need, along with a robust set of tools appropriate for multiple situations. This development work should be a co-design effort by key representatives of the practitioners throughout the organization and program office. (Program office will be discussed subsequently.) The method needs a clear set of phases (or steps), and corresponding requirements, deliverables, tasks, and tools for each. Along with the tasks, a role definition should identify whom the organization holds accountable for what deliverables and tasks, and who is expected to support the work. If the adoption of Six Sigma and PM is in its infancy, and the organization lacks best of breed content, it should begin designing a custom model by starting with a generic industry standard. The design probably should start with a standard Six Sigma approach, and integrate PMI into it. Part 1 of this book discusses key industry standard Six Sigma methods. Highlights of the key elements in PM start with the nine PMI project knowledge areas. These nine themes should be evident across each phase of a methodology. They are as follows: 1. Scope Management 2. Time Management 3. Cost Management 4. Risk Management 5. Quality Management 6. Communications Management 7. Human Resource Management 8. Procurement Management 9. Project Integration A sample PM toolset may include the following components: [Note: Many of the tools listed next are described in more detail in Part II, the tools and technique section of this book.]

Strategic Dire ction—How to Get Started

Scope Management • Project Charter Template—Reference Part II. • Change Control Log—A list of the project scope changes. Best practice suggestions include: • The number and date received of the Project Change Request. • The signatures of who approved (or disapproved) the change, and the date. It may require multiple signatures if the following are different people: the requester of the change, the manager of the resource to deliver the scope change, and the project sponsor. • Who is accountable and responsible to deliver the scope change, and the dates for expected start and completion, if approved. • Change Management Process—A document describing the Change Request process policies and guidelines to modify a project’s scope. • Project Change Request Form—A document describing a project scope change. Such a template should mirror the project charter structure and include the following components: a title of the change deliverable, the objective (or the rationale for, or gap it is filling), the goal or specifications (how success is measured), the specific deliverables, what is out-of-scope, the customer for the specific deliverables, and the required timeframe. The document also accommodates for the following dated signatures: the requester of the change, who approved (or disapproved) the change, the manager of the resource to deliver the scope change, and the project sponsor.

Time Management • Project Schedule Tools and Templates—An integrated calendar of milestones, activities, and tasks that represents the expected and actual dates for start and completion. In addition, often the percentcomplete estimate and the person accountable are also tracked. There are several software packages available to help manage the project schedule. See Also “Gantt Chart” and “PERT (Program Evaluation and Review Technique) Chart,” in Part II on p. 317 and p. 453, respectively. Cost Management Content covering the project budget for expenses on materials, labor and travel, and any one-time capital investments needed to meet the project requirements and produce its deliverables.

809

810

Building Strength via Communitie s of Practice and Proje ct Management

• Project Budget Worksheet—A standard template that defines the cost elements the organization wants the project to track. • Cost/Benefit Analysis Worksheet—A standard template that defines the cost and benefit elements the organization wants the project to track. Some organizations may choose to accept only hard savings rather than soft savings, or may choose to categorize and track them separately. The finance/accounting function should participate in defining the different categories and how to report on them. See Also “Cost/Benefit Analysis” and “Monte Carlo Simulation,” in Part II on p. 238 and p. 431, respectively.

Risk Management Content covering potential risks to meeting the project requirements and producing its deliverables, as well as counter-measures to address those risk events if and when they occur. • Risk Mitigation Plan—A risk management and contingency plan describing the process to identify, quantify, monitor, respond and control. See Also “Risk Mitigation Plan,” in Part II on p. 601 • Failure Modes and Effects Analysis (FMEA)—A comprehensive Risk Management matrix tool that is a super-set integrating many other risk management tools. See Also “Failure Modes and Effects Analysis (FMEA),” in Part II on p. 287 • Other Risk Analysis Tools that help to identify, monitor, analyze and manage risk events if and when they occur. These toolset may include: Cause and Effect Diagram (See Also “Cause and Effect Diagram—7QC Tool,” p. 173), Cause enumeration diagram (See Also “Cause and Effect Diagram,” p. 173), Cause and Prevention Matrix or Cause-Prevention diagram, p. 198, Fault Tree Analysis (See Also “Fault Tree Analysis (FTA),” p. 309). Process Decision Program Charts (PDPC) (See Also “Process Decision Program Charts (PDPC)—7M Tool,” p. 515). Some organizations or situations do not require the ongoing management of a full FMEA; in that case, any one of the following Risk Management tools may be used: • Risk Opportunity Analysis Model (ROAM)—An overall project risk analytical technique that examines multiple risk elements versus several different opportunity elements, then provides a rating and plots it in a two-by-two matrix. • Risk Event Assessment Form—A simple tool to list potential risk events and record the anticipated probability and impact.

Strategic Dire ction—How to Get Started

• Comprehensive Risk Rating Tool—A composite tool that evaluates risk elements from multiple tools, and provides an overall project risk profile aggregate score. • Risk Filter Process—A technique that sorts and diagrams potential risk events by impact to the customer, likelihood of occurrence, likely timeframe impact on project, and level of control (whether within or external to project control). • Risk Watch List—An actively managed prioritized list of potential risk items.

Quality Management Content covering the quality requirements of the project and its deliverables needed to satisfy and meet the acceptance criteria for project completion. • Management Quality Review and Checklist—A tool that assists management in conducting phase-gate reviews and also serves to document the project quality and performance goals. See Also “Checklists—7QC Tool,” in Part II on p. 204 • Customer Interview Guide Template—A tool that provides interview guidelines and outlines the required and elective elements of a customer interview. Often this tool is absent in many organizations’ standard toolsets. However, its inclusion helps ensure, across project teams, the criteria for understanding of customer needs versus wants and for considering the requirement for translation into CTQs when designing questions. See Also “Voice of Customer Gathering Techniques,” in Part II on p. 737 • Customer Satisfaction Template—A survey tool used to understand the customer satisfaction with a specific product and/or service. This tool may be absent from an organization’s standard toolsets; however its inclusion is invaluable. It can be utilized when several customers exist, and for both internal and external customers. Having a template that maintains organizational standards of excellence for satisfaction helps to compare satisfaction levels across projects and facilitates easier gathering, processing, and analysis of the data, and communication of results.

Communication Management Content covering the project communication process plan that identifies the key messages for each target audience and expected timing throughout the project. • Communications Management Checklist and Matrix—A tool that provides communication policy and guidelines and outlines the

811

812

Building Strength via Communitie s of Practice and Proje ct Management

required and elective elements of a communication plan. See Also “Communication Plan” in “Matrix Diagrams—7M Tool,” in Part II on p. 399 • Project Status Report—A template that provides status guidelines and outlines the required and elective elements of a project progress report. • Meeting Agenda Template—A document template with the required and elective elements of a benchmark meeting agenda. • Meeting Minutes Template—A document template with the required and elective elements of benchmark meeting minutes.

Human Management Content covering the project people resources and the supporting management processes, tools and infrastructure needed to meet the project requirements and produce its deliverables. • Project Staffing Request Form—A template that identifies the required skills, knowledge, and expertise to produce a specific set of deliverables, and used to request resources to be assigned to a project. • Project Directory—A list of the project stakeholders (team members, sponsor(s)). • Project Team Organization Chart—If the project team is large, this tool depicts the structure of the team, often illustrating sub-teams. • Project Responsibility Matrix—A tool that defines the various roles and responsibilities within a project team, often structured as an RACI matrix. See Also “RACI Matrix (Responsible, Accountable, Consulted, Informed)” in Part II on p. 554 • Project Close-Out Team Questionnaire—A two-part project manager document that provides: 1) project close-out guidelines, and 2) a survey on each project team. The project manager (or project subteam manager) completes the survey on each project team member evaluating the individual performance on completing deliverables, the timing and quality of work, application of tools, adhering to commitments, and overall team attributes. Note: This is not a functional technical evaluation of deliverables content, but focuses on the method and appropriate toolset application. A completed document should be supplied to the individual’s reporting manager. • Project Manager Evaluation Form—A survey document questions the project team on the project’s successes and weaknesses from the perspective of both the project scope and the project management. This should be completed by the project team and the project sponsor. A completed document should be supplied to the project manager’s supervisor.

Strategic Dire ction—How to Get Started

Procurement Management Content covering the project plan to procure and manage the inputs and services from third party partners needed to meet the project requirements and produce its deliverables. • Procurement Management Plan Checklist—A tool that provides procurement policies and guidelines and outlines the required and elective elements of the procurement process. • Procurement Contract Template—A document with the required and elective procurement terms and conditions.

Integration Content covering the overall coordination and assimilation of the component parts needed to meet the requirements, to produce its deliverables, and to complete the project within scope, on-time, within budget and per the quality performance specifications. • Project Notebook Template—Defines what project content needs to be actively managed and documented—both required and supplemental project elements. Best practice suggestions include: • The project manager develops and actively utilizes both a hard copy and electronic version of a project notebook throughout the duration of the project. • At the project kick-off meeting, the project manager distributes a project notebook to the project team. • The project notebook is to be updated and maintained for (at minimum) the project manager and project sponsor(s); often a project manager responsibility. The project team members often maintain their own notebook. • Work Breakdown Structure (WBS)—A hierarchy of project deliverables and milestones, often structured as a tree diagram. • Issues and Actions Log—A list of the project’s key issues and actions that the team wants to monitor and track. Best practice suggestions include: • The document often includes who the item is assigned to and the dates of the expected start and completion. • Separate document for the issues and actions, wherein the issues may become items to escalate to management, and actions tend to be more administrative and short-term in nature.

813

814

Building Strength via Communitie s of Practice and Proje ct Management

• Lessons Learned Template—A document should contain both guidelines and the template to document the reflections. The guidelines should describe the process for capturing what worked well and what did not, the frequency, how to document them, how to publish or communicate them, and to whom. Best practice suggestions include: • Complete within or at the end of each phase-gate in the methodology. • The template should include concept, potential timing of occurrence according to the project lifecycle, suggested for subsequent projects (counter-measures or planning tips), and person (or two) to contact if additional context setting or background information is needed by future project teams. • Consolidate and archive the Lessons Learned in the project notebook (both the hard copy and electronic versions). A hybrid Six Sigma/PM methodology addresses how best to integrate the two by not only blending the procedural steps and toolset, but also by defining the roles and responsibilities. The integrated method needs a clear RACI matrix to decipher the differences between Six Sigma and project manager roles. For example, the model must clarify which role is accountable to manage the project team. Once the methodology and toolset strategy is defined, the organization needs to design its roles and career path(s) depending on its business needs and the breadth and depth of its resources’ experience and skills. See Also RACI Matrix (Responsible, Accountable, Consulted, Informed)” in Part II on p. 554 In conclusion, once the common methodology is designed and agreed to, an organization needs to define how best to communicate and deploy this standard to its workforce. There are several enablers that an organization needs to address. They include the organization or personnel accountable for using the methodology, implementing, training, governing and sustaining it. These accountable individuals may comprise departments, functional groups, and organizational roles. Any configuration of who is accountable should satisfy the unique requirements of the organization, its culture, and its needs (or starting point).

Ensure Appropriate Resources People are the power behind a successful CoP that builds operational excellence. Operational excellence spans the deployment and ongoing utilization of the practice. An organization’s Human Resources function provides critical enablers to building a community of practice:

Strategic Dire ction—How to Get Started • Defining and implementing a recruiting/selection plan for the key

resources—the practitioners needs to be developed. • Enabling a career path that identifies development opportunities

needs to be defined. • Building a resource pool of competent subject matter experts is critical

to starting and stabilizing a community of practitioners. To that end, key considerations that HR should drive include the following topics.

Job Profiles and Career Path An important decision to consider is whether to create organizational positions and define a professional career path for Six Sigma and PM, or treat each of these disciplines as a competency. Both approaches have proven to be successful. The decision criteria rests upon the organization’s business model and culture. If the organization chooses to execute the competency approach, it wants to ensure that some level of basic Six Sigma and PM knowledge exists in all levels of the organization. Training then often triggers creative application of how to adopt the discipline to the current work practices. If the organization is relatively small in size, this may be the most appropriate approach. If the organization is large, it will reflect a heterogeneous utilization of the disciplines. To ensure some semblance of consistency, it may be easiest to promote the default industry standard approach and definitions to Six Sigma and PM. The firm should encourage and perhaps finance individual membership to ASQ and/or PMI at the international and local chapter levels, thereby allowing those individuals who seek the industry-sanctioned certifications in a given discipline, to bring that external knowledge and inter-company networking in-house to fuel the CoP. If an organization decides to create job profiles that depict the cumulative work experience within a discipline, the precedence has established role definitions for both Six Sigma and PM. In aggregate, these traditional roles reflect a hierarchy that portrays the level of work experience in the given discipline. Benchmark data exists so a firm can modify and adapt the model that best fits its needs. Both project manager and Six Sigma career paths should be built on cumulative assignments. Typical project manager positions and career path might include the following: • Project Coordinator—Coordinates project activities; meets time, quality, and scope goals; provides project plans and controls; and communicates with functional leadership. • Project Manager—Achieves project goals by applying standard approaches and procedures; provides project planning control and leadership; coaches and consults with team members; coordinates preparation for phase-gate reviews; communicates project specifics with stakeholders; and resolves project conflicts.

815

816

Building Strength via Communitie s of Practice and Proje ct Management

• Program Manager—Achieves program goals; mentors project managers and evaluates their performance; provides program planning, control, and leadership; establishes program priorities; resolves program conflicts; communicates with program clients; supports methodology, communicates, and trains; implements group strategic plan; and tracks and controls project costs and benefits. Similar to the Total Quality Management (TQM) initiative, some benchmark companies create new employee roles (such as Black Belt, Green Belt, and Master Black Belt as project leaders). Some also institute a new management or organizational structure, and new or revised project and operational processes to instill the concept. Typical Six Sigma positions and career path might involve the following positions: • Yellow Belt—Participates as a project team member, often with a functional expertise; and participates in data gathering and analysis activities. • Green Belt—Produces simple project deliverables; participates in data gathering and analysis activities; and coaches and consults with team members. • Black Belt—Produces project deliverables by applying the Six Sigma approach and utilizes the appropriate tool-task-deliverable linkage; prepares for phase-gate reviews; and communicates project specifics with stakeholders. • Master Black Belt—Mentors the Six Sigma practitioners and evaluates their performance; provides program planning, control, and leadership; establishes program priorities; resolves program conflicts; communicates with program clients; supports methodology, communicates, and trains; implements group strategic plan; and tracks and controls project costs and benefits. A good developmental position should be established to ensure a feeder base for future project managers and Six Sigma Black Belts. That role may be called a project coordinator or Green Belt, depending on the discipline’s focus. This assignment would be appropriate for someone aspiring to be a project manager or Black Belt. The duration in this position should equate to one cycle, to ensure movement. The coordinator or Green Belt would manage a portion of a project, under the tutelage of the respective project manager or Black Belt. A certified project manager and certified Six Sigma Black Belt, with demonstrated project success for a given period of time, should be eligible for a natural career progression. For example, the next career move may be a director or general manager position, whatever the organization deems appropriate. Hence, leadership attributes are important. PMI suggests that a seasoned project manager also should possess entrepreneurial

Strategic Dire ction—How to Get Started

attributes. However, regardless of the appropriate position options, they should be documented and published. The challenge facing an organization that embraces both disciplines may be how best to knit them together as a complementary, logical career progression. Again, there is no right approach; it depends on the business model and culture. Some organizations integrate the two disciplines as one, and expect the Six Sigma professional to possess PM competencies. Other organizations apply PM to mission-critical, strategic, complex, and relatively large in scale projects. If this is the case, typically the PM resources reside in an overlay function or the PM program office. Some organizations deploy the unique roles to different functional areas, such as Six Sigma professionals in engineering and manufacturing and PM for strategic and filed operations; however, this approach may be a good initial phase deployment strategy to create momentum and interest, and will not yield the best organizational benefits if the deployment stops here. Regardless of the starting point, the deployment plan ultimately should reach all areas and levels of the organization. When implementing both Six Sigma and PM, the organization could consider and select from a hybrid model or a lead model, wherein either the Six Sigma or PM discipline would represent the lead competency. The selection criteria depends on the organization’s strengths and needs. The options include the following: • Hybrid Model—The person accountable to lead a project team would be required to have certifications in both project management and Six Sigma as the core essential skills for such a position. The job positions and career path would mirror this model such that the various positions would reflect a hierarchy that integrated the competencies of and the work experience in the two disciplines. • Six Sigma Practitioners as Lead—The Six Sigma practitioner and project manager would be separate and distinct roles, with the certified Six Sigma practitioner as the project lead. The project manager role would serve as a support function responsible to administer and manage the ongoing project team, its performance, scheduled activities, and budget. Separate job positions and career paths would correspond with the two distinct disciplines. • Project Manager as Lead—Project manager and Six Sigma practitioner would be separate and distinct roles, with the certified project manager as the project lead. The certified Six Sigma role would be deferred to as the subject matter expert on content. Separate job positions and career paths would correspond with the two distinct disciplines.

817

818

Building Strength via Communitie s of Practice and Proje ct Management

Regardless of which approach the organization selects, there are two human resources requirements needed for any of the three implementation models: 1. A team leader, who ties the project’s cross-functional implementation closely to the organization’s vision, mission, and business strategy. 2. Team professionalism in fulfilling client promises and emphasizing performance measurement (on time, within budget) across functional boundaries to ensure successful implementation of even complex and sophisticated solutions. These requirements become one of the responsibilities of the organization’s leadership team, essentially the board of governors that provides strategic direction. The discussion on the board of governors concept is forthcoming.

Salary and Incentives The HR function should conduct an analysis of the salary range commensurate with the job profile and career path strategy the organization selects. If these disciplines are being introduced into the organization for the first time, the appropriate reward and recognition will fail to acknowledge the incremental competency or positions. If the discrete positions and career path strategy is selected, the salary and incentive plan for each profile ultimately needs to adequately compensate for the new required skills and experience and for the new organizational alignment. Seed Strategy A seed program may be introduced to create excitement around the new positions, to incite enthusiasm to adopt the new methodology, and to become a certified practitioner of either or both disciplines. The program could include a bonus incentive to a project that successfully applies the new integrated methodology and its tools, or to the individual that completes certification for either discipline. Selection/Recruitment Plans The human resources group should define a recruitment and/or selection program to identify the appropriate candidates to fulfill the project management and Six Sigma positions. Given the organization’s business needs, several options can be explored: internal sourcing; using a professional recruiter; and/or outsourcing to consultants. The latter option affords the organization the luxury of gaining critical mass of project management resources without incurring permanent commitments. Plus, if he/she matches the needs of any given organization, that operation may have the option to hire the consultant.

Strategic Dire ction—How to Get Started

Learning Continuum Professional development and training is critical to establishing a new discipline and/or methodology. However, learning cannot stop simply with just the project manager and Six Sigma practitioners. The learning needs to penetrate the entire organization. Both disciplines, at some level, need to be viewed as a core competency for the organization. In doing so, the organization’s performance improves. Project team members will be stronger contributors. Managers will conduct more effective phase-gate reviews. Thus, an appropriate learning continuum needs to be developed for all levels of the organization. A skills assessment tool should be developed, particularly for the disciplines’ practitioners, respective mentors, and project sponsors. This can be accomplished by a cross-functional team of human resource experts and the business leaders. The learning continuum should fill the gaps of the skills assessment. The continuum should also prioritize courses based on the business needs. For example, essential skills may define core curriculum. Beyond the method, tools, and best practices workshops, common core courses may include the following: • Business Statistics • Risk and Scope Management covering Due Diligence/Risk Assessment; Requirements Planning and Scope Control; and Project Estimating • Successful Project Management Concepts and Techniques covering Project Management Processes; Planning; Controlling and Communications Techniques; and Time Management • Monte Carlo Simulation enhances and may supplement some Risk Management and Project Management topics, depending on the organization • Effective Negotiation Technique • Design of Experiment (DOE) and/or Conjoint Analysis covering robust design and/or customer preferences and other various types of DOE techniques The organization’s core learning principles should be appropriately reflected in the methodology, its toolset, and related learning events—for example, utilization of on-going reflection (lessons learned) throughout the project to promote quality assurance and manage risk. Peer-to-peer sharing of knowledge (best practices/lessons learned) should be solicited during formal activities, and encouraged to continue throughout the work practices. Formal and informal networking and mentoring/

819

820

Building Strength via Communitie s of Practice and Proje ct Management

apprenticeship programs should be planned to encourage participation from all levels of experience. These principles reinforce the cultural elements necessary for collaboration in which communities of practice thrive.

Align the Resources Resource alignment, along with an organization’s governance and supporting infrastructure, can either enhance or hinder the leveraging of its assets. Multiple iterations exist on how to organize people. Each organizational structure design poses its own set of pros and cons. These various designs feature functional, business, or project configurations. The added complexity of a matrix configuration blends either the functional, business, or project structure. The relationships within a matrix can range from weak, balanced, and strong construction, depending on how flexible and fluid resources are shared. PMI purports that a functional organization’s structure minimizes the project manager’s authority. Often both the project manager and project team are assigned on a part-time basis, where the functional responsibility distracts them. In the Project Management Body of Knowledge (also called PMBOK Guide), PMI states that the most effective organizations are structured by project. In such an organization where project managers are full-time, they are held accountable for and empowered to manage the assigned project. No right configuration exists; it depends on the organization’s business model, culture, size, and available resources. Regardless of the structure employed, an organization should consider how the project is staffed, how the resources support a project, and how the project is governed. One best practice that transcends the various organizational designs is establishing a program office.

Program Office The program office staff should comprise subject matter experts in both Six Sigma and PM. The program office, by design, should serve as the hub of the wheel as the center of excellence. But as with all good masterapprentice models, it should spawn and nurture other pockets of excellence, as appropriate. Hence, the people staffing this function should be carefully selected. The program office individuals should have strong traits in mentoring, coaching, leadership, and strategic thinking, coupled with seasoned experience. Their resume would include a proven track record of demonstrated competencies in operations (a mix of successes and failures) wherein the disciplines were applied, and industry-level certifications (such as Project Management Institute (PMI) and American Society for Quality (ASQ), or equivalent). Given this expert-level of technical excellence and experience, ideally this team of people would wield leadership out of reverence, wherein the fellow practitioners throughout

Strategic Dire ction—How to Get Started

the enterprise admire them for their professional expertise, knowledge, and humility. The philosophy of the program office should be one of guidance of colleagues. Sometimes the program office members may work alongside the other practitioners to role model on a project and other times promote thinking by asking thought-provoking questions. They should initially facilitate peer-to-peer information sharing, and may conduct skills advancement workshops, but are constantly seeking suggestions and input from fellow practitioners, or encouraging peers to take initiative. While the program office may be small in number (i.e., number of people), its budget should allow for it to conduct a series of forums/workshops. These events facilitate mentoring, knowledge sharing, and continued learning/integration into the work practice. Guest speakers featured during a forum event may be sourced from industry associations (for example, PMI and ASQ), clients/customers, or practitioners with an experience to share. These forums will facilitate networking, mentoring, information sharing, and work practice problem solving. Continuous improvement of the toolset will be achieved by soliciting practitioners’ suggestions, which reinforces both their ownership of the toolset and its evergreen nature. The program office should be responsible for investigating additional internal and external learning opportunities that build competencies in both disciplines. It should also explore implementing formal mentoring/coaching programs, experience portfolios, a knowledge repository, publishing newsletters, and negotiating organizational memberships to groups such as PMI and ASQ. The program office charter might include the following: • Build and promote an organization-wide community of practice among Six Sigma and project managers and interested parties, in which learning from one another readily occurs. • Define and implement a learning continuum for both practitioners and others interested in Six Sigma and PM competency and certification. Develop a Learning Continuum and Certification process. • Standardize the organization on a single methodology and toolset. Document the method, roles and responsibilities, and governance process in an easily accessible format. Identify and determine additional documentation and communication material needs, such as job aides and communication collaterals for management, customers, and new project members. • Assist the organization in assessing the healthiness of its projects and the professionalism of the practitioners. • Provide resource tracking and management that matches a need with required skills/experience.

821

822

Building Strength via Communitie s of Practice and Proje ct Management

Board of Governors The program office activities ought to be directed by a board of governors comprised of each of the organization’s operational leaders. The board of governors is formed by a representative cross-section of the organization executive and senior management. It not only directs the program office, but also manages the governance process of the methodology. Hence, any given projects, across the enterprise, most likely, would be sponsored by a member from the board of governors. First and foremost, the board of governors establish the level of excellence for the organization. They determine what good looks like—setting the bar for operational excellence. A board of governors structure should be a formal organizational committee, with a clearly identified director. Membership to the board of governors should be determined by the main organizational position the individual fulfills. Delegates or representative membership should be avoided, and the executive organizational leaders should be expected to actively participate in this board. This board provides the strategic linkage necessary to ensure a strong matrix organization structure that supports project management. The board also provides critical coaching and mentoring to the practitioners. In turn, these management leaders need to walk the talk. They should adopt the standard approach and integrate the summary project performance information in their operations reviews. Also, they should work with the program office to establish escalation guidelines for project team members to elevate issues. The board of governors sets and upholds the standards for operational excellence through the governance process.

Establish the Governance Process The governance process encompasses several aspects: the chartering of project and balancing the project portfolio; the oversight of project phasegate reviews; the overall cultivation of the professionalism and career advancement of the resources; and the steering committee of the program office. Project Selection and Project Portfolio Management The board of governors first must define the project selection criteria used to prioritize and select the organization’s active project portfolio. The criteria should reflect the current organizational needs and strategic direction, as well as available funding and resources. This criteria should be evaluated at least once a year, immediately following or in parallel with the annual planning process. Moreover, the board must establish the project discontinuance criteria used to evaluate project healthiness and to determine when to stop investment and kill a project.

Strategic Dire ction—How to Get Started

The board of governors would convene on a periodic basis to consider, prioritize, select, and charter projects. The meeting frequency may be calendar-driven, or by the initiation of a new project proposal or the closeout of a project. The activation of a project should culminate with the project sponsor publishing and distributing the project charter that describes the project scope and its core team members. This project charter should be a standard template within the methodology’s toolset.

Project Phase-Gate Reviews The project sponsor should conduct project phase-gate reviews primarily to ensure that the project remains within scope, provide any direction and information, approve deliverables, and remove any barriers. The sponsor helps to maintain the project boundaries, and prevent scope creep. The sponsor provides strategic direction, sets expectations for the current and subsequent reviews, and gives any relevant information linking the project’s mission to the organization’s overall direction and goals. The sponsor determines whether the project deliverables have met the requirements. The sponsor resolves or eliminates any issues escalated to him/her. At the close of the review, the sponsor should communicate relevant status to colleagues and key management stakeholders to continue to promote support and interest in the project. The board of governors should define a dashboard, or clear set of metrics, as part of the standard methodology toolset, to monitor each project and provided in both the phase-gate reviews and in interim status communications. The dashboard concept rolls up critical summary project information/performance metrics necessary to manage the business and conduct reviews. The dashboard may be a single or set of active templates used throughout the organization. The benefits of such standard project review templates include the following: • From project-to-project, no rework to communicate important information. • Standard data presentation across projects of key business performance indicators to quickly identify trends or extremes. Key project and business indicators might include: • Project schedule (on or off plan). • Project budget (on or off plan). • Completed project deliverables (since last date issued). • Planned deliverables for next reporting cycle. • Issues/concerns to be escalated to sponsor.

823

824

Building Strength via Communitie s of Practice and Proje ct Management

• Project business case/opportunity: impacted revenues and gross margin as a result of project activities. • Risk management: key areas of status and contingency plans (including customer/client, project scope, project resources, schedule, and budget). • Customer acceptance criteria and customer satisfaction. • Quality Assurance metrics/performance. • Project charter and status report identifiers: project name (and/or number), project sponsor, project manager, expected project close date, status report date. See Also “A Process for Product Development,” by Bill Jewett in Part III, p. 887 for additional information on the phase-gate structure and governance process works within an offering development and launch preparation project.

Resource Guidance In support of resource guidance, the board of directors monitor and promote the practitioners’ progress. These individuals make up the CoP; they are the Six Sigma and PM practitioners and program office members. It guides the resources in both individual project performance, certification, and career progression. In conjunction with Human Resources, the board of governors should determine the criteria for operational excellence and internal certification, and for career advancement.

Credential the Practitioners To achieve a minimum level of performance standards, certification levels should be defined to cover a percentage, if not all of the organization. Certification could include the PMI project management certification and/or a Six Sigma certification from a reputable firm such as ASQ or similar consultancy. If the organization develops a custom methodology that integrated one or both of the project management or Six Sigma disciplines, then it should consider certifying the project members in its own method. Such internal certification may be achieved by an individual being interviewed by (or paneling with) a combination of people comprised of mentors, members from a board of governors, the program office, and/or an industry consultant. This certification panel could use a combination of interviewing and project observation techniques to certify the practitioner.

Summar y

Those practitioners wishing to advance their career in either discipline should be encouraged to not only acquire the appropriate certification, but also join ASQ and/or PMI at the national and local chapter levels.

Summary A standard approach or methodology built on the foundation of best practices and knowledge provides a good start to building operational excellence. Therein, an organization’s approach should embrace the rich and robust data-driven fundamentals of Six Sigma, and the rigor and thoroughness of the project management principles to run the project. However, this is only a start to realizing operational excellence. Operational excellence is reached when the workforce consistently achieves the professional’s standards of work and instinctively knows and uses this methodology and toolset. Building and strengthening a discipline is best achieved through communities of practice. A CoP fosters continuous learning and professionalism, striving to reach and surpass that bar of excellence. The success of implementing a standard methodology and building a community of practice both depend on several key enablers, such as: • Incorporating the new approach and process into the organization’s governance processes. • Ensuring the evergreen nature of the new approach and toolset. • Utilizing the organization’s learning resources to provide content and skill training on the new discipline(s). • Creating both informal and formal mentoring for both disciplines, which is project-based to support the work effort. • Gain proactive sponsorship at the executive and senior leadership level. • Establishing a program office to provide infrastructure support to the community of practice. A small team of subject matter expertise should be assembled to provide mentoring in both disciplines and training. Deliverables may include newsletters, follow-on forums, and other vehicles that promote sharing of information, as well as a change control process to sustain the evergreen nature of the methodology and its toolset.

825

826

Building Strength via Communitie s of Practice and Proje ct Management

Additional References 1. Kerzner, Harold. In Search of Excellence in Project Management. New York: Van Nostrand Reinhold, 1998. ISBN: 0-442-02706-0. 2. Kerzner, Harold. Project Management, A Systems Approach to Planning, Scheduling and Controlling. New York: John Wiley & Sons, Inc., 1998. ISBN: 0-471-2835-7. 3. Project Management Institute. Four Campus Blvd, Newtown Square, PA 19073. Its web site is www.pmi.org. 4. Wenger, Etienne, Richard McDermott, and William M. Snyder. Cultivating Communities of Practice. Massachusetts: Harvard Business School Publishing, 2002. ISBN 1-57851-330-8.

Complex Organizational Change Through Discovery-based Learning By Donna Burnette and David Hutchens Organizations that pursue learning around demanding initiatives are finding success through a counter-intuitive approach:they’re handing the learning over to the learner. In an era characterized by turbulent organizational change, complex initiatives like Six Sigma place extraordinary demands on an organization and on the capacity of its people for learning. A Six Sigma deployment, for example, requires that people develop new capabilities simultaneously in multiple and challenging domains, from the scientific (as in statistical modeling)to the relational (as in team development). Though many organizations continue to equip their systems for change using traditional learning methods that are didactic and “expert driven,” more and more organizations are embracing a “discovery” based model of learning. Steeped in the rapidly emerging field of adult learning theory, discovery based learning is proving to be effective in equipping learners with new models and skill practice. The method shifts the emphasis from the facilitator or expert, and instead taps the learner’s own accumulated experiences and expertise. Importantly, the method is highly experiential and affords learners an opportunity to practice, test, and challenge the new models while also making the critical bridge of illustrating “what it looks like in my world.” Discovery-based learning allows the facilitator or designer to present realworld information in the form of provocative questions, scenarios, role-plays, and other “mirrored realities.” These may take the form of rather passive probabilistic software modeling, such as Monte Carlo simulation, or more interactive games and models to attempt to mimic or mirror a system or behavior. Furthermore, the discovery approach frequently involves the application of narrative, metaphor, images, or even methodologies traditionally associated with the disciplines of advertising and marketing. The developers of the Six Sigma quality process certainly know their way around a metaphor. By designating the expertise of its practitioners using a karate-inspired “belt” system, they acknowledge what is quickly evident to new students of the method: This is nuanced and disciplined stuff. And for many organizations, the learning curve is a steep one and about as comfortable as breaking a block of granite with your forehead.

827

828

Complex Organizational Change Through Discover y-based Learning

Like all methodologies that have achieved a critical mass of popularity, Six Sigma is a frequent subject in organizational horror stories, and no organization wants to become the next case study of how just how badly things can go wrong. For many companies, early stages of Six Sigma deployment often involve a dense “ramp up” period. Becoming a black belt in any of the methodologies, for example, usually requires 16 full days worth of content applied to one or more real work projects. Evoking feelings of subject-matter overload that many haven’t felt since graduate school, the learning agenda typically includes methodology requirements, tool task deliverables, and the governance process. Don’t be surprised if there are additional modules related to dense technical topics such as business statistics, graphical techniques, process and operational management, and also human-oriented “soft” topics such as leadership, team dynamics, project management, and change management. Expect to eat lunch at your work station. The sheer complexity of the subject matter is not the only obstacle of Six Sigma training. The breadth of the menu assures that learners will disengage to some degree around certain topics that are not aligned with their strengths or interests. Additionally, Six Sigma is typically explored using a traditional, didactic approach that reinforces conceptual awareness through a lecture-and-test format but frequently fails to capture the “stickier” kind of learning and mastery that is built through experience.

An Intuitive Approach As the leadership and management landscape within organizations becomes more complex, more and more learning officers are recognizing that a different approach to organizational learning is required. Learning initiatives in this landscape typically have several requisites: They must engage learners personally—and even emotionally—in the conceptual content; they must connect it to their contexts in very immediate ways through mirrored reality scenarios; they must invite the learner to draw from their own expertise to “co-create” the content; and they must provide opportunity to test the content, reflect upon it, and explore it in real time. These approaches are often referred to as “discovery-based,” reflecting a post-modern world in which understanding is created by the individual learner and then tested against a framework or a model for further understanding. (In the “old school,” the framework itself was the takeaway of the learning initiative; here, it is merely a structure to support the process of discovery.)

An Intuitive Approach

If this all sounds awfully “me” centered, that too is a reflection of the post-modern learning context. And when you think about it, this perspective is critical for a Six Sigma project, which moves or stalls depending upon the purposeful actions of each individual employee. Furthermore, the Six Sigma methodology, despite its statistical birthright, is ultimately a human activity that depends upon team collaboration, improved communication, and brave leadership. It turns out that this science is, in the end, an art. A successful deployment of Six Sigma across an organization utilizes the most effective approaches to bring people up to speed and sustain the new initiative. Along with a Communities of Practice approach, discovery-based learning shortens the time to understand, to internalize, and to apply new knowledge, methods, tools and techniques and to incorporate them into one’s daily work practice. (See Also, “Building Strength via Communities of Practice and Project Management,” p. 799, and “The Practice of Designing Relationships,” p. 873.)

This Is Only a Test It would be the mother of all missed opportunities to offer only a written definition of a process that begs to be experienced. In that spirit, we invite you to sharpen your #2 pencil and—without peeking at the answers— complete this short quiz: 1. What is the three-word ad slogan of Nike? ________________ 2. What carbonated beverage used to have the simple ad slogan of “It’s The Real Thing”? _____________ 3. What destination calls itself “The Happiest Place on Earth”? __________________ 4. You are in good hands with ______________________________________. 5. You deserve a break today at ____________________________________. 6. What was the last training program you attended, and what were three key things you learned and took away from that program? ____________________________________ Now stop. Pencils down. How did you do? If you are like most people, you had no trouble answer-

829

830

Complex Organizational Change Through Discover y-based Learning

ing the first five questions: “Just Do It,” “Coca-Cola,” “Disney World ,” “Allstate,” and “McDonald’s.” And if you are like most people, that sixth question was a stumper. Now consider: Why do so many of us have easy recollection of advertising slogans, which are of little value to us, and yet are unable to readily access the critical learning that was deemed important by someone within our organizations? An even more pointed question: How can learning and organizational change agents leverage Madison Avenue techniques to speed the change and learning process within organizations? Increasingly, organizational leaders, who have a knack for quickly embracing anything that delivers results, are flocking to discovery-based programs that marry the worlds of adult learning theory and old-fashioned Madison Avenue persuasion. But upon closer consideration, perhaps these two fields really aren’t so different.

Lessons from Advertising Let’s examine the core objectives of advertising more closely. Consider the three goals of advertising, illustrated in Figure 1. As illustrated in Figure 1, an advertisement attempts to capture the attention of its target audience to inform them of a topic—that a product or service exists along with its key features and functionality. Next, an advertisement illustrates in highly personal ways how the offering answers a particular need or desire. Its ultimate objective is to persuade—that is, present a case so that the audience is moved to take action. This is not persuasion of the arm-twisting variety, but rather a more sustainable kind that invites the audience to act upon an internal logic or felt need. Organizational folks like to Figure 1: Three Goals of Advertising call this “buy in.” We should also point out that advertisements aim to be memorable. The best ones elicit a high recall rate among the target audience to remind them of its main message and their call to action. The objectives of informing, reminding, and persuading are equally applicable to a commercial for potato chips, an organizational Six Sigma

Learning = Action

deployment, or any initiative that seeks to equip a population with new awarenesses, tools, and best practices for working.

Learning = Action Let’s take a closer look at the process that moves people to take action. Here again we will view instructional design through a marketing lens and the classic “AIDA” model. AIDA represents the internal process by which an audience moves from knowing nothing about your initiative to that prized state of “buy in.” Specifically, the AIDA acronym stands for attention, interest, desire and action, as shown in Figure 2. (Start at the bottom of the pyramid and read up.) The AIDA model suggests that the most effective messages begin with attention; generate interest; develop desire; and initiate action. As you have experienced yourself in watching those attention-getting Super Bowl ads, this is accomplished through a post-modern toolset that includes stories, visuals, emotional hooks, humor, sensory stimuli, gaming techniques, varied media formats, surprise, a savvy for the cultural vernacular, and so on.

Figure 2: The AIDA Model to Create an Advertising Message

Okay, so perhaps we overstate our case. The differences between a discovery learning program and a catchy jingle for floor wax have as many significant differences as they do similarities. The point is, organizational change (and it is our belief that the purpose of learning is ultimately to produce change) requires a change in human perception. Where sophisticated organizational communications and learning initiatives differ from advertising is that they are anchored in solid instructional design models and processes. These learning design activities include contextual analysis, program design and development, strategic implementation, and a feedback or evaluation mechanism for future improvement.

Treating Learners Like Adults Even more fundamental to discovery learning approaches are the

831

832

Complex Organizational Change Through Discover y-based Learning

assumptions of adult learning theory. The design elements build on the work espoused by the adult education expert Malcolm Knowles (1913–1997) who popularized the concept of andragogy—the importance of self-directed learning for adults. Adult learners are shown to share some key characteristics that distinguish them from the world of childhood learning (or pedagogy): • Adults come from diverse backgrounds with their own preexisting

knowledge and expertise. • They learn in different ways and at different speeds. • They learn best when they are motivated to learn and when they can

help direct their own learning. • They learn from each other. Learning is a cooperative and collabora-

tive process. • Learning takes place in the realms of intuition and emotion as much

as it does in the realm of reason and rationality. • Adults learn best when they are involved in diagnosing, planning,

implementing and evaluating their own learning. • Adult learners have a need to be self-directing. • Readiness for learning increases when there is a specific need to know. • Life’s reservoir of experience is a primary learning resource; the life

experiences of others enrich the learning process. • Adult learners have an inherent need for immediacy of application. • Adults respond best to learning when they are internally motivated

to learn. • Learners need a comfortable learning environment.

There are plenty of organizations that are resistant to elements of play, gaming, humor, and other learning approaches. But it is surprising how quickly those biases dissolve as the discovery approach engages learners with an almost magical effectiveness. It is an invigorating thing indeed for learners to take the lead in posing the questions, solving the conundrums, and making the applications to the tough challenges that await them back in the “real world.” In other words, they are liberated by a methodology that assumes that education is a process and not a set of facts. Benchmark discovery learning programs can be either off-the-shelf or customized offerings. A custom discovery-based Six Sigma learning pro-

Treating Learners Like Adults

gram simulates your organization’s specific business model by building a program around a set of specific simulations that emulate your work environment. Strive to deploy a program that embraces as many of the adult-based learning tools and techniques as possible. A program that uses stories—in the form of a Six Sigma project scenario—emulates real situations utilizing a range of media and formats such as board games, role plays, and computer technology. Moreover, visuals are critical to create the simulation. The program should use humor, some surprise, gaming techniques, and varied media to attract and engage the learner. And it includes exercises that force the learner to apply the key learnings about the given topic to their own real world companies. Florida-based learning company Solutions House is an example of a firm that offers both standard programs with topics applicable to a Six Sigma deployment and customized services. For example, the firm addresses a similarly complex challenge—the quest for financial literacy at all levels of an organization—using a “board game” approach called Full Throttle . The game puts learners in the driver’s seat (literally!) of a fictional motorcycle company. In the process, the learner-as-CEO is empowered to make strategic decisions, try alternative approaches, and lead the imaginary enterprise to either riches or ruin. Of course, the power of the experience is neither in the dice rolling nor in the winning or the losing. It is in the process of participating in constructing knowledge with their colleagues in an environment that encourages experimentation, risk, and even play. Offbeat? Perhaps. Successful? Certainly. The Florida-based firm is serving an increasing roster of blue-chip organizations that are seeking a “different way to learn” as they face complex organizational initiatives whether it is as “hard” as finance or as “soft” as creating trust in an organization. As liberating as the discovery approach can be for learners, it has even more dramatic implications for leaders and managers. Those who shepherd high-stakes initiatives such as Six Sigma can’t be faulted for wanting to tightly manage the learning piece of the work. But witness the swelling ranks of those who have applied discovery learning and landed safely on the other side. They will testify with near-religious zeal that they used to relish in their role as “the expert” but that something amazing happened when they invited the learner into the change process: They discovered that the learner possessed the will to change all along.

833

834

Complex Organizational Change Through Discover y-based Learning

About the Authors Donna K. Burnette is a co-founder and CEO of Solutions House Inc., located at Gateway Executive Center, 8601 4th Street North, Suite 212, Saint Petersburg, Florida 33702. Its web site can be found at: www.solutionshouse.com. She has written several feature articles in trade publications such as The New Face of the Project Team Member, which was published in the November 2000 issue of PM Network. She contributes to her profession through the presentations she delivers at local and national conferences, a few of which include • Leadership for Womenat the National Executive Women’s Interna-

tional Conference • Business Acumen for Training Professionals at Training 2001 (sponsored

by Training Magazine) • Advertise to Your Employees! Techniques for Energizing Organizational

Change at Training 2001 and the HR Champions’ Summit 2001 (sponsored by Linkage, Inc.) • The Power of Discovery Learning at Training 1999 (sponsored by Train-

ing Magazine) Ms. Burnette can be contacted by phone at (727) 568-0021 or (866) 5252130, extension 912 or by email at [email protected]. David Hutchens is a writer who has crafted messages for some of the world’s greatest organizations, including IBM, The Coca-Cola Company, General Electric, Wal-Mart, and many others. His series of books known as the Learning Fables includes the perennial favorite “Outlearning the Wolves.” The books have been translated into more than a dozen languages and have sold over a quarter-million copies. Learn more at www.DavidHutchens.com.

Lean Six Sigma for Fast Track Commercialization High Risk-High Reward, Rapid Commercialization: PROCEED WITH CAUTION! By C.M. Creveling This article defines the conditions and requirements for conducting a product commercialization project faster than your normal phase-gate process would allow. The key up-front requirement is a well-executed technology development process that transfers safe and mature new technology into the candidate Fast Track project. You can’t pull off a Fast Track project with immature technology. The second key requirement is that you know exactly what tasks and enabling tools, methods, and best practices you are electing to apply later after product launch—this helps define your risk due to postponing some key tasks right up front as you commit to your planned flow of work. The third key requirement is that you have a pre-existing expertise in applying DMAIC Six Sigma methods to problematic designs that have made their way into production, including the tasks you postponed in the previous requirement. You are breaking some established rules within your lean phase-gate process by design. The big difference is that you are doing so honestly so that everyone accountable for the project knows and recognizes the shortcuts that are being “designed” to control the risk as much as possible. It takes both courage and very capable people to pull this off. Just as expert trick skiers break fundamental rules to accomplish their tricks, so do the leaders and functional practitioners of a Fast Track project. Trick skiers are so good at their craft that they can take risks that ordinary skiers simply cannot. You have to be really good at technology development, project planning, and DMAIC problem solving to justify even trying one of these projects. You have to know exactly what DFSS methods are going to be withheld and transferred to the DMAIC team—to be applied later for the sake of time. You are going to intentionally launch an immature product. If you do it skillfully, you will gain early market share and do minimal damage to your brand—a risky but sometimes useful way to grow!

Six Sigma Applications for Fast Track Commercialization Projects Sometimes there is an opportunity to really cash in on a set of circumstances that make it worth the risk to push a project through your phasegate process at a very high rate. You might call this “rushing” a project. 835

836

Lean Six Sigma for Fast Track Commercialization

Most projects should be “hurried” along and not rushed. When we use the word “lean” in the context of product commercialization, we are referring to a project that is hurried along through the proper design of a balanced flow of value-adding tasks that focus on fulfilling requirements that are directly tied to recently gathered customer needs. A Fast Track project is really beyond what you would call “lean.” The main characteristic of a Fast Track project is that some value-adding work is not going to get done fully and completely—and some not at all until a post-launch DMAIC Six Sigma Team completes these tasks. We are making the assumption, and we could be wrong, that you already have put in place some form of Six Sigma competence in the DMAIC problem-solving process tradition. To do a Fast Track project, you have to have a history of success in applying DMAIC Six Sigma to fixing weak designs that are in production. Some Fast Track projects are going to violate other lean principles because they are not going to be driven by direct customer data as such but rather by a new breakthrough in technology that drives such value that customers will simply be blown away by what the product offers. These are cases where customers did not know they wanted or could use such a paradigm-breaking innovation. This is “lead user” turf: when a new innovation breaks down old paradigms. In the case where you want to absolutely dominate the market for this new opportunity, justification can be made for a Fast Track project. This article contains a carefully considered opinion on how to do this and cause the least amount of damage to your brand and reputation in the market. This is high risk-high reward territory. It is rational to do some projects this way—but not many. If your business has developed a habit where you have to rush all new products to market, you need to make a significant effort to control how much of this you actually do. Your product portfolio renewal process needs to be structured to better balance risk profiles across your growth projects. We get asked about the design of tool-taskdeliverable groups that one can use to minimize risk while a project is put on a fast track to launch. We will explain how to use a combination of DMAIC methods in the post-launch environment and DFSS during CDOV product commercialization phases to conduct a Fast Track project. CDOV stands for four generic phases one can use to define the work that is done to commercialize a product (Concept development, Design development under nominal conditions, Optimize the product for stressful conditions, and Verify the product and production system capability against requirements). The IDEA process is a generic set of phases to renew a product portfolio. IDEA refers to Identify opportunities, Define requirements and product portfolio architectural alternatives, Evaluate the alternative portfolio mixes, and Activate the ranked and prioritized portfolio that best meets your growth goals. The IIDOV process is a generic set of phases to

DMAIC Six Sigma Proje ct Capability to Support a Fast Track Proje ct

develop safe and mature technology for transfer into product commercialization. IIDOV refers to Invent and Innovate new technology concepts, Develop stable and tunable technologies, Optimize the technology for robustness when applied across ranges of applications, and Verify the robust and tunable technology is able to be integrated into a product design. All three of these methods (CDOV, IDEA, and IIDOV) assure you have the right strategic environment and data to justify a Fast Track Commercialization project. See Also “Design for Six Sigma (DFSS),” in Part I, p. 45, and “Six Sigma for Marketing (SSFM),” in Part I, p. 67.

DMAIC Six Sigma Project Capability to Support a Fast Track Project In part, our answer to this fast track issue of being able to rapidly correct the problems you elected to create by not doing certain tasks is grounded in your expertise in classic DMAIC Six Sigma and how fast and well you can apply DMAIC tools (this includes those DFSS tools you elected NOT to use during commercialization) to clean up unfinished tasks in a post-launch context. If you are really good at DMAIC/DFSS Six Sigma-enhanced projects and have the dedicated resources to conduct work this way, then you have what we consider a foundation of skills to justify accepting the risk of doing Fast Track projects. Core DMAIC Six Sigma competencies include the following: 1. A hierarchy of Master Black Belts, Black Belts, and Green Belts that work well together as a team. Do you have enough of them to conduct unfinished design tasks whose problems will be fully documented in your FRACAS (Failure Report And Corrective Action System) or Problem Report database? You may choose to eliminate some tasks and happily get away with it. More likely, you will trim tasks and get clobbered with problems—some will be predicable and others will be complete surprises. These DMAIC experts should be told exactly what tasks and tools you are not going to have time to apply so they can get ready to do them after the launch is executed “prematurely by design.” Remember that there is a compounded effect because these people are going to have to unravel the interactive sensitivities and variation transmissions that your design team rushed past. Think “pit crew” or SWAT team levels of competence and team work. They are going to get a bit of a mess handed to them. Their project will have to deal with two key areas: 1. finish tasks that were known to be “postponed by design”; 2. conduct tasks that are surprises because things that were not anticipated actually happened and now you have to deal with them. The more you can document what was postponed, the better chance they will

837

838

Lean Six Sigma for Fast Track Commercialization

have of quickly diagnosing what the shortcomings are and what tasks and tools are needed to close the gaps. 2. A track record of successfully completed projects that have fixed problems similar to the ones you are about to intentionally create. Don’t fool yourselves—Fast Track Commercialization projects create a rich set of problems. You just happen to be choosing which ones you are going to let happen and then fix by telling the DMAIC Team what the gaps are likely to be on a proactive basis. You are going to “design” your problems. 3. Training on Critical Parameter Management (CPM). They are going to need all the data you can give them to finish optimizing the product. The DMAIC team should be handed all CPM data including the Y = f(x) relationships that are under-defined. They will literally be finishing the Critical Parameter database. 4. Robust design, tolerance design, and system integration and balancing competencies (traditional DFSS skills) are a must in your post-launch DMAIC troubleshooting teams. They are going to have to complete these tasks that were postponed or partially completed during commercialization. Some DMAIC training programs do not address these tool sets. You may need to transfer a few DFSS Black Belts into the post-launch DMAIC team to lead these tasks. Eventually you will need to cross-train your DMAIC and DFSS teams so they share skills in these crucial areas.

Technology Development Capability to Support Fast Track Projects If you are really good at developing stable, tunable, and robust technologies, then you have the second foundational skill to justify rushing a select few commercialization projects. This is what Chapter 5, “Strategic Research and Technology Development Process,”from Six Sigma for Technical Processes describes. If you are not good at developing testability, tenability, and robustness, then rushing will unfortunately get you in the end—in the post-launch environment where the DMAIC team gets the mess. If the project you want to put on a fast track is dependent on a new technology, then you must pay your dues on it by not rushing the new technology through your R and TD process (IIDOV, as mentioned earlier as an appropriately rigorous Technology Development Process). We don’t see any way to rush both R and TD as well as commercialization. You must slow one down for the sake of the other. It is always better to get the technology done properly then take the risk of rushing the product in which it is embedded. The strategy we are trying to teach you is to take your time during portfolio renewal and R and TD—then you can rip

Concept Phase Risk Profile s and Tool-Task Re commendations

through commercialization at a pretty good clip depending on competitive pressures and market opportunities to either rush or just hurry a product to market. If you are weak at portfolio management and technology development, you have to depend upon your post-launch DMAIC skills alone.

DFSS Skills and Product Commercialization Capability to Support Fast Track Projects Let’s walk through the phases of the CDOV Product Commercialization process and look at our options for tool-task combinations that are a bare minimum of things that will protect you as you rush a project. As previously mentioned, CDOV stands for four phases used to control risk during product commercialization: Concept Phase, Design Phase, Optimization Phase, and Verify Phase. We will use these four phase names as a guide for what detailed recommendations we are making for conducting a Fast Track project. As a reminder, we are going to focus on critical needs, requirements, and functions that are “NUD” = New, Unique, and Difficult. They are the value-added elements that make the project worth doing. The opposite of “NUD” requirements are requirements that are Easy, Common, and Old—referred to as “ECO,” or non-critical but necessary requirements that must also be fulfilled during the project.

Concept Phase Risk Profiles and Tool-Task Recommendations Rushed projects have the following risk characteristics in the Concept Phase: 1. Low Voice Of the Customer (VOC) needs content—no time to dig down to get fresh VOC details. We have opinions, filtered input from marketing and sales professionals, trends we have picked up on, and secondary marketing data that is mainly historical in nature. At a bare minimum, use KJ and QFD methods to structure and rank the data you have. See Also “KJ Analysis” in Part II, p. 375 and “Quality Function Deployment (QFD)” in Part II, p. 543 2. New technology critical parameter data—not well linked to the current VOC data because it (VOC) is very limited. We may not have a good idea of performance tuning ranges, so the tuning range for the new technology may be off course—but hopefully not. If our new technologies are ranged properly, this risk will be low. Our technologies must be proven, with capability indices and transfer functions that contain tuning capability, fully scalable to whatever target performance we ultimately need to launch the product for a given application. Tunable and robust technology make Fast Tracking projects viable.

839

840

Lean Six Sigma for Fast Track Commercialization

3. Competitive benchmarking will be low and likely anecdotal. This will present two areas of risk: missing technical requirement detail that would have come from deeper knowledge of what you are up against; and improper ranking and prioritization of technical requirements because you are not sure what matters most. At a bare minimum, define the key attributes and performance characteristics for the best competitive product you are up against. 4. Technical requirements will be ambiguous and changing as we move through the commercialization process. Targets and tolerances and their allocation down through the product’s hierarchy will be changing because we will be discovering them as we go rather than having a stable set from the beginning, as we would in a non-rushed Concept Phase. At a bare minimum, apply QFD to the requirements you have. It will be easier to change if what you have is at least well documented. 5. Concept generation will be truncated to produce fewer alternatives than we would normally consider. Maybe we only have one concept, and we bet the farm on it. If the technology enabling the single concept is stable, tunable, and robust, then you have a better chance of seeing this one concept survive and make it through to an acceptable, but less than perfect, launch. At a bare minimum, try to structure two concepts from which you can hybridize a final product concept. 6. Concept evaluation and selection will be limited. Concept evaluation criteria will be anemic due to limited VOC data and the constraints that are placed on QFD and the translation of customer needs into stable technical requirements. Your concepts will suffer from added vulnerability to competitive threats and a lack of true superiority compared to alternative concepts. At a bare minimum, use the NUD classification approach from your QFD work to stabilize the high-priority requirements. 7. Very limited system-level modeling, requirement budgeting and allocation, and early risk assessment of the system will be the norm. At a bare minimum, use QFD NUD requirement flow down traces to define NUD Functions for the product. With this in hand, conduct abbreviated FMEA (Failure Modes and Effects Analysis) on the NUD (New, Unique, and Difficult) functions to get a preliminary document of high-risk areas.

Three Key Marketing Tasks The three key marketing tasks needed during the Concept Phase include:

Summar y

1. Verify the surrogate NUD customer needs by reviewing them with a few key customers for correction, revision, and validation. 2. Verify NUD technical requirements by reviewing them with a few key customers for correction, revision, and validation. 3. Verify final product concept by reviewing it with a few key customers for correction, revision, and validation. All of this work is done to quickly check to see if there are any major misses. It is just the bare minimum to be sure, on a “check back with the customer” basis. Normally we would take more time and really dig in with a broader set of diverse customers to gather the data from them on a proactive basis. Here, we used internal opinion and past experience to generate surrogate VOC data. You can justifiably claim you are customer driven with this approach—it’s just that the cart is in front of the horse!

Summary What Must Be Done at a Bare Minimum During the Concept Phase? The minimum activities required during the Concept Phase include: • Document fresh needs data from your marketing and sales experts

that gives you their best field experience and judgment on customer needs. You are basically using them as a surrogate for the real VOC. Apply KJ to structure, rank, and prioritize this data about the surrogate customer needs. Check back with a few key customers to be sure you are reasonably representing their latest needs. • Document the characteristics of the best product in the market that

you know from your internal opinion. What is going to present the biggest threat to your new product? This is surrogate benchmarking with no actual tear-downs or in-depth analysis of various competitive products. • Use KJ and QFD to process your limited data sets to structure, rank,

and prioritize the customer needs for translation into a documented set of product technical requirements. Check back with the same few key customers to be sure the technical requirements are in basic alignment with the NUD customer needs. • Document at least two alternative concepts to compare against the

one best product you are going up against. Use the NUD requirements from the House of Quality to help establish concept evaluation and selection criteria.

841

842

Lean Six Sigma for Fast Track Commercialization • Select the best of the two concepts and refine the one concept you

have to establish as the most competitive concept possible to go up against your competitive benchmark. If you can, blend synergistic attributes from both concepts to enable one superior concept. Check back again with the few key customers to be sure they are in alignment with where this product is headed. • Conduct an FMEA on the NUD functions within your selected prod-

uct concept to characterize risk.

Key Tools, Methods, and Best Practices to Consider as You Control Risk The Concept Phase contains a core set of tools to utilize; each of these tools is detailed in a separate article in Part II: • KJ Method for Structuring, Ranking, and Prioritizing Customer Needs (See Also “KJ Analysis,” in Part II, p. 375) • QFD and the House of Quality (See Also “Quality Function Deployment (QFD),” in Part II, p. 543) • NUD Screening of Customer Needs and Technical Requirements (See Also the “NUD versus ECO” section within the “KJ Analysis,” Part II, p. 375) • Pugh Concept Evaluation and Selection Process (See Also “Pugh Concept Evaluation,” in Part II, p. 534) • Design FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” in Part II, p. 287)

Design Phase Risk Profiles and Tool-Task Recommendations Rushed projects have the following risk characteristics in the Design Phase: • Stability of functional performance down inside the subsystems and

subassemblies absolutely must be proven and documented. This is especially true of any influences new technologies have on leveraged designs. New materials are particularly important if they are used in existing hardware form and fit structures. Here the documentation that proves stability for each critical sub-level function is the SPC chart known as an Individuals and Moving Range Chart (IMR Chart). • The next level of risk is in the area of functional performance capa-

bility under nominal or best-case conditions. So here we must insist on documenting Cp and Cpk values for critical sub-level functions. The focus is on non-stressed performance—what we call nominal conditions where there are only random sources of non-induced variation active in the data. Here, capability studies are required for

Summar y

all New, Unique, and Difficult functions that are the basis of customer satisfaction in the sub-level designs. • The final required documentation for this phase is tunability. What

one critical adjustment parameter (CAP) is proven to have a statistically significant effect on shifting the mean of each critical functional response in each sub-level design? If we were not rushing, we would take the time to define numerous significant CAPs, not just one dominant CAP. We are NOT building a comprehensive Y as a function of multiple Xs here—just one, simple regression model where we spend only enough time to nail down one critical X for tuning our mean performance onto the desired target.

What Must Be Done at a Bare Minimum in the Design Phase? Document base-line performance stability and capability at the subsystem and subassembly level. If the designs do not possess stability and acceptable capability under non-stressful conditions, then you do not have the building blocks needed to integrate a product in the Optimize phase. The design’s critical Ys should also be tunable to hit the targets defined in the product requirements document. If you cannot scale functional performance at the sub-levels, then you cannot adjust and balance functional performance at the system level. Without this, you simply do not have a viable product by any standard of measure. Key Tools, Methods, and Best Practices to Consider as You Control Risk • Statistical Process Control for Stability of Performance (I-MR Charts) (See Also “Control Charts—7QC Tool,” in Part II, p. 204) • Regression Analysis for documenting key Mean Shifting Parameters (CAPs) (See Also “Regression Analysis,” in Part II, p. 571) • Capability Studies under nominal conditions (See Also “Process Capability Analysis,” in Part II, p. 486) • Design FMEA to update risk for the NUD functions that made the design worth doing from a customer’s perspective (See Also “Failure Modes and Effects Analysis (FMEA),” in Part II, p. 287)

Optimize Phase Risk Profiles and Tool-Task Recommendations Rushed projects have the following risk characteristics in the Optimize Phase: • Here we will do just one robustness optimization screening experi-

ment per sub-level design IF it possesses a NUD function that is critical to customer satisfaction for this new product. We would

843

844

Lean Six Sigma for Fast Track Commercialization

normally prefer to iterate through a sequence of robustness experiments for all our major subsystems and subassemblies—but a quick screen for critical Xs that have a stabilizing influence on the standard deviation of each critical Y under non-random, induced stress conditions is a must. If we don’t do this at a minimum, then we are definitely going to get slowed down during system integration—and we just cannot afford to let that happen. • Once we have the NUD sub-functions somewhat robust to stressful

sources of variation, then we can take the risk of rapid system integration. Here we will run one nominal, non-stressed performance test and document Cp and Cpk for the critical Ys at the sub-level AND system level. Then we will run just one stress screen using a designed experiment to find out where the system integration sensitivities reside. We tune the sub-level design parameters to try and balance performance under stressed conditions—thus minimizing “k” shifting in the mean performance of the sub-level Ys and consequently stabilizing the variability of the system level Ys that are critical to the customer and our business case.

What Must Be Done at a Bare Minimum in the Optimize Phase? Rushed projects should include, at minimum, the following activities during the Optimize Phase: • Conduct limited stress tests on the customer-critical subsystems and

subassemblies to flush out sensitivities that can be rapidly identified and fixed. The best thing you can do is give the system as robust subsystems and subassemblies as you can afford—in our opinion, at least one round of robustness screening. This helps make integration go faster and smoother. The less work you do here, the more trouble you will have during system integration and final product verification. Do as much robustness optimization as you can. Let the DMAIC teams focus on the final tuning and optimization of mean performance—they are better at that kind of work. Your goal is to tame the standard deviation during intentionally induced stress evaluations. • Conduct a limited system integration stress test to identify sensitivi-

ties that can be rapidly identified and balanced. Here you are making a first cut at preventing escaping integration problems. You won’t get them all, but you’ll get the really bad ones out of the way depending upon how much stress you choose to induce. • Document Failure modes for the latest data so the DMAIC team can

see emerging risk trends and tasks you did not complete. They have

Summar y

to structure their project plan to fill your gaps. • Conduct Process FMEA to identify risk in the relationships between

design functions and process functions that make the parts and materials that will affect the design performance, fit, and finish Cps and Cpks.

Key Tools, Methods, and Best Practices to Consider as You Control Risk • Robust Design of Experiments (See Also “Design for Six Sigma (DFSS),” in Part I, p. 45, and “Design of Experiment (DOE),” in Part II, p. 250) • Post-robustness SPC charts (See Also “Control Charts—7QC Tool,” in Part II, p. 217) • Capability Studies before and after robustness stress evaluations (Cp and Cpk indices) (See Also “Process Capability Analysis,” in Part II, p. 486) • Design and Process FMEAs to link process risk to design performance risk (See Also “Failure Modes and Effects Analysis (FMEA),” in Part II, p. 287)

Verify Phase Risk Profiles and Tool-Task Recommendations Rushed projects have the following risk characteristics in the Verify Phase: • NUD Tolerances and nominal set points down through the inte-

grated system have to be finalized. Here we use tolerance analysis aided by Monte Carlo simulations and designed experiments to balance the integrated system—with a constrained focus on the New, Unique, or Difficult functions that matter most to customers. Careful identification of just those set points at the sub-level design arena and system output should be targeted for focused tolerance optimization. Let the rest of the set points be specified for what the supply chain can reasonably hold at the most reasonable price. We will let the post-launch team clean up the final specifications that have to be balanced for cost vs. performance. • Final reliability evaluations can be run at this time to screen for limi-

tations and focused corrective action. We know postponing a lot of system-level reliability work until this final phase is scary, but we think it is a better trade-off to spend your limited time on robustness development rather than just depressing your selves with premature reliability assessments that do nothing to improve reliability. They chew up precious prototype hardware evaluation time and human resources. You already know that you’re intentionally developing a

845

846

Lean Six Sigma for Fast Track Commercialization

less than ideal product for rapid launch—let the post-launch team apply aggressive reliability growth tasks in harmony with cost vs. performance balancing. If their DMAIC skills are strong, this work will be reasonably quick and straightforward for them vs. robustness development tasks, which are far more specialized to DFSSenhanced commercialization teams.

What Must Be Done at a Bare Minimum in the Verify Phase? Rushed projects should include, at minimum, the following activities during the Verify Phase: • Conduct tolerance balancing analysis and designed experiments to

finalize specifications for the NUD functions that matter the most to customers. • Conduct capability studies at the system level on the product to

identify current, integrated performance. The DMAIC team will need this data to begin their improvement projects on the weak areas across the system. • Conduct reliability assessments. Document Failure modes for the

latest data so the DMAIC team can see emerging risk trends from the tasks you did not complete. They have to structure their project plan to fill your gaps.

Key Tools, Methods, and Best Practices to Consider as You Control Risk • Analytical Tolerance Design (Monte Carlo Simulations for tolerance balancing) (See Also “Monte Carlo Simulations,” in Part II, p. 431, and “Selecting Project Portfolios Using Monte Carlo Simulation and Optimization,” in Part III, p. 921 .) • DOE for integration Sensitivity Analysis (See Also “Design of Experiment (DOE),” in Part II, p. 250) • Reliability Assessment • HALT and HASS Testing (only if you have time) • Capability Studies (See Also “Process Capability Analysis,” in Part II, p. 486) • Design FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” in Part II, p. 287) • Process FMEA (See Also “Failure Modes and Effects Analysis (FMEA),” in Part II, p. 287)

Summar y

Macro-Summary of Fast Track Projects Fast Track projects can lead to helpful additions to sustainable growth. They alone cannot produce sustainability—only a balanced portfolio of new products can accomplish that goal. The non-negotiable, critical few requirements for a Fast Track project include the following. Conceptual Design Phase

• Clear and stable requirements that are derived, as much as possible,

from real customer data. In our case, it is going to be mainly “check back with the customer” oriented. We are using needs that are truly New, Unique, and Difficult (NUD) to drive criticality versus Easy, Common, and Old requirements. • A superior product concept that was hybridized out of at least two

alternatives that were compared against one best-in-class benchmark. The concept evaluation and selection criteria must be based upon the NUD requirements that are traceable back to real customer preferences and needs. • A risk assessment based upon the potential failure modes of the

product, the market, and the project management tasks that are being planned to complete it. Nominal Design Development Phase

• Base-line stability and tunability of all critical (NUD) sub-level

design elements. • Cp and Cpk data for all critical functional responses from the sub-

level designs under nominal, non-stressed conditions. • A risk assessment based upon the potential failure modes of the sub-

level designs, the market, and the project management tasks that are being planned to complete the next phase. Optimization and Integration Phase

• Sub-level design robustness screening to identify sensitivities and

parameters that diminish their effect on the standard deviation of the NUD critical functional responses. • System integration tests under both nominal and stressed condi-

tions—prior to system-level reliability assessment tests. System-level sensitivities are diminished and balanced across the NUD critical functional responses.

847

848

Lean Six Sigma for Fast Track Commercialization Product and Production System Verification Phase

• Sub-level and system-level NUD tolerances are balanced for cost

and performance. • Production tolerances and nominal set points are capable of support-

ing the NUD functional responses within the integrated product. Boiling this all down, you can see that we truncate our tasks to a bare minimum that is totally documented by Critical Parameter Management of just the New, Unique, and Difficult requirements and functions of our Fast Track product. The interaction of the NUD functions and the Easy, Common, and Old (ECO) functions will be dealt with by the post-launch team. If we get the dominant sensitivities under control for the NUD functions, we are going to minimize the damage of poor performance across the items that really make the product value adding to the customer. The remaining sensitivities and reliability issues will have to be cleaned up very quickly using the DMAIC team. Pick, or better yet, design your problems wisely and proactively, be sure your DMAIC team is well aware of their role in this kind of project, and don’t hope for the best—PLAN for it! Proceed with caution…this is not the way you should be commercializing products as a rule—only do this as the exception when you can prove the risk is worth taking from good portfolio management discipline, as outlined in the IDEA process briefly mentioned earlier.

Additional Readings 1. Six Sigma for Technical Processes; Creveling. ISBN 0-13-238232-6. Specifically refer to the following chapters: a. Chapter 5, “Strategic Research and Technology Development Process” b. Chapter 6, “Tactical Product Commercialization Process” 2. Six Sigma for Marketing Processes; Creveling, Hambleton, and McCarthy. ISBN 0-13-199008-X. Specifically refer to Chapter 4.

About the Author Mr. Creveling is currently President of Product Development Systems and Solutions (a full services product development and product lifecycle management consulting firm). Prior to that, he was an independent consultant, DFSS Product Manager, and DFSS Project Manager with Sigma Breakthrough Technologies Inc., where he served as the DFSS Project Manager for 3M, Samsung SDI, Sequa Corp., and Universal Instruments.

About the Author

Mr. Creveling was employed by Eastman Kodak for 17 years as a product development engineer within the Office Imaging Division. He also spent 1 1/2 years as a systems engineer for Heidelberg Digital as a member of the System Engineering Group. During this 18+ year period, he worked in RandD, Product Development/Design/System Engineering, and Manufacturing. Mr. Creveling has five U.S. patents. He was an Asst. Professor at Rochester Institute of Technology for four years, developing and teaching undergraduate and graduate courses in Mechanical Engineering Design, Product and Production System Development, Concept Design, Robust Design, and Tolerance Design. Mr. Creveling is a certified expert in Taguchi Methods. He has lectured, conducted training, and consulted on Product Development Process improvement, Design for Six Sigma Methods, Technology Development for Six Sigma, Critical Parameter Management, Robust Design and Tolerance Design theory, and applications in numerous U.S., European, and Asian locations. His clients include 3M, Merck and Co., Motorola, Samsung, Applied BioSystems, United Technologies, ACIST Medical Systems, Beckton Dickenson, Mine Safety Appliances, Lightstream, Kodak, NASA, Iomega, Xerox, Sequa Corp. (Atlantic Research Corp., MEGTEK, Sequa Can), GE Medical Imaging Systems, Bausch and Lomb, Moore Research, IIMAK, Case–New Holland, Maytag, Cummins, Schick, Purolator, Goulds Pumps, INVENSYS, Shaw Carpet, Heidelberg Digital, Nexpress, StorageTek, Universal Instruments Corporation… among others. He has been a guest lecturer at MIT, where he assisted in the start up of a graduate course in Robust Design within the MS in System Development and Management program. Mr. Creveling co-authored the following texts: • Engineering Methods for Robust Product Design (Addison-Wesley,

1995; ISBN: 0-201-63367-1) • Design for Six Sigma in Technology and Product Development

(Prentice Hall, 2003; ISBN: 0-130-09223-1) • Six Sigma for Marketing Processes (Prentice Hall, 2006;

ISBN: 0-13-199008-X) Mr. Creveling authored the following texts: • Tolerance Design (Addison-Wesley, 1997; ISBN: 0-201-63473-2 ) • Six Sigma for Technical Processes (Prentice Hall, 2006;

ISBN: 0-13-238232-6) Mr. Creveling can be reached by email at [email protected].

849

This page intentionally left blank

Listening to the Customer First-hand; Engineers Too By Bill Jewett

Product development teams need much more than a set of requirements provided to them by marketing or management. Teams need to understand how customers will use the new products and what features and functions will provide the most benefits to them. The costs to customers of owning the products and the aggravations in their malfunctions must be interpreted accurately. How customers perceive value as they compare products from competitors must be understood well. The conclusions from these evaluations contribute to improved requirements for new products. However, there can be a bias against product development teams interacting directly with customers. Marketing and sales people may distrust what engineers will say or how they will interpret what they see and hear, and may be threatened by a perceived invasion of their domain. Management may not believe that the time “off the job” is justified and may be concerned that engineers talking directly with customers can increase the risks of committing the company to product capabilities or schedules that cannot be delivered. Product development teams are composed of creative people developing solutions to customers’ needs and problems. By involving enlightened customers directly with those teams, development projects can have a higher probability of providing the necessary capabilities, quality, and costs in new products, at the right time. This article recommends several tactics by which development teams can be more involved with their customers. It may be difficult for internally-focused organizations to accept some of these tactics. However, a more inclusive approach to involving customers can provide substantial benefits toward the objective of delivering superior value to customers. By doing so earlier, development teams can avoid critical changes later in the process and position their new product to be more successful in the market.

Direct Customer Involvement Is Not Always the Best Approach The focus of this article is on the requirements for your new products and the development of design concepts to satisfy those requirements. Customer engagement by your engineers has clear value when there is a significant potential for unfulfilled needs not to be understood well enough or for the opportunities to provide new benefits not to be identified early enough. However, a direct conversation between engineers and customers is not always the best tactic to obtain information for a development project. For example, it may be useful to determine acceptable prices, but not 851

852

Listening to the Customer First-hand; Engineers Too

to forecast sales volumes or market share. Broad studies of the competitive market situation would be more appropriate. A small, although enlightened, set of customers may not be able to indicate which product variant will sell the most. Actual sales when the products are in the market will be an accurate indicator, requiring your teams to develop product designs that are flexible and production processes that can react quickly to market feedback. Depending just on information from customers may not be wise if your proposed product concept is a radical departure from current concepts and thereby not familiar to customers. When your situation does justify direct interactions with customers, recognize that there are several creative and valuable methods for doing so. When practiced effectively, the benefits of customer involvement have a high potential of being critical to the success of your product development projects.

The Voice of the Customer Can Be Critical to Engineers The “Voice of the Customer” (VOC) is just what the words imply: inputs directly from customers about their needs for new or improved benefits from products and services, and the preferences customers value that guide the trade-offs that they make when comparing competitive offerings, particularly when constrained by budgets. Customers may express their needs as, for example, problems that create difficulties for their business or interfere with activities that they’d like to pursue more easily. They may desire lifestyle improvements or capabilities that reduce the stresses in their lives. They may feel that prices are too high when compared to the benefits that they expect. They may want to enable improvement initiatives for their own business, but worry about the certainty and timeframe of the payoff. They may need to comply with environmental regulations, or want to develop a new business opportunity. Certainly the scenarios are endless. The point is that customers have problems or desires for which they are willing to spend money for solutions. Factors that are critical to the success of product development projects1 include, among others: • Requirements that define clearly differentiated benefits and costs that are superior to those from competitive products. • A sharp and early definition of the best product concept. • A strong orientation toward those markets and customers chosen to be served. 1. Robert G. Cooper, Winning at New Products: Accelerating the Process from Idea to Launch (New York: Addison-Wesley Publishing Company, 1993).

The Voice of the Customer Can Be Critical to Engineers

An important objective of product development teams is to understand the basic needs, preferences, and timeframes of customers they choose to serve. Those understandings can then be translated into technical requirements that can be acted upon by the teams developing new or improved products. Although not a topic for this article, the same can be said for new or improved services to be delivered to customers. A key principle is that the information is “voiced” or demonstrated by those with the needs, in the context of their business or activity, with an understanding of the stressful conditions that are relevant. That says a lot. It implies that the dialog with customers must have a rich and deep content, more than you might expect for conversations with marketing and sales people who, rightly so, are more interested in an acceptable price or intent to purchase. How often have you found yourself, as a customer, wishing that developers of products would not only listen to you, but actually understand what you said and meant? Consider an example scenario of shopping for a flashlight (more generally, a portable source of light) in your local camping supply store. What features do you need? How will you use the flashlight? What aggravates you about flashlights that you already own? What causes flashlights to fail just when you need them the most? How do you compare alternative flashlights that the store offers? A value proposition is an unambiguous description of benefits to be provided to chosen customers, compared to their costs. As a developer of new products, what value proposition does your team propose to offer to the market? How will it be differentiated from existing value propositions? You may choose to develop a product with easier handling, more functionality, or higher reliability. What are the conditions under which it needs to operate? Who will use it? Is the user agile? In the flashlight scenario, are variable intensity and variable beam width important parameters? What will the user do with the lighting? If your development teams do not understand these criteria, it’s unlikely that the solutions that they develop will satisfy customers well enough that sufficient sales at a competitive price will generate an acceptable financial return. That is an important objective of new product development, isn’t it? So here you are back in the camping and hiking supply store looking at flashlights, and a nice person with “vendor” printed on their vest introduces herself and asks you probing, open-ended questions about how you want to use “portable light sources” and what characteristics aggravate you the most about ones that you already own. As a customer, you’d probably think, “At last someone asked.” As a producer, you would think “At last I found out.”

853

854

Listening to the Customer First-hand; Engineers Too

The Voice of the Customer Does Matter to the Business There are two fundamental questions that your development teams need to answer early in their project: 1. How will your new product (or service) deliver value to your target customers that will be superior to alternatives available to them after your product’s market entry? 2. How will your new product return value to the corporation better than the next best alternative investment? The first question is at the heart of why a customer would choose to purchase your product instead of that from a competitor, and to do so repeatedly. The second question focuses on what makes a good financial justification for the project, as presented in the project’s business plan. The business plan presents to management the project’s expectations for the investment in product development, market entry, and production, and the returned revenues from sales, customer support, and service. Management teams, who manage portfolios of development projects, must judge whether or not the proposed project is expected to generate higher returns for the corporation than the next best alternative investment. That’s one of the jobs of management. What solutions at what price and at what time will generate the highest flow of revenues? Higher price and/or higher placement volume with acceptable manufacturing costs are strong contributors to acceptable financial forecasts. But what features or functions or level of quality are necessary to command a higher price or to generate a higher level of demand for the product? Many customers in your target markets need to decide that the benefits of your new product fill real needs, that the price is acceptable, and that the resulting value—for example, benefits compared to cost—of your offering is superior to alternatives available to them. So, delivering superior value for many customers is critical to your business. In addition, the timing of the availability of your new product may be critical. Certain markets are seasonal, such as with the agricultural industry or for products tied to holidays. That new flashlight you plan to develop may make a popular gift. What is the timing of the buying behavior of your customers? Customers have options. They can choose among alternative offerings. They can wait. They can do nothing. So your solutions must be perceived not only as being superior to alternative products, but also superior to customers using their money for other purposes.

Who Are Your Customers?

Who Are Your Customers? It is surprising that often this is a question with an uncertain answer. One answer is that a customer is the next person down the value chain. For consumer products, that might be the distributor or retail sales channel. For an industrial product, it might be the buyer or those who advise the buyer. Or is it the user or operator of the product who have to become more productive? In the case of a high volume printer, is it the end user of the printed material—that is, the customer of your customer? A broad view suggests that “customers” are all of these people, since they all have needs to be satisfied, although probably not the same needs. The operator of a product may care little about the price but will care a lot about the functionality and usability. The owner may care more about the environmental impact or service policy, and even more about the cost of ownership than the purchase price. A buyer may focus more on price and delivery. All of these people are customers who can contribute insight relevant to the range of requirements. Do you have to talk with all of these people? Of course, if you want to satisfy their needs. When you add a lifecycle context, you have to include the experiences of your customers in learning about your product and comparing it to alternatives, the processes for purchasing the product, its consumable materials, or maintenance parts. Depending on the nature of the product, it may include its eventual decommission and disposal. So there are many “voices” to be heard. Their differences need to be understood, respected, and integrated. To make matters a little more difficult, most products are developed not for specific customers but for market segments. Those are large groups of customers with similar value drivers. They may be global. So many representative customers need to be heard, with their collective needs and preferences internalized so that new products respond to the needs of market segments. In our example scenario, as a producer of flashlights wanting to learn about the market, you can ask the people who are buyers for large retailers. They will tell you about street prices and demand volumes. They may talk about their dealings with suppliers and the tactics used by your competitors. If you ask the customer support people, they may tell you about display strategies, packaging features, and possibly even questions asked of them by end users. Neither of these sources will know much about how customers use the flashlights or why they favor one product over another. However, their information is important to the requirements needed by the development team. The solutions to those requirements may take the form of advertising, packaging design, delivery processes, or launch plans developed by your multifunctional teams. Customers who can tell you about what they need and who have a vision of future applications will provide information that you can use

855

856

Listening to the Customer First-hand; Engineers Too

directly to formulate requirements for new products. For example, suppose a major outdoor outfitter wants to provide “portable light sources” for people who explore caves or climb rocks. How would they turn on the light, or adjust the beam? To what extent would that need to be one handed, or no handed? What if the application was a stressful military application? What if the user was injured? As a product development engineer, how would you know what to do?

Development Teams Need to Listen to Their Future Customers Your company’s view may be that it’s the job of marketing or sales to talk with customers. They already know customers and have responsibilities for revenue forecasts and achievements. How about market research? They have responsibilities for forecasts of market price, sales volumes, market share, and other parameters in the business plan. Those interactions are necessary. However, as mentioned earlier, a factor critical to the success of product development is for requirements to be clear, differentiating, and value adding. For product development engineers, the filtering of information through marketing and sales people has high risks of getting it wrong, incomplete, or at least not good enough. For example, people in sales may focus on the inadequacies of the current products. In-bound marketing specialists may focus on high-level value drivers for market segments. People in market research may focus on broad market parameters rather than those specific to a product. An improved paradigm is that those people who develop solutions for customers should be the ones who need to be directly involved with those customers who can speak about their needs, describe how they make choices, and explain their aggravations. Engineers developing products need to think like their customers and make decisions, every day on the job, so their customers win. Multifunctional product development teams include several types of people who need to interact with customers. They can include the following: • Engineers, designers, and other creative people who develop the products and their packaging • In-bound marketing people who can act as customer advocates • Out-bound marketing people who must manage the promotion of the new product and its entry into the market • Quality assurance people who conduct development and verification tests, and who are advocates for quality parameters most important to customers

The Involvement with Customers Is an On-Going Proce ss

• Service engineers who develop diagnostics and tools to keep the product operating as intended • Customer support people who develop training and help systems, and who are the initial contacts to resolve customers’ problems Management must promote an understanding of their customers’ businesses and activities, oversee the development projects, and provide the resources to involve customers directly in the process. So, key managers must also engage customers at an appropriate level.

The Involvement with Customers Is an On-Going Process The processes for involving customers apply to three basic timeframes: 1. The planning for the portfolios of new products and the stream of new or improved technical capabilities that enable them 2. The development of specific products with inputs and feedback that influence the detailed requirements, the selection of the design concepts and architectures, and the optimization of specific features and functions 3. The support of the new products in the market, with feedback that can improve those products, or provide requirements for future products These stages provide a reasonable way to explore practices that can make customer involvement a routine, highly valued element of your business practices.

Prior to Product Development Advanced product planning is an on-going activity. Deliverables from it include not only portfolio plans for a stream of new products. An objective of advanced planning is to identify those opportunities that can add differentiated and superior value to customers. That value must have the context of customers’ changing activities and usage patterns, as well as the expectations for actions from competitive companies. The early plans for portfolios of future products need a broad understanding of customers’ applications, needs, and costs. New or improved technical capabilities may need to be developed or acquired. These long-range plans must reflect a vision of how relevant value propositions will evolve over time and be competitively superior. A proposal for the investment in new products and technical capabilities must include an initial business plan to estimate the benefits, costs, and timelines for development projects. For example, in our flashlight scenario, do you see a growing market for higher-value devices

857

858

Listening to the Customer First-hand; Engineers Too

that have applications and expectations that are clearly beyond the conventional notion of a flashlight? How will they be used? What technologies will be required to enable those designs? What retail channels will be appropriate for them? Product development projects can benefit greatly if those concerns are resolved prior to their start. Otherwise, the projects will have to take on that work, adding time, costs, and risks. The value propositions must be important to the businesses or activities of the customers. For example, suppose your company’s vision is for “portable light sources” that can be recharged after long use, are adjustable from a wide angle to a focused beam, are operable by one hand, and can tolerate submersion underwater at sub-zero temperatures. What is the persona2 of customers for these products? Describing the character and activities of these customers can be a useful tactic for capturing the intent of the requirements. These are not just features and functions that are more extensive or less expensive than those of competitive products. The objective is not just “faster, better, cheaper.” The benefits must enable the users of your products to do something that they could not do otherwise, or do as well or as easily. The price positions must be perceived as being acceptable, given the benefits and competitive alternatives. Otherwise, why would they purchase your products? Prior to product development, the work of market research is very important to understand the various elements of the competitive market situation. The future purchasing decisions will be influenced strongly by many factors, such as market economics, product and price positioning, customer demographics, and so on. In addition, customers’ future applications, needs, and costs need to be studied. This can be the responsibility of in-bound marketing people experienced in this work and assigned to the planning team. Another valuable approach is to assign engineers, scientists, or future project leaders to advanced planning teams. They will learn about customer-driven requirements with a view of their consequences for product development projects. They may actually join those projects later. The knowledge derived from this involvement becomes the basis of the value propositions that are planned for future products and the plans for the development or acquisition of new technical capabilities required by those future products.

During Product Development The involvement of customers with development teams can satisfy several important objectives: 2. Alan Cooper, The Inmates Are Running the Asylum (Macmillan, 1999).

The Involvement with Customers Is an On-Going Proce ss

• It can ensure that the value proposition specific to a new product is an accurate reflection of customer-driven needs for benefits and costs. • It can enable and validate the translation of those customer needs into the more technical requirements for the development of the product. • It can guide the selection of the product’s architecture and its detailed development. • It can provide feedback from the use of development prototypes at both system and subsystem levels to improve those designs prior to their completion. • It can determine whether or not the product under development will be both acceptable to customers and superior to competitive alternatives. • It can provide feedback to improve the delivery of the product and its supporting services. Customers can provide much insight that is not expected to be known by those developing the solutions. The key is for development engineers to obtain those inputs when there’s still time to react. For example, the clarification of and concurrence with the requirements are needed in the early development phases. The validation of design concepts is needed before designs are refined for production. It might cause a project to go back to an earlier phase to resolve a flawed design concept. Feedback from the use of prototypes can be reacted to until the design is declared to be complete and frozen for launch and production. Feedback that can influence the design architecture is needed much earlier than feedback about the detailed features. Inputs that have consequences for mechanical tooling, for example, are needed much earlier than inputs that affect software, because of the longer lead times to react to their change. Feedback that can be implemented by changes to an adjustment, to a user’s procedure, or by a training course can be handled closer to market entry since their response time and consequences are much less. So, customer involvement is viewed to be not only important, but also critical throughout the development process.

After Market Entry The real proof that advanced planning and product development were done well, and that the business plan described an achievable business opportunity, is in the repetitive purchasing and use decisions made by customers when faced with competing solutions. This may be in their initial purchases, in their purchase of additional products and accessories, in their

859

860

Listening to the Customer First-hand; Engineers Too

use of consumable materials, and in other ways that they spend money to solve their problems or to enhance their activities. Their feedback is very valuable. It can trigger corrections to the current product, some of which, unfortunately, might have to be mandatory. It can provide insight into the needs for future products, features, functionality, reliability, or costs. It can identify ways in which the early VOC knowledge was either on-target or flawed, leading to process improvements. This is not about improving your “Customer Satisfaction Index.” It is about learning how well the benefits of your product will satisfy the needs of your customers and thereby deliver value better than that expected from the products of your competitors.

A Variety of Methods Can Be Productive for Development Teams I’d like to suggest some approaches for involving customers. Some may be new ideas, while others may already be in practice. In the context of your products and customers, you may have additional creative ways of gaining VOC information. There is much guidance available in professional literature and coaching about several of these methods. It would be beyond the scope of this article to explain them all in detail. Probably more than a single approach will be necessary. They do not need to be tremendously expensive or burdensome. However, the processes do need to be deliberate, funded and staffed, designed well, at the right time, and improved with practice. Here are brief overviews of several tactics that may find application as you improve your involvement with your customers.

Conduct Market Research Studies of target markets can analyze competitive dynamics, relevant power among suppliers and buyers, and trends in price and product positioning. Market growth and competitions for market share can be characterized, along with their economic factors. Trends in technologies, product functionality, industrial design, packaging, and other competitive parameters can be identified. This research may define or clarify value drivers for market segments and enable forecasts of changes in those drivers. In our flashlight example, what trends in the industrial design of outdoor equipment are expected to be influential? Is that market expected to be interested in hand tools of higher value? Are market segments being redefined? What retailers are expected to move into them?

A Variety of Methods Can Be Productive for Development Teams

Benefits: This information can contribute integrity to key parameters in the business plan. It may also be crucial to the definition of the best portfolio of future products. Cautions: It is unlikely that these data will contribute much to product requirements. Worse yet, the people doing this work often believe that their job is to report to management, rather than to provide insights for engineering. So, market research is necessary, but not sufficient.

Evaluate Competitive Products Certain competitive products may have a well-deserved reputation in the market. What is the value that they deliver to customers? Comparisons of product capabilities can be performed either by internal benchmarking initiatives or by external organizations and trade publications. The most valuable process is an internal one that usually involves the testing and disassembly of competitive products and their evaluations by engineers and other people challenged to develop superior solutions. Often good ideas can be gained from several competitors and integrated into a superior new product. As you study flashlights, in our example, you’ll probably visualize how the ideal device will be used. Possibly you’ll want to use it outdoors in the rain and mud. You may want to use a focused beam to read maps or see the details of a repair, but also a bright and wide beam for hiking in the woods or camping. You may need the time between battery changes or recharging to be exceptionally long, since the access to recharging power may be uncertain. And you may need the light to be able to hold a stable and adjustable position when set on the ground. No current product may satisfy all of those requirements. However, what features do you see in available products that achieve those requirements separately? Do you see features that indicate requirements that you had not considered? Benefits: A strategy of benchmarking with internal teams enables many questions to be asked, contributing to project decisions and development tactics. If done externally, there’s no control over the character and depth of the information learned. An objective of product development, at both system and subsystem levels, is to design capabilities that are superior to those design concepts in competitive products. That differentiation may be in performance, robustness, costs, usability, appearance, and other requirements important to customers. Cautions: These evaluations can face challenges in access to competitive products with freedom to take them apart. That’s not much of a problem for an inexpensive consumer product, but what about large costly product systems? Can you lease them and then return them in their original condition? Can you obtain service literature? Do you have the people

861

862

Listening to the Customer First-hand; Engineers Too

with sufficient knowledge and bandwidth to do the work internally? I’ve seen an extremely valuable organization for this process devalued and dismembered by budget cuts and management indifference. What a loss! Another caution is that these evaluations only reveal what competitors have already accomplished in products. They say little about their future intentions and say nothing about what customers will need in the future. If your strategy is that of a “fast follower,” then competitive information is very useful. A concern is that the information is too late for those who want to lead the market.

Select Lead Users for Focus Group Discussions A focus group is a collection of friendly lead customers who agree to meet periodically and act as advisors to your business on behalf of their industry or market segment. Its participants can provide substantial wisdom about their businesses and markets, and the trade-offs that they must make. Their discussions are enriched by their interactions and by the familiarity of a routine advisory process. In our example scenario, imagine the conversations among scouting leaders talking about portable light sources for weekend backpacking trips. Imagine similar conversations among extreme mountaineers. I can imagine their needs to be quite different.

Benefits: The participants and discussion topics can be customized to the needs of both portfolio planners and product-specific development teams. Depending on how visionary the participants are, market trends and value drivers can be understood, along with business and technical trade-offs that are characteristic of the market segment. If the meetings are planned well, the engineers designing the solutions will be included, rather than these meetings being just the domain of marketing and management. The outputs may add integrity to parameters in the business plan and even validate the value propositions for future products. Cautions: Conclusions may be more relevant to the advanced planning of product families than to a specific product. Their insights may be personal opinions and difficult to generalize to an entire market segment. Lead users may visualize a future beyond the lifecycle of a new product in question, or focus on current problems in need of fast solutions. The facilitator of the discussions will have to ensure clarity. For large product systems, these customers may have been selected because they are most important to the business. The environment of those meetings may be very sensitive and constrain the participation by development team members. Often these discussions tend toward determining the customers’ intent to purchase rather than focusing on an understanding of the value drivers and trade-offs among requirements. Given that travel and scheduling can be a

A Variety of Methods Can Be Productive for Development Teams

problem, the planning of meetings may need substantial administrative efforts. Professional facilitation may be necessary.

Ask Customers to Participate in Surveys Surveys can be conducted in many ways, such as by the Internet, email, or even by snail mail or telephone. The strategy is to ask questions that have a reasonable probability of providing information that can be useful to the development of requirements. Benefits: A large population and variety of customers can be covered, with the survey costs being dependent on the method used. Cautions: Although the strategy can be cost effective, its value can be very dependent on the ability to design questions that draw out the details and perspective desired. However, with no ability to ask followup questions, there is a risk of not asking the right questions. The value of the information will depend on the willingness of the customers to respond in detail and to be honest and thoughtful. The probability of response can be low and the data collection can be slow.

Analyze Databases of Suggestions and Complaints Feedback from current products can provide information about missing features, price concerns, failure rates, and requests for customer support, for example. It can also provide verbatim comments from customers. These may contain requests for specific improvements in the current products as well as ideas to be developed into new products. Benefits: For many companies, this information is collected and analyzed routinely. It’s more relevant to product extensions for which customers expect current problems to be resolved. Certainly it should not be ignored. Cautions: This information looks backwards, only telling you what’s wrong today. It tends to lack insight for the value propositions and requirements for future products.

Work in Customer Support or Product Service These two organization functions deal directly with the problems that customers have with current products. By being exposed to these interactions, people in product development can gain a greater appreciation for how customers perceive and react to problems. Work the telephones. Answer emails. Travel with service representatives. I remember carrying the tool kit for a service representative in New York City where the walking distances were long, the work spaces small, and the patience of customers little. I

863

864

Listening to the Customer First-hand; Engineers Too

returned very motivated to reduce the number of tools required and to improve the ease of repair and maintenance. Benefits: A lot can be learned from experiencing the discontent of customers. Their perceptions as product users will likely be different from those of product developers. Cautions: Remember that this feedback is relevant to current products. Your job is to develop new products. So the feedback needs to be interpreted to be relevant to the requirements for the new products.

Have Customers Test Prototypes of the New Product The prototypes to be tested can represent partial solutions during several of the development phases. In early phases, customers can evaluate conceptual illustrations, mock-ups, or subsystem prototypes. Once system-level prototypes are available, customers can test them under real use conditions, when there is still time to change the designs without jeopardizing production preparations. Customers can also participate in the verification of designs to ensure that they not only satisfy customer requirements but also will be superior to competitive alternatives. In your portable light source business, how about giving prototypes of your new robust, all-purpose “backpacking torch” to troops of scouts for weekend camping trips? Then interview them at their next troop meeting. After you’re comfortable with the early design, give more mature prototypes to your mountaineering friends for their use in week-long treks. The key is to obtain useful feedback when there’s still time to react—for example, improve the design. If the design architecture or a feature concept is flawed, you certainly want to know that very early in the process.

Benefits: Early feedback from customers evaluating design concepts can contribute to trade-offs among alternative design concepts, validate design architectures, and identify design flaws during development. Later customer trials can also evaluate services such as training, customer support, product repair or maintenance, and instructions for use. Cautions: Chosen customers may not represent the market segment well, so a variety of customers will be advisable. Some customers may not be willing to convey bad news freely, particularly if they feel like a guest or think that you already know the information. The engagement process needs to foster open, honest feedback. If they somehow pay for the use of the prototypes in their business, or are encouraged to feel important to the process, their feedback may be more assertive. Often there are concerns about showing a product concept to potential customers too early. It may create expectations prematurely. It may jeopardize the confidentiality of the development project. Engineers may not

A Variety of Methods Can Be Productive for Development Teams

want to submit partial solutions to these evaluations. These concerns tend to delay the customer tests until the design is complete, when it’s too late to change the design without causing major consequences.

Have Customers Join Your Team Some of your customers will appreciate the opportunity to contribute ideas and opinions that can guide your development teams. These inputs in the early development phases can contribute to the principle of “getting it right the first time” and reduce time-consuming changes in later phases. The tactics for having customers act as an extension of the development team can range from participation in project reviews to contributing to decisions. The strategy is for direct involvement, particularly by customers with substantial insight and a healthy spirit of cooperation in developing solutions for their own benefit. Benefits: Feedback from customers can validate value propositions, clarify requirements, assist trade-offs among alternative design concepts, validate design architectures, and help development teams to think like customers. Cautions: As with the cautions about prototype testing, chosen customers may not represent the market segment well. This type of involvement requires resources to manage these relationships and their logistics in order to be value adding and reasonably efficient for development teams, as well as fostering the open and timely participation by customers. This approach may aggravate the risks in maintaining confidentiality, so the customers chosen need to have a sense of commitment to the project as if they were employees.

Spend a Day in the Life of Your Customers This is the title of a noteworthy article3 that advocates visiting your customers’ businesses and observing their activities in the context within which your products would add value. Small cross-functional teams can tour customers’ operations, observe relevant products being used or abused, understand their applications, and identify conditions that are stressful to a product’s performance. They can ask probing, open-ended questions of key people about workflows, key difficulties, improvement initiatives, business metrics, buying criteria, perspectives about costs, and other topics of concern to the value proposition and the product requirements.

3. Francis J. Gouillart and Frederick D. Sturdivant, “Spend a Day in the Life of Your Customers,” Harvard Business Review (Jan.–Feb., 1994).

865

866

Listening to the Customer First-hand; Engineers Too

How will this help your flashlight development project? Go on a camping trip. Go on a trek. Observe your customers using the prototype designs under stressful conditions. Ask a lot of questions. Listen well to their feedback.

Benefits: By listening to customers and watching them in their own environment, much can be learned about difficulties that can be overcome or advancements that can be provided through clever product designs. The task is to listen and watch, letting your customers do the “show and tell,” and to do it better than your competition. Cautions: Your engineers may have to learn how to ask probing, open-ended questions and to listen reflectively, without judgment. Your sales reps will need to enable the visits, but not get in the way of the learning process. Let the customers lead the process. Your teams may need to become culturally competent in their customers’ world and not just look for problems with their own preconceived solutions. The cost of travel may be a concern for the accountants, but the real barrier may be the “get up and go” that gets your people to move out of their own space.

Work for Your Customer If you develop products for businesses, find customers who can welcome you in to operate your products and those of your competitors within your customer’s workflow. Work there for a week. You’ll be amazed at how much you’ve learned. How about spending a weekend working at your local outdoor adventure store? Sell flashlights. Talk to their customers. Listen to their decision processes.

Benefits: This will not only make you familiar with the position of your product in your customer’s business, but also enable you to understand more about the customers of your customer and how value is perceived for their intended applications. Your intent is to think like your customer and to speak their language, to understand what gets in the way of them getting their job done, to feel their pain. Cautions: Of course, you’ll need cooperative customers with the right opportunities. That may limit your experience but not devalue it.

Be a Customer Be a routine user of your own products and of those of your competitors. Read the instructions. Take the training. Call the help desk. Interact with your products and services as do your customers. If you are not a user of flashlights, you may have little insight into what makes a really good one. Flashlights can be very frustrating. They can roll around, get lost easily, have short battery lives, be difficult to turn

What Do You Do with the Voice of the Customer?

on, and not throw a strong enough beam. They can leak and rust, break when dropped, be weak when it’s very cold, and be heavy. What else? Well that depends on the applications and how stressful they are.

Benefits: You’ll learn about the pros and cons in the same way as your customers do. Better yet, you’ll develop the insights that identify new opportunities to improve your customers’ activities, to make them more enjoyable, more convenient, or more productive. Cautions: Be careful since you may have some biases that can prevent you from seeing your products in the same light as do your paying customers, or those who refuse to pay. All of these tactics can contribute necessary information. The question is which practices should be part of your culture. In many companies, there may be pockets of excellence in customer involvement. Patagonia, the outdoor clothing company, has made it their culture for employees to be active outdoors and to use their products and those of competitors. Look at the knowledge they have about their industry. In general, however, good practices are far from universal and systemic, and are vulnerable to reduction or deletion when budget or time pressures are high. Not all of them may be practiced or respected in the project management plans or development budgets. Not all of them are needed. The last four noted previously may not only be the most valuable to both development teams and customers, but unfortunately may be the least practiced. A VOC project may be viewed as an option for development teams, competing for resources and time with technical development work. Its justification needs to reflect the expected benefits to the value of the new product. It also needs to reflect efficiencies the development teams expect from gaining better information sooner and from making better decisions throughout the development phases. Often “technology push” initiatives feel that R&D people know more than do customers about market needs. This may be the case in highly innovative development situations for which customers are not aware of the product concepts or how they might be used. In “market pull” situations, however, detailed, accurate, and early listening to the voice of the customer is critical to success. In order to provide a competitive advantage, the involvement of customers with internal teams needs to be an element of a culture of partnership between those teams developing solutions and representative customers who will benefit from those solutions. The paradigm is for customers to be an extension of the development team.

What Do You Do with the Voice of the Customer? Quality Function Deployment (QFD) is a tool that can be helpful in developing requirements, in ensuring that they are a sufficient translation of

867

Listening to the Customer First-hand; Engineers Too

customer needs, and in deciding how the new product can be superior to its competition. • Higher-level statements of customer needs are decomposed into lower levels of detail. By asking follow-up questions, your teams get to the “A-ha” moment of real understanding, and your customers will say “That’s right, you got it.” • Preferences that customers express and the parameters of comparison among competitive offerings are translated into the goals for satisfying particular customer needs and the target values for requirements that are derived from them. • The QFD process translates those needs, stated in the language of the customers, into requirements stated in the language of the solution. Here’s an example for our “backpacking torch,” partially completed to illustrate the method. Figure 1 shows a portion of the QFD matrix known as the “House of Quality.”

Improvement Ratio

Sales Advantage

Scores

% Score

9

3

Competitor 2

9

Bulb Brightness

Rugged Filament

Focusing Lens

9

New Product Goal

9

3

Competitor 1

Securely Held

Durable Battery

3-legged base

9

Current Product

Resistant

Drop 4 ft without damage

Planning Matrix

Customers

Drop

Tight Seals

Customer Needs Represent a Market Segment

No slip handle

Product Requirements

Rugged Case

868

10

9

8

8

9

1.0

1.0

10

16

4

4

6

8

8

2.0

1.2

9

16

8

7

7

8

7

1.0

1.0

8

12

5

3

9

9

9

3.0

1.5

23

36

Rugged Weather Proof

Wide Beam Bright Light Narrow Beam

Operates at -20F

9

Operates to 20 meters under water

9

9

Readable text at 100 ft

3

9

Illuminate 50 ft dia . 100 ft away

9

9

Focus on finger nail at arms length

9

3

Readable text at arms length

3

3

Target

Figure 1: Voice of Customer Portion Within a QFD Matrix

Planning Score = Customer Value x Improvement Ratio x Sales Advantage

What Do You Do with the Voice of the Customer?

The structure in Figure 1 shows: a. The Customer Needs Matrix has the voice of the customer decomposed into lower levels of detail so that they are actionable. These insights would be derived from early involvement with customers and their applications. b. The Requirements Matrix illustrates that some of the needs will be addressed by more than one technical requirement. The numbers in the cells represent the extent to which a particular technical requirement contributes to a specific customer need. This mapping would be derived from internal teamwork that translates the voice of the customer into requirements, independent of their solution. c. The Planning Matrix shows the results of prioritizing, benchmarking and goal setting. The scores represent the conclusions from the work of teams closely involved with customers and competitive product assessments. The details of the QFD methodology help to establish measurable product requirements, resolve conflicts among them, and deploy the requirements to subsystems, components, and manufacturing processes. Later, if a particular technical requirement cannot be achieved well, the consequences for the customer can be understood through this mapping. It’s important to expect that the voice of the customer will not be complete in content. When you ask customers what they need, they will tell you about things that are on their mind. That may reflect their current problems, or activities that they’d like to enhance. However, there are two categories that you probably will not hear about: those needs that they did not think to tell you about and those needs that they did not know to tell you. The first category tends to represent those elements of a product that customers assume to be implemented well and are outside of their purchasing decision. In the purchase of a car, for example, the fundamental structure, suspension, and weather protection may be unmentioned requirements, unless it’s an off-road vehicle for which these will be sources of competitive differentiation. The competition among passenger cars may very well be with characteristics such as elbow room, visibility, acceleration, and reliability. Achieving these well tends to keep customers happy in the long run. But what gets them to purchase the car? Often customers can be impressed by new features or functions that they did not know enough to ask for, but were identified by the insight of people in marketing, engineering, or industrial design. Characteristics in this category can include turning radius, sound and vibration isolation, audio system quality, comfort, styling, and warranty terms. Delivering these features and functionality not only surprises customers and differentiates your product, but

869

870

Listening to the Customer First-hand; Engineers Too

may be the key to winning the sale and establishing a superior reputation in the market.

Summary There are four fundamental outputs contributed to by the involvement of customers: 1. A portfolio of new product projects that will deliver superior value to target customers via a stream of future products. 2. Requirements that differentiate a specific product that is to be developed. 3. Product designs that can be demonstrated to satisfy those requirements. 4. A business plan that forecasts how a development project will return superior to the corporation. I encourage you to adopt customer involvement as a principle of your business, to decide how to do it efficiently, to integrate it into your project plans and budgets, and to learn to do it so well that it influences your reputation in the marketplace. This perspective puts your customers at the heart of your product development process. They have the needs. Your may have solutions to those needs. They make trade-offs among alternative sets of benefits and costs. For customers to decide to spend their money on your products, they must perceive that your products will deliver value to them that is superior to alternatives available to them.

Additional Reading Sheila Mello, Customer-centric Product Definition, The Key to Great Product Development (New York: American Management Association, 2002).

About the Author Bill Jewett is a consultant to businesses engaged in the development of new technologies and multi-disciplined products. With insights into important paradigms and advancements in practices, he assists improvement teams in upgrading their engineering and management processes, project management practices, cross-functional teamwork, and the governance of development programs. For many years, Bill worked for Eastman Kodak Company and Heidelberg Druckmaschinen, with focus on the development of high-volume electrophotographic copiers and printers. Among his division-level

About the Author

responsibilities were the management of product development programs and of competency centers for mechanical and systems engineering. At the corporate level, he was one of the authors of the processes for advanced product planning, technology development, product commercialization, and their related governance. For over a decade, he taught the processes and coached teams in their adaptation and implementation. As the process steward, he evolved the models to incorporate lessons learned from internal practice and external benchmarking. Currently, Bill and an associate are writing a book to integrate strategies and methods for the development of increased robustness and higher reliability in products. They expect their book to be available in the last half of 2007. Bill can be reached by telephone at 585-705-3100, or by email at [email protected].

871

This page intentionally left blank

The Practice of Designing Relationships By Mike Cook

This article is primarily excerpted from the author’s recently released book, THRIVE: Standing on Your Own Two Feet in a Borderless World, published by St. Lynn’s Press, Pittsburgh, PA.

Executive Summary Seeing an article dealing with “relationships” at first may seem a bit odd, given the context in which you have been engaged, that of course being Six Sigma and its several means of being expressed. It may be artistic license to include this article here, or it may be a legitimate perspective to be developed further. I’ll let you be the judge of that. Six Sigma as it is commonly understood is most often spoken of in at least three recognized contexts: • As a metric • As a methodology • As a management system Which one is it? Might it be one, two, or all three? Most experts in the world of Six Sigma would likely agree that it is all three, at the same time. So there you have it—what the experts say, at least. However, is theirs the last word? There may yet be other contexts to consider and even validate as we continue to research and explore questions of not only how to continually improve the quality and quantity of our outputs but also the quality and impact of the Six Sigma approach itself. In this article, I will suggest another possible context in which to consider Six Sigma. I’ll venture to say that on one hand, this context is the most obvious and thereby has been overlooked, as the obvious sometimes is, and on the other it is likely the least explored, at least until now. Could Six Sigma be used to evaluate critical human interfaces—i.e., working relationships within the organization, individual and business roles, the working environment, the quality of the product, and the 873

874

The Practice of De signing Relationships

connection of the business to the customer? I assert that as we continue to be more and more knowledge based in what we offer our clients and customers, the individuals in our employ and the connections among them (relationships if you will) become more and more key to the delivery of optimum value. Certain positions in all organizations have traditionally been more obviously part of our value propositions, sales, and customer service as examples. However, as we face the reality of the true assets of our business going home every night to who knows what risks and temptations, (Will the employees be back in the morning?), we must address the questions of how to continually improve the quality of the following: a. An individual’s relationship to our organization b. An individual’s relationship with or “fit” for the role they are playing in our process c. Our employees’ readiness to develop and sustain working relationships grounded in a shared respect for the Six Sigma approach This article offers at least a model within which to explore a Six Sigma approach to “connections,” or relationships, and perhaps even a pathway to follow for further exploration.

Distinguishing a Conscious Approach to Being Connected in the Workplace There is a conscious approach to be learned and a craft to be developed in mastering the practice of Designed Relationships. This is a pretty bold statement to make from the very beginning. Aren’t we all adults? Don’t we know how to interact successfully with our co-workers and others we encounter in our daily lives? As I have pursued my consulting practice over nearly 20 years, I have gathered pretty unequivocal evidence that strongly suggests that: a. Whether we are all adults is universally a matter of debate, and it doesn’t look like the final verdict is going to be “yes.” b. Almost without hesitation, every person I have encountered—executive, manager, supervisor, or frontline employee—agrees that the toughest part of being at work is dealing with people. The rest of our working issues may be challenging, but we are generally pretty confident that there are definite answers and eventually we’ll find them. When it comes to people, it seems like it’s an “every man for himself” and “good luck with that” kind of world.

Distinguishing a Conscious Approach to Being Conne cted in the Workplace

So, where do you need to be connected, who do you need to connect with, how do you need to connect, and finally, are you sufficiently connected in all these cases, or not? Are these even valid questions? Are relationships like bits and bytes, either on or off? A quick assessment of your life will probably reveal that “yes,” you are related, in some cases more than others, and that there are gradients in relationships. Sometimes I have misread where I was on the relationship gradient with a given person. Maybe I have a great co-worker and it’s very satisfying working with him or her, and so I assume I’d get as much satisfaction from having a beer with them. I make the invitation for something social and find that they feel our working relationship is great and it is derived from the shared context of being at work—but they have no desire to have it be otherwise. Pardon me! Something like this has happened to all of us, I imagine, either professionally or socially. Relationships, we find through experimentation, are context-bound. If you presume to place the same expectations on the relationship elsewhere as you do in the shared context, you may quickly find the limits. Nevertheless, I am getting ahead of myself. Is there a gradient in relationships? I think so. The gradient is the measure of power available in the relationship in the context in which it was formed. I imagine that if you are honest with yourself, you have wondered about how some people, obviously not as smart or talented as yourself, have managed to become so successful? Of course you have—we all have! My observations over years of consulting have led to the conclusion that there are some cases that can be accounted for with the “right place, right time” theory—that is, dumb luck! This theory does not, however, account for enough cases to warrant developing an attitude suggesting that you are either lucky or you aren’t. There are significant numbers of people we know who have not “gotten lucky” just once but seem to be able to reproduce their luck in a variety of circumstances. These people and their habits and practices do merit closer examination. Figure 1 illustrates the conceptualization of my own observations. The repeatedly successful people, the companies with enviably high customer retention, and those same companies having remarkably high product satisfaction or low product return figures, all follow similar patterns: They determine very clearly and precisely how they want to be known by their customers and employees, what they will produce, and who they will produce it for. They have a vision. The second essential element in the formula for success that is either sustainable or repeatable is responsibility. The product of the relationship between vision and responsibility, POWER, is ultimately the source of success in whatever the chosen endeavor.

875

To save you the trouble of going to a dictionary to establish my meaning for the terms “engagement,” “responsibility,” and “vision,” allow me to introduce my meaning for these words:

Po we A r to Fu M a nc tio ke th n in of En gs H ga a ge pp m en en : A t

The Practice of De signing Relationships

Responsibility

876

Engage—To choose to involve oneself in or commit oneself to something, as opposed to remaining aloof or indifferent. Responsibility—Both the ability to Vision respond and the willingness to answer Figure 1: Power as a Function of for; something we all can or do have. Engagement Vision—The act of perceiving mental images, again, something we all can do. On one hand, we may have vision without responsibility and could be thought of as wishful thinkers. Someone who operates in this fashion is often called a dreamer. On the other hand, responsibility without vision may make us a “good person,” but in the world we live in, being a “good person” is not a solid foundation for thriving when things come to a matter of economics. In combination, responsibility with vision offer an experiential definition of engagement, with power as their product. Power—The capability of producing an effect. I prefer to think of power as the ability to get things done. For any position on the gradient of power, I believe there are distinguishable levels of engagement, but for the most part, we are unaware of the distinctions in any way that makes them useful. If I ask you if you could borrow money from certain people in your life, you would probably say, “Sure.” If I asked if you could borrow money from everyone in your life, you would say, “Of course not!” If I asked, among those you could borrow from, could you borrow the same amount from all of them (aside from their ability to lend), you would undoubtedly say, “No!” You could probably borrow cab fare from some people you know without a problem, but beyond that you are not sufficiently related. We intuitively and obviously do know these things about our relationships, but most often do not see this knowledge as profound. If I asked you if you could take one of the cab fare relationships and turn it into a $5,000 six-month loan relationship (again, assuming ability to pay), would you know what to do? Hmm, what would you do? The challenge of taking a Six Sigma approach to our working relationships is no less daunting than

Distinguishing a Conscious Approach to Being Conne cted in the Workplace

that. For those of us without the willingness or ability to generate relationships, our response to this challenge can look like “Falling with Style,” as Woody said to Buzz Lightyear in Toy Story, when commenting on his flying ability. We intuitively know that there are levels of engagement with the people around us, whether at home or at work, but we do not necessarily know how to intentionally design relationships to operate at the levels we really need them to be. We sort of either have the relationships we need or we don’t. If we have them, we use them to our advantage. If we don’t, we often concoct some story about how some people are luckier than others, easier to work with than others, etc., and maybe it’s not going to work out for us this time around—i.e., we do our best, and we get lucky often enough to leave us thinking we know what we are doing. Whatever the story is, it frequently comes down to you being some kind of victim in this big game of chance called life. The “victim perspective” is a very complex subject area, and I am going to save it for another day, except to comment that this is a fundamental point of differentiation between the perspective of those who will continue to be successful in the expanding knowledge economy and those who will not. It goes something like this, for those who get the game now: “I will thrive to the extent that I understand and apply the rules of engagement in the world today. I recognize that my intent to deliver value to others is the entry point to engaging effectively with those who are resources to me. I will expect no return to me until I have delivered value to another.” The bound-to-not-be-successful might verbalize this alternate perspective: “The opportunity to work with great people is something I am offered or not; likely it’s the luck of the draw. Hopefully I will be awake if they ever show up. In the meantime, I will work really hard, do as much as I can by myself, and hope that I get halfway decent people to work with.” Maybe this sounds silly to you, and you can’t imagine anyone really saying something like this to themselves, particularly the latter statement. Yet, some version of these statements—even if it remains only a thought—is fundamental to the perspective from which each of us engages the world. I am just asking you to check and see if your point of view isn’t pretty close to one of these. In addition, if you can see where I’m going with this, can you also see how your point of view affects your actions and attitude toward, and willingness to engage with, those around you at work? Now back to the notion of engagement and an approach (Six Sigma or otherwise) to designing relationships. In the post-WWII economy, when

877

878

The Practice of De signing Relationships

“most people” migrated toward steady jobs, their first level of being related to their places of work might have been referred to as Satisfying Personal Economic Needs. This level was pretty much complete when we secured a job, and it might have come as a result of the type of relationship where my abilities and my “I need a job” met up with an opening for “someone who can do this job.” My need was satisfied by fulfilling a need. There wasn’t much negotiation, if any. Since I was in need, I took what was offered. Since the other had a specific need, I offered only what was asked for, and nothing more. There was never any interest in or discussion of my value—that is, what I could provide if given license to do more. Once the position was secured, I could, in many cases, maintain the relationship for many years, again without negotiation, and make a living out of the fruits of my labor as they were arbitrarily doled out to me over time. The biggest risk I faced in this economy was the next level of being engaged, “Getting Along.” Because I had virtually no say about who my co-workers or supervisors were, nor they about me, the luck of the gods of economics came into play. If I was fortunate, I rolled co-workers I could get along with and drew a supervisor who at least liked me. If not, woe unto me because, “I need this job.” Many jobs were very narrow in scope, at times ridiculously so, and limited collaboration was called for. Most of us could “Get Along” with most of us given the limited level of engagement required. It was boring, but it was steady. Probably most of us have wound up in circumstances where we could survive and make a living. There were enough people we got along with for us to get our work done. In addition, while we may have not been the favorite son or daughter of our manager or supervisor, they had enough of a sense of fair play that we got a bump in our favor every now and again. I know all this sounds pretty dismal, and I am offering extremes here to provide emphasis on a point that has been frequently too painful to discuss, given the price we may have personally paid. However, not to worry, because things have changed. Not only has the game changed but also the skills needed to be connected in life have now become the skills to be connected in the workplace, so you can get a double dip! However, like many other things, being connected has at the same time become more freeing and more complex. Connecting (engaging with) and the ability to connect are now equal in value to what you know or can do. In many instances, being related will be even more important than what you know. Peter Drucker has mused that we will likely not see this fundamental truth until the light of “choice” in our lives makes it visible. Other people do not limit the choices available to us; other people are involved in the choices we make as a natural consequence. If this new age is truly the age of “choice,” then it is now our destiny, I believe, to discover just how much choice we really have. If we take on

De signing for Engagement: There Are Levels of Engagement to Be Addre ssed

this learning curve, we will run smack into another fundamental truth: As human beings working in an organization bounded by policies and procedures, we are truly interdependent. To the degree we can willingly choose to appreciate the interdependence, we might leverage this truth. One opportunity we have in our immediate futures is to use our places of work as our classroom. Other people do not limit the choices available to us; other people are involved in the choices we make as a natural consequence. Whether that consequence results in people as assets or people as obstacles is entirely up to us.

Designing for Engagement: There Are Levels of Engagement to Be Addressed Here is a short description of each of the levels I distinguish and I believe need to be addressed when designing for engagement. Allow me first to show you my model, in Figure 2, and tell you the method from which it has been derived:* Design for EngagementTM

RESPONSBILITY

T p EN ap EM gs H G n A i G Th EN ke Ma o t r we Po

en

Generating Connections

Taking Effective Action

Being Coachable and Coaching Others

Appreciating Interdependence Recognizing Myself

VISION

Figure 2: Design for Engagement * A more comprehensive description and discussion of this model and the notions it represents is contained in the book THRIVE, by Mike Cook, published by St. Lynn’s Press, Pittsburgh, PA.

879

880

The Practice of De signing Relationships

These several levels are something I designed one evening when trying to answer the question: “Why do people who are obviously capable, fail to do what they have said they would do?” This is a pretty basic question, yet one I’ve noticed “most people” do not consider deeply. If they don’t get what they have been promised from someone, they a. Stop counting on that person. b. Ascribe the failure to deliver to a character defect. c. Use the outcome to reinforce the comfortable and familiar belief that if you want something done right, you do it yourself. Each of these assignments of cause accomplish something similar; they serve to protect us from future disappointment while unfortunately disconnecting or distancing us from the offending party. See if this addresses the concern you mentioned; if not, feel free to make changes that do. You seem familiar enough with where I am going for me to allow you that freedom. On my way to creating an answer to this question for myself, I started with what may seem to be a naïve assumption: When people say they are going to do something for you, in the moment of saying this, they truly mean to do it. No one ever accused me of being anything but naïve, so why would this be different? To conserve my limited resource budget, I used myself as the control subject in this investigation. You would have to admire my Six Sigma approach here. Using me as the control subject eliminated variances in responses. It was also an inexpensive set-up. I began by asking myself what were the sources of my not doing what I promised to do. I assumed here that I possessed the ability to deliver what I had promised. What follows is an outline and brief description of what I came to understand about myself and since then by inference, what I now believe I understand about the larger category called “people we work with.” I started out with the premise that when I did not do as I had said I would, it involved some sort of character issue. I was surprised to discover that something else was more likely true. A brief aside here: In the expanding global economy, there is definitely an alternative to the “Me” and “My Needs” level of being connected to all aspects of the world I operate in and with. Whereas the old economy’s context was pretty much a one-dimensional relationship—where employers had the upper hand, and the desire for security was the driving motivation of “most people”—the new economy’s prevailing context is rich with the possibility of personal fulfillment…though apparently less certain. This view of richness depends of course on whether you are willing to consider that there is nothing wrong with the sense of uncertainty that pervades the experience of many people in the workforce as the global economy continues to take shape. There are multiple points of view that can lead to there being nothing wrong with this new world. One might be

De signing for Engagement: There Are Levels of Engagement to Be Addre ssed

to consider the situation from a systems perspective, where in a capitalist system, an expanding global economy is a natural outcome of the development of markets and technology over time. Alternatively, you could view it from an egalitarian perspective that says, “It’s about time everyone has increased access to increased opportunity,” and so on. Personally, I don’t think it much matters what you tell yourself as long as you tell yourself something that leaves you fully engaged with the way things are now and are going to be. This may seem like a big leap, yet if it is now the way of the world, all that is at risk is your point of view. Therefore, back to my experiment with myself, the five levels came to me immediately after I released the belief that broken promises were somehow a character issue. In that new light, I saw that my failures to perform were often rooted in something that was missing, not something that was wrong. As I settled into the inquiry, I saw that in many instances, I did not treat my relationships as precious; I often treated them as givens, or as taken for granted. In the context of taken-for-granted-ness, there would also always be understanding, there would be forgiveness, and there might be consequences, but certainly nothing really final. After all, I meant well, even though I did not deliver. That had to count for something; right? What was missing, then, was masked over by what was assumed— that being that sincerity has value, and good intentions should carry much the same weight as good results. If this sounds familiar, it should. This is the theme song of a large percentage of our population who believe working hard renders you deserving of reward. This is a carryover notion from the now bygone era when employers paid mainly for your time, not your performance. These same people believe that jobs in this country should be both created and protected by our government. Yikes! I thought we were capitalists? When I let go of any notions of “deserving” and replaced them with responsibility, it became apparent that many of my working relationships were ripe for disappointment since they had never really been designed; they had either been inherited from previous situations or conceived in a swamp of mutual assumption. As I sat at my desk that evening, I simply asked myself over and over that question: “Why don’t people who are obviously able, do what they have said they will do? What was missing?” The following set of distinct levels to be addressed while designing for engagement rolled out of the fog of my consciousness. I did the inquiry backwards from the failed promise, tracking it to its logical source. I began to see it in terms of levels. I’ll present them with the caution that relationships really cannot be broken apart as neatly as I am describing, but they can be viewed through multiple perspectives. Here are the Big Five in a nutshell, just so you’ll know what the earlier diagram is attempting to describe.

881

882

The Practice of De signing Relationships

Designing for Engagement Level 1: Recognizing Myself The relationship most often stepped over is the one we must have in place before we can effectively begin a process of engaging with the world around us. I am referring to the relationship we have with ourselves. I’ve termed this level of engagement…Recognizing Myself. I do not, of course, mean recognizing myself in the mirror, I mean understanding yourself deeply at many levels, from skills to talents, from strengths to limitations, and especially at the levels of purpose and vision for your life. Here’s an abbreviated description of the level: • Creating my life purpose. (What am I about?) • Creating my life work and work mission. (What is my vision?) • Understanding what matters to me. (What’s my deal?) • Knowing my motivators. (What is automatically important to me?) • Understanding my style. (How do I go about getting things done?) • Knowing my natural work. (What is the sound of my song?) • Understanding and celebrating my talents. • Recognizing and accepting my limitations.

Level 2: Appreciating Interdependence The second level to design for in the new economy is characterized by two-way relating—you and others. The level arises quite naturally. From the moment you distinguish yourself as your desire to contribute and accomplish, your awareness of not being able to do everything alone becomes quite vivid. I’ll focus on your part in this level for now: • Realizing that “independence” is an illusion. • Recognizing and legitimizing differences in styles, values, and competencies. • Seeing my talents as what I have to offer others. • Seeing my limitations as opportunities for connecting and engaging others

De signing for Engagement

• Intentionally meeting the needs of others. • Accepting “interdependence” as the state of being for everything in our world.

Level 3: Generating Connections If you can manage these first two levels, you may begin to see for yourself the next level in any given relationship. You may find yourself asking if something more might be available. Could I make something out of this or any other relationship if I made more of an investment? Thus, the third level begins to emerge because people want to create something together with others. “Goodbye, Lone Ranger. Hello, Collaboration!” You begin to employ the power of commitment as something mutual, something generative, and you are now moving from being circumstantially related to being related on or with purpose. • Sharing visions of a desired future in which we all win. • Leveraging differences. • Developing an authentic desire to know and appreciate the perspective of others.

Level 4: Taking Effective Action When you move to this level of designing relationships, you have crossed a threshold, and you now have the power to add value inside of already working relationships, as well as create new value-adding relationships. The fourth level to be addressed is almost obvious but tricky nonetheless. This and the fifth level may be the ones where the opportunity for a Six Sigma approach is easiest to discern. Once you have “generated” a relationship, you will undoubtedly desire to see if you can make something happen. Something happening comes as an outcome of Taking Effective Action as a way of being related. This level, like the others, has its own set of tools, some that will seem pretty familiar and some maybe not.

Level 5: Being Coachable and Coaching To make something happen together, we must do the following: 1. Say we are going to do certain things by certain times and mean it. 2. Ask for certain things by certain times and mean it.

883

884

The Practice of De signing Relationships

However, how about the question of whether we can do it? Do we have the requisite knowledge, skills, resources, etc.? Is this obvious? Do you always do what you say; do others always do for you what they say? The tricky part comes in when we, either you or I, have said we are going to make something happen and we are not doing what we said we were going to do, or not asking for what we really need. When designing for engagement, this level is where the rubber really meets the road. This is where improvements in performance can begin. Any exchange of commitments introduces the opportunity for measurement and management. With the opportunity for measurement comes the opportunity for assessment and so on. The move to this level is a natural outcome of the previous. When you commit to make something happen, you move toward its happening and you also open the question of its not happening. As we all know when making commitments, about the only thing you know for sure is that the outcome is rarely predictable. Therefore, when it isn’t going the way you predicted, and you know that because you are measuring, when your assumptions prove to be flawed and your plans don’t work, you likely need a coach. When someone isn’t delivering what they said they would, as determined by measurement and not by feeling or opinion, it may be time for coaching. The coach may be you; it may be someone else. This level is called Being Coachable and Coaching. Here are some of its elements. Coaching • Saying what you are up to • Being responsible • Making offers • Committing to active support of another’s success Being Coachable • Inviting support, rather than attempting to get things done exclusively on your own. • Saying what you are up to; allowing others to connect with your purpose; make it personal and not just a circumstance of being at work. • Being accountable for having made commitments, you have not been made to do anything; and for delivering what you have said you would do.

Sugge sted Reading

• Being responsible for your limitations, emotions, and attitudes; no blame or excuses. • Welcoming the opportunity to have your strengths employed. • Being open to the contribution of others. I have looked at a lot of literature over the past 27 years in the arena of interpersonal concerns and the value of digging in to learn more. This article has provided a brief overview of one person’s view of an approach to self-education in the area of developing powerful and continually improving working relationships. There’s not a lot of “how to” here. I am aware of that. My hope is that I have stimulated an interest or at least a curiosity that you can begin to pursue on your own. There is certainly enough literature out there and the opportunity to educate yourself several times over. I have provided a short bibliography of recommended readings to that end. My purpose is to provide a framework from which you can make a connection with other areas of Six Sigma technology. If I have done that, I am satisfied. If you conclude your consideration of my brief presentation here by being reflective toward your everyday circumstances for a moment, I think you’ll have to admit that the ultimate success of any Six Sigma initiative is balanced precariously on the brink of the level of engagement that can be harnessed for the objectives in mind. I assert that historically we have settled for less than the best that people have to offer because we feared, did not know how to control, their differences. The time when that was acceptable practice has passed us by; now we need to appreciate the whole person. To take full advantage of the strengths of others, a fundamental acceptance must take place. We must welcome the entire spectrum of what people bring with them to our workplace, and not merely accept but invite the quirks and traits, interpersonal styles of communication, and even learn to accept the ambiguity of people, situations, work, and environments that sometimes confound and astound us all.

Suggested Reading Practical Skills for Everyone at Work Crucial Conversations: Tools for Talking When the Stakes Are High, Joseph Grenny et al, McGraw-Hill. Thought Provoking Maslow on Management, Abraham Maslow, Wiley. The Fifth Discipline: The Art and Practice of the Learning Organization, Peter Senge, Currency.

885

886

The Practice of De signing Relationships

The Empowered Manager, Positive Political Skills at Work, Peter Block, Jossey-Bass. The Answer to How Is Yes: Acting on What Matters, Peter Block, Berrett-Koehler.

Logical, Thought Provoking, and Humbling Out of the Crisis, Edwards Deming, The MIT Press. Inspirational Stewardship: Choosing Service Over Self Interest, Peter Block, Barrett-Koehler. Leadership Is an Art, Max DuPree, Dell. Transformational Freedom and Accountability at Work: Applying Philosophic Insight to the Real World, Peter Block, Pfeiffer. Educational Now Discover Your Strengths, Marcus Buckingham and Donald O. Clifton, Free Press. Management Challenges for the 21st Century, Peter Drucker, HarperCollins.

About the Author Mike Cook lives with his wife, Pat Jackson, in Anacortes, Washington on Fidalgo Island, north of Seattle. Mike and Pat have four grown children, one daughter and three sons, plus a granddaughter in Portland, Oregon. Mike has worked as an independent consultant and later founder of Vitalwork for over 19 years. In that time, he has focused his energy on illuminating the things people do to and with each other that makes their life at work less than satisfying. THRIVE is Mike’s first book. He intends to continue writing, especially with an intention to create tools that support individuals in mastering the skills of self-management.

A Process for Product Development By Bill Jewett

This article describes a process for product development with depth to enable the understanding of activities by which marketing and service work with engineering and management. Processes for advanced product planning and for technology development are also described to characterize the work prior to the beginning of product development. These processes are broadly applicable to a wide range of product types and development projects. Many of the activities and deliverables can be enhanced by the application of better tools and methods. However, to be useful within a company, the processes need to be articulated with terminology familiar to the business and with details relevant to the specific technologies, products, and business. Many companies invest in the development of new products. Those who are consistently successful follow processes that are standardized across their projects and adapted well to the characteristics of their business model. Their plans for future products enable new technical concepts to be developed in anticipation of the needs of those future products. Likewise, they develop organizational and functional capabilities that help them to achieve higher quality and lower costs in their new products, with projects that have more predictable and shorter cycle times. Whether or not their processes are good ones is an open question for many companies. Likewise, whether or not a company actually achieves excellence in following their processes is also a question for assessment. Those companies that do not have standardized processes or do not follow them well tend to lack consistency in how they approach development work. This makes their projects difficult to manage and vulnerable to large variations in outcomes. Efforts to improve the results find it difficult to understand systemic problems and to make changes that can achieve sustainable advancements. Product development is inherently cross-functional. It integrates the activities of engineering disciplines such as software, mechanical, electronic, optical, and materials. There are parallel development activities by various organizations, such as marketing, service, packaging, manufacturing, and by other companies that act as partners or suppliers. It is a process of creating value-adding solutions to new, unique, or difficult requirements. The value to be delivered to customers may be achieved by the engineering that goes into the product designs and their packaging, or into their manufacturing processes. It may be provided through product service and customer support after the product is in the market. 887

888

A Proce ss for Product Development

Customers may also receive value during the purchasing experience from effective information provided by marketing and sales. However, with most markets being competitive, customers have alternative solutions to compare. They choose those products and services that they expect will deliver the most benefits for the least costs—that is, the highest value. So, the goal of your product development project is to deliver value to your target customers that will be superior to alternatives available to them after the market entry of your product. Product development is also viewed as the creation of a new or incremental business, as represented by the new products and services. The business objective of a project is to return value to the corporation that is not only acceptable, but also better than that expected from the next best alternative investment. Although this value is usually thought to be financial, it may also reflect other objectives such as to establish a market presence or to obtain feedback to benefit a new product platform. When designed to be suitable to the internal culture, the product development process establishes consistent, more predictable methods to accomplish the development of new products and their related services. With the integration of internal and external knowledge, and the lessons learned from experience with previous development projects, it enables superior quality, costs, and timelines to be achieved. With excellence in execution, it provides sustainable competitive advantages. Descriptions of the popular development processes can be found in selected references. In their generalities, they are similar phase/gate models, although their phase (or stage) descriptions vary in detail and emphasis. For example: • PACE® from PRTM1 stands for Product and Cycle-time Excellence. It employs five phases to cover the timeline, providing a framework for functional details to be added by client companies. • Stage-Gate®, developed by Robert Cooper,2 uses the term “stage” to refer to logical groups of activities. It also employs five stages, with marketing and financial elements well represented. • Don Clausing3 also uses four phases to describe the development of new products, with an emphasis on activities and methods particular to how engineering and marketing can work together. 1. Michael E. McGrath, Michael T. Anthony, and Amram R. Shapiro, Product Development: Success Through Product and Cycle-time Excellence. Boston, MA: Butterworth-Heinemann, 1992. ISBN 0-7506-9289-8. 2. Robert G. Cooper, Winning at New Products, Accelerating the Process from Idea to Launch. Reading, MA: Addison-Wesley Publishing Company, 1991. ISBN 0-201-56381-9. 3. Don Clausing, Total Quality Development: A Step-by-Step Guide to World Class Concurrent Engineering. New York: ASME Press, 1994. ISBN 0-7918-0035-0.

COM Revisited

• CDOV©4 describes the process in four phases, with a couple subphases to capture sequential details. It places even more emphasis on details relevant to engineering development. Figure 1 shows a comparison of these processes with an alignment of the phases based on the similarity of the activities that are described by their sources. Product and Cycle-time Excellence (“PACE”®) Phase 0: Concept Evaluation

Phase 1: Planning and Specification

Phase 2: Development

Stage 1: Preliminary Investigation

Stage 2: Detailed Investigation

Stage 3: Development

Phase 3: Test and Evaluation

Phase 4: Product Release

Stage 4: Testing and Validation

Stage 5: Full Production and Market Launch

Stage-Gate® Process

Basic Concurrent Engineering described by Don Clausing Concept Phase

Design Phase

Production Preparation

Production

Concept – Development – Optimize – Verify (“CDOV”©) Phase 1: Superior Product Concept (C)

Phase 2: Baseline Design (D)

Phase 3A: Subsystem Robustness Optimization (O)

Phase 3B: System Integration, Stress Testing and Balancing (O)

Phase 4A: Product/System Design Capability Verification (V)

Phase 4B: Manufacturing Process and Supply Chain Capability Verification

Figure 1: Published Product Development Processes Are Comparable Although Somewhat Different in Terminology and Emphasis

These process descriptions all have their merits. There’s no claim here that one is better than the others. To provide competitive advantages, they all depend on being adapted to the business model of a company, with insight to those elements that contribute to higher quality, lower costs, reduced risks, efficient project management, and effective use of resources. There are no magic remedies here. The processes all expect rigorous involvement with customers and suppliers. They expect functional excellence, aligned to achieve improved business results. They all benefit from high-performance teams working with enabling support from management. Descriptions in this article provide more details for the development process with a slightly expanded scope. They emphasize the development of solution concepts prior to their design for production. Marketing and service capabilities are developed in parallel to the development of 4. C. M. Creveling, J.L. Slutsky, and D. Antis, Design for Six Sigma in Technology and Product Development. Upper Saddle River, NJ: Prentice Hall PTR, 2003. ISBN 0-13-009223-1.

889

890

A Proce ss for Product Development

the product and its manufacturing processes. Activities generate information that is cross-functional in content and evolves in maturity and completeness. Progressive freezing of partial information enables new work to begin while refined information is being developed. The descriptions extend to the processes for advanced product planning and for technology development, which provide a solid foundation under the process. They enable product development projects to begin when necessary and have shorter cycle times. Although it is not a topic for this article, these processes have the umbrella of portfolio management, since development projects are managed in the context of other development projects, all competing for resources and management attention. Figure 2 shows an example of a portfolio of product development projects. It depends upon a stream of new design concepts and technologies developed in anticipation of the needs of product development. There are also projects to develop technical resources and organizational capabilities to add value to product development. The initiative for those advanced developments comes from the plans established by the work of advanced product planning. Advance Product Planning: market situation analyses, early Voice of the Customer, product platform plans, product line plans Portfolio of projects to improve capabilities

On-going, consistent investments

Technology Development: new or improved design concepts, manufacturing processes, enabling technologies Development of functional capabilities: new or improved skills, tools, processes, infrastructures, facilities, partnerships, channels, suppliers

Anticipation of the needs of product development projects

Plans, design concepts, enabling technologies, functional capabilities to enable product development projects

Development of Product A Portfolio of Projects to Develop New Products

Product Management for Product A

Development of Product B Development of Product C

Project charter decisions

Product Management for Product B

Portfolio of Products in the Market

Product Management for Product C

Product launch decisions

Figure 2: The Portfolio of Product Development Projects Is Enabled by the On-going Investments in Improved Capabilities

As illustrated in Figure 2, each product development project is managed in the context of a portfolio of development projects, chosen for

Advanced Product Planning Proce ss

their potential value to customers and for their expected returns to the corporation. The basic foundations for these forecasts are in the analyses of competitive markets and their value drivers, in the financial analyses of the proposed business opportunity, and in the capabilities of the future products to be developed. The ability to develop these products successfully is in the functional and technical capabilities of the organization, its partners and suppliers, and in the design concepts and technologies that can be developed, or acquired, and commercialized. These on-going planning and development activities can have dramatic benefits to product development projects either by improving their chances of being successful or by changing their direction as market or technology factors change.

Advanced Product Planning Process What used to be thought of as the “fuzzy front end” does not have to be fuzzy. There are specific activities to be staffed and managed, information to be gathered, plans to be developed, and decisions to be made. Longrange plans that are developed for product families enable the authorization of the following: 1. Product development projects beginning at the right time. 2. Technology development (or acquisition) projects in anticipation of

the needs of those future products. 3. Long range, integrated investment plans that enable the develop-

ment or acquisition of technologies, products, and manufacturing processes at the right time. It starts with basic decisions that clarify the vision and strategies of the business and the target markets it chooses to serve. It draws conclusions from evaluations of the following: • Value drivers for selected market segments • Unfulfilled needs and preferences of target customers • Competitive situations in those market segments • Trends in competing technologies, including expectations for obsolescence • Business opportunities for which you have sustainable competitive advantages • Creative ideas for new products and services to offer to those markets

891

892

A Proce ss for Product Development

The company must have a vision for the business and a strategy to achieve it, with which the plans need to be aligned. A series of differentiating value propositions can then characterize its intent to serve the customers in those markets in ways that will be perceived as being superior to available alternatives. This takes a lot of knowledge and insight, not guesswork. The winners include those who see the market opportunity first. Figure 3 illustrates a linkage of activities as an on-going process, rather than one with a phase/gate structure. Information is gathered and evaluated continually, with adjustments made to the plans for portfolios and projects as needed. The work is the responsibility of those people assigned to study markets and customers, to establish plans for new products and capabilities, and to develop or acquire new technical concepts. Product family plans are maps of solutions to initial, high-level requirements that are derived from the value propositions. The capabilities that the products are challenged to commercialize are the sources of the needs for new design concepts or manufacturing processes, with their enabling technologies. They are also the sources of plans to leverage existing platforms for applications to new market needs. The timelines for the various product development projects, working backwards from the required market entry milestones, determine when each project needs to be started. That then determines when a new or improved technical capability needs to be ready for transfer into a product development project. The passage of Gate 0 provides the project with a deliberate charter and the necessary allocation of resources. This must be early enough so that the “window of opportunity” for market entry can be achieved, with the value proposition implemented by an acceptable development effort. The activities that develop the information need to be funded, staffed, and managed consistently. If this information is not developed in advance, it will be the burden of product-specific projects, adding time, costs, risks, and potential conflicts to each project. These are highly dependent plans. Given the risks that they embody, alternative approaches and contingency plans must be included, such as more than a single new design concept to be developed to answer a particular new requirement. As illustrated in Figure 4, projects start at times that depend on their development cycle time and the target date for market entry. They depend on resources being available and on required technical capabilities being ready. The roadmaps of technical concepts are plans for the migration of new capabilities into products and for alternative solutions from which the best ones are chosen for the product’s applications. Figure 4 uses an example for electrophotographic printers to illustrate the idea.

Advanced Product Planning Proce ss Clarify the Strategic Vision

Define Market Segments with Value Drivers

Clarify the Participation Strategy

Study the Competitive Market Situation

Identify Superior Business & Technical Opportunities

Select Market Segments & Customers to be Served

Translate into “initial” New or Difficult Product Requirements

Study Customers’ Value Drivers, Needs & Preferences, Images

Develop Differentiating Value Propositions

Clarify the Innovation Strategy

Generate Ideas for Solutions

Portfolio Plans Phase 0 Develop Product Platform Plans

Develop Product Family Plans

Create Technology Development or Acquisition Plans

Technology Development

Charter the Development of a New Product Concept Product Development Project

Develop Integrated Investment Plans

Develop or Acquire Resources for Commercialization Programs

Input to Annual Funding Budgets

Develop or Acquire Robust New Technical Capabilities

Assess the Robustness of Required New Technologies

(on-going stream of new capabilities)

Figure 3: Advanced Product Planning and Technology Development Enable the Best Product Development Projects to be Chartered Early Enough and To Be Enabled by New Technologies that Are Robust for the Application Expected Product Development Projects High Quality, Black, 2-sided Images

Heavy Weight Papers Higher Quality, Color, 2-sided Images

Product Family Plans

Medium Speed Printer Platform

Magnetic Image Characters

High Quality, Black, 2-sided Images Heavy Weight Papers

High Speed Printer Platform

Very Higher Quality, Color, 2-sided Images

On-going Technology Development Projects Toner Materials Toning Process Image Transfer Process

Technology Roadmaps

Product capabilities are dependent on these new design concepts and their enabling technologies, mapped to the timeline of the project, funded and managed.

Time

Cleaning Process Fusing Process Legend: Paper Handling Development milestones with follow-up refinements

Project Start Development Milestones Market Entry

Figure 4: Examples for Technology Development Plans Derived from Product Plans for Electrophotographic Printers

893

894

A Proce ss for Product Development

The objective is to develop a stream of new capabilities that will deliver improved value to customers and return acceptable value to the corporation. The tasks of technology development are guided by a phase and gate process, although not as cross-functional as that for product development. The information developed contributes to early business plans specific to the concepts being developed, and as such needs support from marketing and financial expertise.

Technology Development Process The objective of technology development is to prepare a stream of new or improved technical capabilities to be robust for a range of future applications and stressful conditions. Those applications are the ones described by the product family plans and their related technology roadmaps. The new capabilities may be commercialized in new products or manufacturing processes. Robustness developed early may be critical to the success of those future products. The technical capabilities that are to be developed may be at the level of a design component or material, or its manufacturing process, or at the level of a subsystem design concept, or even a full-system product architecture. The technology development process applies to what you might call “advanced technology development,” or “advanced research and development.” To clarify, it is not intended to apply to the processes of basic research or invention. Those tend to be much more open-ended processes without a timeline that can be managed or predicted well. The process represented here is one to develop technical capabilities to serve future projects that have planned timelines. The process for technology development has four phases, shown in Figure 5 by alphabetic characters to differentiate them from those of product development. Gate decisions judge whether or not acceptance criteria, specific to the phase, have been satisfied well enough to enable the project to move into its next phase. The following table describes the objectives and acceptance criteria for each technology development phase.

Te chnology Development Proce ss Technology Development gate decisions

Technology Development activity phases

Alternative Design Concepts

Gate A: Select candidate concepts for further development

Phase A: Clarify Requirements

Gate C: Select the best design concepts for transfer

Gate B: Screen out or postpone immature design concepts

Phase B: Develop Performance

Phase C: Develop Robustness

Phase D: Document Technology

Single-pass Duplex Fusing

Single-pass Duplex Fusing

Single-pass Duplex Fusing

Hybrid External & Internal Heating

Hybrid External & Internal Heating

Hybrid External & Internal Heating

Hybrid External & Internal Heating

Externally Heated Rollers

Externally Heated Rollers

Externally Heated Rollers

Externally Heated Rollers

High Intensity Internally Heating

High Intensity Internally Heating

High Intensity Internally Heating

Transfer design concept to new product development project

Preserve technology for later application

(Examples for illustration only) The development of single-pass, duplex fusing was postponed due to difficulties elsewhere in the process.

The life expectancy of high intensity lamps was difficult to achieve.

The robustness of the hybrid concept was demonstrated to be superior to that of rollers heated only externally for the new products’ applications.

Figure 5: The Technology Development Process Has Four Phases That Manage the Development of Performance and Robustness for a Range of Future Applications.

Technology Development Process Phase A: Clarify Requirements

Objective:

Translate the value proposition and the initial requirements of future product applications into requirements at the level of the concept or technology to be developed. Select candidate concepts for development as potential solutions to those requirements.

Acceptance Criteria:

To be selected for development, a technical concept must be judged to have a reasonable probability of being able to contribute to the value delivered to future customers, as well as to contribute to the value returned to the corporation.

Phase B: Develop Performance

Objective:

Develop the performance of the selected technical concept to be controllable and to achieve its requirements over the range of intended applications.

895

896

A Proce ss for Product Development Acceptance Criteria:

The nominal performance of the technology must be controllable by functional parameters that are available and practical. The technical concept must continue to be a good candidate for future commercialization, satisfying criteria such as intellectual property rights, expected manufacturing process capabilities, and costs.

Phase C: Develop Robustness

Objective:

Develop the robustness of the selected technical concept to be controllable under those stressful conditions that are expected.

Acceptance Criteria:

The new technical capabilities must be superior in performance and robustness to: • Alternative concepts available for commercialization. • The expected competition in the future markets.

Phase D: Document the Technology

Objective:

Prepare the information, equipment, and other resources for transfer to a product development project.

Acceptance Criteria:

There is no decision for Phase D. The responsibility to select a new or improved technical concept is with the subsequent product development project. If the concept is not selected for commercialization, the technical reports, specifications, and development prototypes will be stored for potential future uses.

If a new technical design concept or manufacturing process is developed to be robust, it will have controllable parameters that are practical for manufacturing and effective at reducing the vulnerability to stressful conditions. Some of these parameters will be effective at adjusting the average performance to its required target value. Ideally, the results of technology development will be sufficient to enable the product-specific development to focus on tuning the adjustable parameters to be appropriate for the subsystem and later to optimize system integration. In reality, there may need to be further development of the technology in the context of the product-specific configuration and stresses. The specifications of those functional parameters must be transferred into the product development project. Technology transfer would include test reports, concept descriptions, risk assessments, and other information important for integration into a product system or manufacturing

Product Development Proce ss

process. It may also include development breadboards and critical personnel. In the process for product development, the selection of the baseline design concepts occurs in Phase 1.

Product Development Process The products to be developed can be expected to vary widely, from highly leveraged product enhancements to new platforms with many new design concepts to be integrated and optimized. The content may be software only, and thereby not have the development of new manufacturing processes or early commitments to tooling. On the other hand, the projects may need to develop systems of hardware, software, and materials that are highly complex and integrated. For some products, the manufacturing processes may need to accommodate very high production volumes, with short process flows. For others, the production volumes may be modest or the manufacturing flows relatively long. This range of project characteristics places a challenge on the product development process to be scalable so that it can guide both large and small projects. The range of potential technical content demands the process to be flexible enough to adapt its principles to the product-specific challenges. Key strategies and their principles must be incorporated in order to provide competitive advantages in achieving the objectives for quality, costs, and cycle time. Design for Six Sigma and Lean Product Development are two such strategies. In addition, your product development process must work with the processes of other business units or companies, acting as collaborative partners or suppliers. It can be expected that no two development projects are the same in technical difficulties, business challenges, or project risks. It makes sense, then, to have a standardized process written so that all organizational functions can understand its relevance to their responsibilities. In Phase 0, the project leadership decides how to adapt the standard process to the characteristics of the particular project. They identify those expectations of the process that are either not relevant or are already satisfied, as with a leveraged design. They may expand the process with additional activities, reviews, and decisions where necessary to accommodate unusual characteristics that may not be clearly addressed in the standard process. With their responsibility to be efficient and to meet their project milestones, they also decide where activities can be accelerated or started earlier. Likewise, the needs for rigorous discipline can be identified, which may place certain reviews and decisions on the critical path. Figure 6 shows the same comparison of processes as in Figure 1, with an additional outline that is described below. Additional phases are valuable to describe cross-functional objectives with more details and to manage risks more closely.

897

898

A Proce ss for Product Development Product and Cycle-time Excellence (“PACE”) Phase 0: Concept Evaluation

Phase 1: Planning and Specification

Phase 2: Development

Stage 1: Preliminary Investigation

Stage 2: Detailed Investigation

Stage 3: Development

Phase 3: Test and Evaluation

Phase 4: Product Release

Stage 4: Testing and Validation

Stage 5: Full Production and Market Launch

Stage-Gate Process

Basic Concurrent Engineering described by Don Clausing Concept Phase

Design Phase

Production Preparation

Production

Concept–Development–Optimize–Verify (“CDOV”) Phase 1: Superior Product Concept (C)

Phase 2: Baseline Design (D)

Phase 3A: Subsystem Robustness Optimization (O)

Phase 3B: System Integration, Stress Testing and Balancing (O)

Phase 4A: Product/System Design Capability Verification (V)

Phase 4B: Manufacturing Process and Supply Chain Capability Verification

Elaboration on the Cross-functional Product Development Process Phase 0: Clarify business objectives; charter the project

Phase 1: Clarify the requirements select and develop architecture and subsystems

Phase 2: Complete the requirements optimize system development

Phase 3: Design product specifications, processes for manufacturing, service, sales. customer support

Phase 4: Verify product designs; begin production and supply chain preparations

Phase 5: Phase 6: Phase 7: Prepare for Verify readiness Launch production for market entry production and product launch, customer sales, service and support customer support

Figure 6: To Be Useful, Product Development Processes Must Have Details that Guide Cross-functional Activities, Dependencies, and Deliverables to Satisfy Phased Acceptance Criteria

These descriptions do not provide a sufficient statement of a new product development process. To be useful in practice, it must be adapted to the character of your business model with details and flexibility added to be most relevant to the challenges in your projects. Although the language may be more familiar to mechanical and electrical engineering, it must be easily interpreted for the development of software, materials, and the other technical elements of the product system. It may be that the real challenges in a product development project are for new manufacturing processes or facilities, so the cross-functional guidance must be clear and viewed to being just as important as the product design. For that matter, the real challenges may be for marketing or service capabilities. By any way of looking at it, the process must be inherently cross-functional in its objectives, activities, deliverables, acceptance criteria, and decisions. It must integrate business and technical deliverables so that the operations after market entry satisfy the financial expectations of the corporation. Similarly, the product must satisfy the market’s expectations for features, functionality, performance, reliability, usage life, costs of ownership, and any other characteristics that customers use in purchasing products and judging their value. That places high expectations on the development teams and their leadership.

Product Development Proce ss

Let’s look at the structure of the process. A quick view of Figure 7 shows that: a. The phases are expected to overlap, with activities starting when they are enabled by information and resources rather than when a gate review is passed. b. The gate reviews are not equal in importance (smaller diamonds) and are not necessarily on the critical path. Caution should be used in overlapping phases too much. Risks can increase when commitments are made prematurely and later changes force rework that can be costly and time consuming. Suppose, for example, that commitments to tool designs are made before part designs are completed. The scrapping or rework of tools and prototype parts can be painful. Once chartered by Gate 0, development programs follow six development phases to market entry, with a seventh phase to resolve post-launch problems and to obtain feedback from the market. In this description, the “number” of the gate decision review is that of the phase that it follows. Some companies use the number of the phase that the gate precedes. That may be to emphasize that the passage of the gate enables or even authorizes the work of the next phase. Phase 0

Phase 1 Develop Requirements

Charter the Product

Develop Subsystems Phase 2 Gate 0

Project Governance Gate Decisions with Acceptance Criteria

Optimize Systems Develop Implementation Plans Phase 3

Phases of Cross-functional Activities and Deliverables

Design Product and Manufacturing Processes Develop Marketing/Service

Phase 4 Verify Product Designs

Formal Authorization of Market Entry Phase 5

Phase 6

Develop and Verify Capabilities for Production, Distribution, Service, Customer Support, Marketing and Sales

Verify Readiness for Market Entry

Post-Launch Review Phase 7

Scale-up Production, Supply Chain, Sales, Service and Customer Support

Figure 7: Product Development Is Managed with Overlapping Phases from the Project’s Charter Through the Product’s Launch into Production

899

900

A Proce ss for Product Development

Phase and gate structures for product development are often viewed as being too sequential and bureaucratic. The criticism argues that the work of a phase cannot begin until there’s completion and approval of the work of the previous phase. In the process described here, project teams should start activities when the prerequisite information is available and trustworthy, and when the risks of rework due to late arriving information are low. With similar license, project teams may see advantages in combining phases or in managing less important gate reviews by more informal means. This flexibility may provide benefits in reduced cycle time and development costs, without adding unacceptable risks. It all depends on the circumstances of the project. The difficulties are in the potential for added risks. There’s no universal right answer. It must be the judgment of the project leadership and their governance bodies to identify those adaptations of the standardized process that are appropriate and acceptable for a specific development project. Figure 8 shows a view of product development as a process of creating information. Data, analytical conclusions, forecasts, and implementation plans, for example, evolve and mature over the early phases, illustrating that the initial development work has to be accomplished with partial information. Progressive freezing forces the preliminary information to be stable while future work adds refinement and completeness. Requirements mature from “preliminary” in Phase 1 to “final” in Phase 2. Controllable parameters evolve from critical parameter specifications in Phase 2 to product designs for production in Phase 3. Project management plans are preliminary in Phase 1, while in Phase 2 they are complete enough to enable the project team to commit to a market entry date. Likewise, the business plan evolves to be complete, approved, and committed to by the end of Phase 2. You also see negotiations among the deliverables of information, indicated by the bidirectional arrows. Not all product requirements can be achieved by the available design concepts. Functional implementation plans may have to be changed to establish project management plans that will achieve the market entry milestone. An acceptable business plan may need substantial changes to the design concepts or marketing plans. This information is highly linked, since there are many dependencies among the deliverables. For example, the revenue forecasts in the business plan depend on the market entry date. They also depend on how well the product requirements represent differentiated and superior value to customers, and how well the developed product will satisfy those requirements. The pricing inputs to the financial model will depend both on the competitive position of the product and on the manufacturing costs, which in turn are enabled by the product designs and achieved by the manufacturing processes and supply chains.

Product Development Proce ss Phase 0

Phase 1

Customers’ Needs & Preferences (advanced)

Customers’ Needs & Preferences (detailed) Program Requirements (initial)

Value Proposition

Phase 2

Validation of Requirements Program Requirements (prelim.)

Product Concept Descript’n

Program Requirements (final)

Baseline Design Descript’n

Project Management Plans (initial)

Business Plan (initial)

Product Performance Spec’ns

Critical Parameter Spec’ns

Functional Implemen’n Plans (prelim.)

Portfolio Plan

Phase 3

Product Designs

Functional Implemen’n Plans (final) Project Management Plans (prelim.)

Business Plan (prelim.)

After completion, documents are approved & under change control. Project Management Plans (final)

Business Plan (final)

Figure 8: Highly Linked Information Is Developed, Clarified, and Validated Across the Early Project Phases

The following overview describes the phases of product development. The work of Phase 0 determines whether or not the project should be chartered and resources allocated to it. Gate 6 is the formal approval to release the product to revenue-bearing customers. The gate reviews separating the phases may be thought of as quality checkpoints for the project. However, they will provide higher value if designed to focus not so much on the completeness of deliverables, but more on the value developed for customers, and for the corporation, and the readiness to accomplish the work of the next phase with less risk. They are decision processes that are aimed at guiding the direction of the project and ensuring that the project has a high probability of success.

901

902

A Proce ss for Product Development Product Development Process Phase 0

Objective:

Decide whether or not to charter the proposed product development project and to allocate the resources required for Phase 1.

Activities and Deliverables: • Clarify the opportunity to deliver superior value to target customers. • Describe the visions for the product, with its value proposition and related high-level requirements. • Propose candidate design concepts, enabling technologies, and functional capabilities. • Evaluate the strategic fit of the product concept and its enabling technologies. • Develop project management plans for Phase 1, with rough estimates of plans to achieve the window of opportunity for market entry. • Recommend project-specific adaptations to the standard product development process. • Develop an initial business plan, with financial analyses and forecasts, a project risk assessment, and key lessons learned to be implemented. Acceptance Criteria:

The decision should be based on analyses and forecasts that show that the: • New products and services are expected to deliver value to target customers better than alternatives forecast to be available to them after our market entry. • The project is expected to return value to the corporation better than the next best alternative investment. • The project is prepared with knowledge, technical concepts, resources, funding, and so on to achieve the objectives of Phase 1 within an acceptable timeline.

Phase 1

Objective:

Develop an achievable business plan derived from the preliminary requirements, the selected baseline design concepts, completed subsystem-level development, and the selected approaches for manufacturing, distribution, marketing, and service.

Activities and Deliverables: • Clarify and validate the “Voice of the Customer.” • Translate the needs and preferences of target customers into achievable requirements for new products and services. • Incorporate requirements derived from corporate, industry, and regulatory sources.

Product Development Proce ss • Decompose system-level requirements to the level of subsystems, components, software, materials, and so on. • Select the baseline system architecture, subsystem design concepts, and their enabling technologies. • Complete the development of new concepts to be robust for the product applications. • Select the baseline approaches for manufacturing, product service, customer support, marketing, and distribution. • Develop functional implementation plans with integrated schedules, dependencies, milestones, and resources. • Develop an achievable project management plan for Phase 2, with an update of the estimate of market entry date. • Develop a preliminary business plan with updated financial analyses, forecasts, and risk assessments. Acceptance Criteria:

Preliminary requirements: • Represent a superior and differentiating value proposition. • Are complete enough to support the selection of the baseline design concepts for the product. Baseline system architecture, design concepts, and enabling technologies: • Are chosen to be the best available solutions to their requirements. • Have been demonstrated to be robust for the product-specific applications. Preliminary project management plans, integrating the functional implementation plans and risk management plans, are judged to be achievable and acceptable. Preliminary business plan is judged to be acceptable.

Phase 2

Objectives:

Develop the capabilities in the product to deliver differentiated, superior value to target customers. Develop implementation plans that will return superior value to the corporation. Prepare the project to accomplish the activities leading to market entry with fast, deliberate implementation, with a high degree of predictability, and with the support of organizations, partners, suppliers, and channels committed and aligned to achieve the requirements of the project. Commit to the target date for initial market entry.

continues

903

904

A Proce ss for Product Development Activities and Deliverables: • Complete the requirements for the new product and its services, with validation by customers. • Integrate subsystems, modules, and accessories to develop full-system configurations. • Validate system architectures and design concepts with feedback from customers. • Optimize system-level performance and robustness, with specifications of critical functional design parameters. • Complete and approve functional implementation plans. • Complete project management plans through product launches into production. • Complete the business plan, with acceptable financial forecasts and risks. Acceptance Criteria:

Requirements are approved, frozen, and under change control. System-level product performance and robustness are optimized. Project management plans are approved and supported by resources through product launch and production start-up. Product launch dates for various markets are acceptable. Project-level risks and their management plans are acceptable and supported by resources. Project business plan is approved and committed to.

Phase 3

Objective:

Complete and release the final designs of the product system.

Activities and Deliverables: • Develop product designs with feedback from internal reviews and from trade-offs to achieve required manufacturing and service capabilities. • Test and refine design prototypes to comply with product requirements and to be robust under stressful conditions. • Upgrade product designs with feedback from customer acceptance tests. • Complete the specifications and release the product designs with cost-effective manufacturing tolerances. • Develop production and supply capabilities. • Develop capabilities for product service, customer support, and marketing. • Develop product launch plans for target markets. • Ensure that the expectations for the product are aligned with the business plan.

Product Development Proce ss Acceptance Criteria:

Product designs satisfy their requirements. Customer feedback has validated the product designs. Product prototypes are updated and ready for verification tests. Manufacturing suppliers, processes, and materials have been selected. Manufacturing processes have forecasts of acceptable quality, costs, and cycle times in production. Service processes are expected to have acceptable quality, costs, and cycle time.

Phase 4

Objective:

Verify that the product designs satisfy their requirements.

Activities and Deliverables: • Begin the procurement of production parts, components, materials, and so on. • Test the full-system product prototypes to demonstrate compliance with their requirements. • Identify design improvements required prior to market entry. • Develop marketing communications to describe the expectations for the new product and the timelines for its launch into target markets. • Ensure that the expectations for the verified product are aligned with the business plan. Acceptance Criteria:

Unbiased tests have demonstrated that the released product designs: • Are acceptable to customers. • Satisfy their product requirements over the expected range of applications and operating conditions. • Have a reliability growth rate that is acceptable. • Satisfy requirements for manufacturing and service quality, costs, and cycle time. • Are superior to their expected competition. Plans for corrective actions required by product launch are supported by resources.

Phase 5

Objective:

Prepare for market launch and for sustained production, sales, and product support.

continues

905

906

A Proce ss for Product Development Activities and Deliverables: • Develop the production and supply chain capabilities and capacities. • Improve and verify manufacturing processes to achieve acceptable quality, costs, and cycle times. • Verify and manage corrective design changes that are required prior to market entry. • Manage the processes for regulatory and statutory approvals. • Develop the capabilities for product distribution, service, and support with feedback from early customer trials. • Develop the readiness for marketing and sales. • Complete preparations for product launches into selected markets. • Ensure that the expectations for the production and support processes are aligned with the business plan. Acceptance Criteria:

Corrective actions required for market launch have been implemented. Products built with production tools and methods satisfy their requirements. Factory and supply chain operations are stable and repeatable, and: • Satisfy launch criteria for quality, costs, and cycle time. • Have capacities to meet the demands of production scale-up. Service and customer support preparations are complete and have sufficient capacity to support initial customers.

Phase 6

Objective:

Authorize the delivery of products to revenue-bearing customers.

Activities and Deliverables: • Demonstrate the product’s performance reliability, usage lives of consumables, and so on to provide updated expectations to customers and to sales, service, and support organizations. • Achieve necessary regulatory and statutory approvals. • Confirm that the corporate requirements are satisfied for target markets. • Confirm that the product and its supporting processes are ready for market entry. • Obtain corporate authorization for the release of the product to revenuebearing customers.

Product Development Proce ss Acceptance Criteria:

Manufacturing, procurement, supply chain, and assembly processes are ready for production. The product is available to customers in sufficient quantities. Support organizations are prepared to provide service and customer support in the initial markets. Marketing and sales capabilities and capacities are ready for product launches into target markets. All regulatory and statutory requirements have been satisfied. All corporate requirements have been satisfied. The product and its supporting organizations are prepared to deliver the value proposition to target customers.

Phase 7

Objective:

Manage the transition into stable, routine production and product support.

Activities and Deliverables: • Manage the scale-up of production and supply chain operations. • Develop and verify improvements to the quality, costs, and cycle time for production and supply chain processes. • Manage improvements to product service and customer support. • Develop feedback from product service and customer support to guide product improvements and requirements for future products. • Evaluate how well manufacturing, sales, service, and customer support have satisfied the expectations of the business plan. • Identify lessons learned that can benefit future product development projects. Acceptance Criteria:

The production processes are mature and stable, meeting requirements for quality, costs, and cycle time. Feedback from the market and customers no longer demands improvements to the current product. The financial returns are aligned with the expectations of the business plan. Remaining corrective actions can be handled by routine production and product management initiatives.

After the start-up of production and the entry of the product into its target markets, a post-launch review is very valuable. It is a formal opportunity to see how well the forecasts of progress in the business plan have

907

908

A Proce ss for Product Development

been achieved and to initiate corrective actions if necessary. Feedback from customers can identify improvements required to the current product or for follow-on products under development. It also is an opportunity to identify lessons learned that can benefit future projects. During Phase 7, responsibility for the product configuration may be shared between product development management and production management. Product development is not just a process of turning engineers loose on challenging problems. It is an investment in the development of a new business. It has customers willing to spend money for solutions to their problems. Those customers look for the solutions to be better than alternatives available to them. It has investors—that is, those managers of corporate funding and other resources who control portfolios of investments and who are challenged to achieve superior returns on those investments. It has the minds of people who have to integrate technical and business capabilities into a well-functioning value delivery system, available at the right time. Many people have a stake in the success of the project and how well it uses its investments. There are many decisions that have to be made. They add value through their effective implementation and their intended consequences. As with the activities, there are expected to be standardized processes for making decisions, with the scalability and flexibility to be adaptable to the range of project characteristics.

Project Decisions Product development decisions can range from the choice of the project objectives, to the selection of requirements that are achievable and that will define a differentiated and superior product. They select the design concepts or architectures to be developed. The many parameters in the project management plan and the business plan are the results of decisions that establish target metrics of an acceptable business and how they will be achieved. Risks that are identified require decisions about whether or not and how they will be mitigated. Decisions include whether or not to charter a project, or later to cancel it. The quality and timeliness of these decisions and the effectiveness of their implementation have dramatic consequences for a development project. The right decisions made at the right time can enable teams to achieve higher-quality products at lower costs and with on-time market entry. However, a company’s inability to make decisions well or when needed can be a critical weak link. Decisions that are based on data with low integrity, or whose implementations are not supported by resources, can be very costly. Decision processes that are very burdensome, time consuming, or difficult to arrange will be perceived as not adding value.

Proje ct De cisions

Decisions require sound processes to gather relevant information and to ensure the integrity of the related data. They need to integrate the wisdom of people who have broad relevant experience with those doing the detailed project work. Without efficient and value-adding process elements, the development project may fail to meet its objectives, the risks may increase, the customers may reject the solutions, and the investments may be squandered. Those are avoidable wastes and should not be tolerated. Decisions are made at different levels in the organization. Certain decisions are the responsibility of the corporate or business unit’s management. For example, these might apply to strategic objectives, portfolio changes, and funding levels. Other decisions are the responsibility of the local governance body for product development. In certain cases, there can be decisions to be made by customers, such as the accuracy of requirements or the completeness of solutions. It depends on the nature of the business. So who makes what decisions? a. Acting on the principle that decisions are to be delegated, they should be made at the lowest level in the organization where both the knowledge and the responsibility for implementation coincide. b. Acting on the principles of empowerment, decisions should be made by those who have the responsibility, authority, and accountability for their consequences. Many decisions are the responsibility of the development teams. The empowerment of development leaders, rooted in their charter approved in Phase 0, places important expectations on development teams, such as the following: • Act with responsibility, authority, and accountability for the development project: • Commit to the project’s success; don’t “pass the buck.” • Negotiate achievable requirements. • Select the best available approaches and solution concepts. • Integrate development activities into plans and capabilities. • Manage the actions to reduce risks. • Close the gaps between “actual” and “plan.” • Deliver superior value to customers: • Make decisions so the customers win. • Implement functional excellence.

909

910

A Proce ss for Product Development

• Practice the behaviors of high-performance teams. • Return superior value to the corporation: • Be efficient and lean in all practices. • Leverage proven capabilities. • Implement new capabilities where their benefits add highpriority value. • Manage deliberate actions to reduce risks. • Respect the uncertainty due to factors beyond control. The absence of clear responsibilities for actions and decisions can be a major impediment to the efficient management of a project. There are two easy tools that can be helpful. An RACI matrix shown in Figure 9 helps to clarify roles for activities, recognizing that teams do not work in the absence of external wisdom and oversight. For each type of activity, it guides the mapping of people who are to be responsible or accountable for the completion of the work, who provide consultation on critical inputs, or who must be kept informed of the results achieved. I’ve shown an example just for the purpose of illustrating the tool. Product Design and Project Execution

Allocation of Resources Ensure Excellence in Execution

Performance Appraisals of Core Team Members

Business Unit Manager

I

I

C

Technical and Business Functional Managers

I

C

R

Project Core Team Leader

A

A

R

Core Team Members

A

A

I

Development Team Leaders

R

R

C

Technical, Business and Process Specialists

C

C

C

(Example for Illustration)

R = Responsible to execute an activity A = Accountable for the completion

C = Consulted to provide critical inputs I = Informed of the output

Figure 9: A “RACI” Matrix Clarifies Responsibilities for a Range of Activities

The RAPID analysis shown in Figure 10 is a similar tool that applies to decisions. There are people who recommend decisions and those who

Proje ct De cisions

make decisions. Certain people provide necessary inputs. Some decisions need a variety of supporting agreements. What decisions need to be escalated to a high level of authority? The authorization for the release of a product to revenue-bearing customers, for example, may need to be made by a corporate officer. Decisions are without value unless people or teams have the responsibility and resources to implement them. The RAPID analysis identifies people in those various roles. As with the RACI example, I’ve shown an illustration of a gate decision for a particular organization model. The conclusions for your business may be very different.

Recommend

Project Core Team • Coordinates the work during the development phase • Develops information of high integrity relevant to the objectives of the phase • Derives conclusions from analyses • Makes sound recommendations

Agree

Perform

Functional Managers Project Core Team and • Ensure that functional Management plans are achievable • Responsible to • Ensure that functional implement the excellence is achieved decision and make • Enter gate review with it stick knowledge and a viewpoint of project’s recommendation • Can intervene to stop a decision

Input

Technical and Business Experts • Consulted by the Project Core Team to ensure data integrity and the soundness of the recommendation • Consulted for the decision, with no obligation by the decision maker to act on advice

Decide

Governance Body • Facilitate a high quality decision process • Ask open-ended probing questions • Set a climate for open discussion and evalution of data • Negotiate a decision whose consequences are acceptable

Figure 10: A “RAPID” Analysis Clarifies Who Gets to Decide, Who Has Input, and Who Gets it Done

Decision Processes Generally, decisions that add value and are timely are the result of a structured process. Here are some critical elements that you may find to be useful to improve your own decision processes: • The decision schedule should be critical to enabling subsequent work. • A decision must have alternatives, among which the best one is selected. • Acceptance criteria must be relevant to the objectives of the work being completed. • Information, data, analyses, and perspectives must be available with standardized contents and formats to facilitate their comparison to appropriate requirements. • The setting and atmosphere should be conducive to unbiased, honest evaluations.

911

A Proce ss for Product Development

• There must be open discussions among knowledgeable and responsible people, with ample opportunity for response to probing questions. • There must be a process for the implementation of the decisions, with responsibilities, resources, and follow-up methods. • Decisions must be communicated clearly to leaders and work groups, with appropriate rationale. These are thoughtful, cross-functional processes, not competitions among powerful voices. The consequences of bad decisions and late decisions can be severe for a development project. An easy tool to facilitate the choice of the best decision is derived from Stuart Pugh’s Concept Generation and Selection Process.5 Figure 11 illustrates the tool. For each decision criterion, alternatives are compared. The objective is for the chosen decision to be superior to the alternatives for all criteria. Alternative Decisions Criteria Relevant to Gate Decision

Ensure that the product is superior; customers win Ensure that the business plan is acceptable Ensure that business and technical risks are acceptable Enable the phases to market entry are more predictable Ensure alignment with product development portfolio Comply with funding budgets Conclusions

Keep Going Reference Concept for Comparisons

912

Go, w/ Redirect Corrective the Actions project

Stop the project

+

+

–

+

–

–

+

+

S

+

–

S

S

S

+

S

–

+

+ =3 – =0 –S = 2

+ =2 – =3 S =1

+ =2 – =2 S =2

Figure 11: The Best Decision Must Be Selected from Practical Alternatives

Project Monitoring, Enabling, and Control In a typical “phase/gate” process, the gate review is considered to be an important method by which management reviews progress and makes relevant decisions. For high-risk, critical projects, the reliance on gate 5. Pugh, Stuart, Total Design, Integrated Methods for Successful Product Engineering. New York: Addison-Wesley, 1991. ISBN 0-201-41639-5.

Proje ct De cisions

reviews may not be sufficient. Management may need to remain continually aware of progress and intervene when necessary to meet project milestones. For other projects, many formal oversight meetings may be excessive. In Phase 0, the project leadership and management agree on how the standardized process will be adapted to the project and how the project will be monitored, enabled, and controlled. Suppose that project risks are low, the process is repeatable, and the progress of the project team is fairly predictable. In that case, a reasonable strategy may be to define acceptable limits for temporary deviations from progress metrics. The project teams would be responsible to monitor their own progress. As long as their progress remains within the pre-agreed tolerances, their project status would be reported but without formal management review. However, if the progress of the project drifts beyond the acceptable limits, the leadership team will be obliged to call a meeting with management to consider alternative ways to get the project back on track. This same approach is valuable within long phases, between gate reviews. The example in Figure 12 may be thought provoking. It illustrates that not only can well-laid plans not achieve their intent, but that it may take a crisis to get management to act. It’d be far better for a project team to be able to make small corrections in progress along the timelines than to be left with major corrections when close to market entry. Almost never are those late corrections good experiences to be repeated. Market Entry Forecast Slip

Project Management Plan

Critical Path Activities

Current Plan Recovery Plan Enabled by Additional Resources

Conflict with another project jeopardizes market entry

Gate Reviews

Accelerated test plan recovered critical path

Rate of progress is too slow

Project Time Line

Figure 12: Example of Progress that Lags the “Critical Path” May Require Intervention by Management with Additional Resources or a New Plan

913

914

A Proce ss for Product Development

Gate Decision Reviews The construct of the gate review is intended to be a value-adding process in the context of the many things that can go wrong with a project. For many development projects, decisions may have to be based on incomplete work, uncertain information, and unnerving risk assessments. The teams may not have all the resources or knowledge they need. There may be conflicts among projects in a business unit, or among business units in a corporation. So, people involved in reviews need to develop capabilities to make important and timely decisions when faced with partial information and with uncertainties about the future. A gate review is a phase-based decision process. It enables the company to invest in new products progressively, in the context of long-range funding plans. By management being routinely involved with teams, expectations can be clarified and a sense of urgency can be reinforced. It establishes an agreement between the recommendations of the project’s leadership and the wisdom of the managers on the governance body. The process of gate reviews is an important way for management to share responsibilities for the success of the development project. It balances project-specific concerns with those of the overarching business. Through this process, senior management leads product development, implements the strategy of the new product portfolio, empowers the project leadership team to manage the investment, and takes actions to improve the probability of success for the project. It also is a process by which a project can be redirected or cancelled. Gate reviews are positioned in the development process at natural points of convergence of critical information and the needs for projectlevel decisions. Their schedule is not based on the passage of time. An exception would be when a phase is very long, for which interim project reviews might be warranted. However, unless there’s a decision to be made, that objective would relate more to risk management. The concept of a “gate” implies that there’s a respected barrier to be opened, rather than a “speed bump” to be treated casually. However, this does not mean that gate reviews can’t be combined or treated as being off the critical path. In fact, principles of lean product development encourage that. It becomes a matter of the objectives for each gate review and how the consequences of the decisions are reflected in the work of the team. The decisions at gate reviews have alternatives: • Continue the development project with the work of the next phase. • Go forward, but with specific conditions attached and follow-up check points. • Remain in the current phase to resolve specific risks and revisit the gate at a planned later date.

Proje ct De cisions

• Redirect the project: • Accelerate • Re-target • Re-configure • Delay • Place on hold • Stop the project and redeploy the resources to higher-valued projects. The potential consequences for delays in market entry must be judged against the delays in revenue and the extended costs. Decisions are agreements, with rationale that are acceptable so that all involved organizations can live with the consequences and support the implementation. They are not consensus decisions—for example, for which everyone has to approve—based on their own functional interest. Some organizations may have to compromise their own interests for the benefit of the enterprise and its customers. Particularly if submitted early, a recommendation to stop an ill-fated project should be rewarded. People at all levels should be motivated to avoid wasting resources and time on projects that have a low probability of technical or business success. The agendas have higher value when they focus more on improving value and reducing risks than on the inspection of deliverables or on compliance with work standards. With a focus on business concerns, the agenda for a gate review needs to be concise. The discussions are better aimed at improving the quality of future work, than just on evaluating the work completed. The agenda for a 1–2 hour meeting should, for example: • Clarify the objectives for the current phase and its acceptance criteria. • Compare the achievements and forecasts against the expectations for the phase, ensuring that they are supported by clearly interpreted, high-integrity data. • Evaluate the business consequences of technical data, organizational capabilities, and competitive market forces. • Evaluate business and technical risks, such as the consequences of unfinished work from the previous phase, or risks from work of the next phase started too early. • Ensure that resources and partner organizations are committed and aligned to the objectives and timelines of the project.

915

916

A Proce ss for Product Development

• Ensure that the program is prepared to achieve the work of the next phase with fast, deliberate implementation and with a high degree of predictability. • Approve strategies and action plans for the next phase. • Identify specific actions necessary to make future work more predictable and of higher value. • Clarify the expectations for future work, including near-term corrective actions. • Ensure alignment with the business portfolio and its improvement initiatives. • Evaluate the recommendations of the project’s leadership team. Some companies focus the review agenda just on answers to questions that are derived from the review of advanced documentation. There are no presentations. Other companies depend entirely on reacting to evidence presented in the meeting. There’s no best way. Local management style probably rules, as long as the end results are value-adding and efficient. One way to focus on those topics that represent risks is to use a checklist of deliverables with assessments against their acceptance criteria. The “traffic light” form, shown in Figure 13, is developed by the project leadership team to identify weaknesses for attention. The criteria are reasonably straightforward and familiar: • Green (G) = Acceptance criteria are satisfied; the work of the next phase can proceed without risk. • Yellow (Y) = Acceptance criteria are not completely satisfied; the work of the next phase is handicapped, but the risks are acceptable; corrective actions will be completed with near-term target dates and a high level of predictability. • Red (R) = Acceptance criteria are not satisfied; the work of the next phase is handicapped to the point that the risks for proceeding are not acceptable; additional work with uncertain timelines is required to reduce the risks to an acceptable level. A gate review is also an important opportunity for management to enable the project leadership team. So, it is very appropriate for management to accept action items to improve the probability of success, such as • Provide specific resources or additional funding to recover the critical path or to reduce project risks.

Proje ct De cisions

Deliverables

Owner

Achievement versus Criteria

Program Recommendation

G

G

Y

R

Y

Remarks re. Risks

R

Complete, approve, and freeze program requirements

Risk will be reduced by Customer Acceptance Testing in next phase

Validation of superiority of value proposition

Value Proposition has been validated by focus group

Program Requirements Document

Requirements have not yet been frozen by Marketing

Validation of Program Requirements

Funds are not budgeted for customers to validate requirements

System Requirements Document

Requirements deploy the assumed value proposition

Subsystem Requirements Documents

Requirements deploy the assumed value proposition

Figure 13: Traffic Light Form Uses Familiar Symbols to Note Those Project Rsks that Are Important to the Gate Decision

• Maintain sufficient capacity and clear priorities at sources of queues, such as centralized testing services. • Resolve conflicts over resource allocations or project priorities. • Approve major purchase orders when needed, such as for long lead tooling or capital projects. • If necessary, participate in the redesign of the project. • Initiate the redesign of processes to increase efficiency or to reduce waste. • Provide guidance from lessons learned by previous development projects. • Delegate decisions to the appropriate level of knowledge and responsibility. • Make expectations specific, clear, and prior to the work. Contrary to typical management incentives, larger financial gains can be achieved by getting products to market on time with higher quality than by complying with development funding budgets. Unfortunately, individual performance expectations for managers create incentives for the opposite behavior.

917

918

A Proce ss for Product Development

Gate Keepers Development projects within a business should have a governance body comprised of specific managers. Who are these managers? They are senior managers who act as a small decision-making body for all gate reviews, and for all projects in the business unit. With them being responsible for the portfolio of development projects, you can think of them as investment bankers, worried more about how well the business elements of development are progressing than about the technical details. Gate keepers can be characterized as being: • Knowledgeable of the business and its technologies. • External to the project, but internal to the business. • Able to ask better questions and know good answers. • Familiar with best practices and metrics, as well as their implications. • Owners of resources whose allocation can be changed as needed. • Possessors of position authority to remove organizational or bureaucratic barriers and resolve cross-project conflicts. • Stakeholders in the success of the portfolio of development projects. • Links to higher-level governance bodies within the corporation. A typical group of gate keepers may include the leaders of design engineering, manufacturing, marketing, service, and finance. The business unit’s leader may be the gate keeper who leads the process and has ultimate decision responsibility. The leader of the project’s core team would coordinate the presentation of the project’s recommendations and supporting arguments. Some companies include expertise external to the business to act as a corporate conscience or “devil’s advocate.” To emphasize their role, some companies call these “gate keepers” or “key enablers.” Similar to the commitment of core team members, members of this committee are stable for the business and participate with commitment to the business. As with the core team, there are expectations that the product development process places on these managers: • Ensure the success of the business: • Promote the understanding of the business of the corporation and that of its customers.

Summar y

• Ensure alignment with business and technical strategies and portfolios. • Set high and achievable expectations for excellence in execution. • Enable the project’s progress: • Provide the skills and funding required, when required. • Provide guidance from lessons learned by other projects. • Eliminate excessive constraints on the project. • Remove unnecessary bureaucratic barriers. • Maintain focus and a constancy of purpose. • Be decision makers: • Exercise independence in the judgment of recommendations. • Ensure the integrity of the data that influence decisions. • Work for the success of the whole business. • Manage the follow-up implementation of decisions. • Act with responsibility for the cross-functional product development process: • Reinforce the discipline of the process. • Encourage flexible adaptations to improve effectiveness and efficiency. • Coach teams in the expectations of the process. • Institutionalize corporate lessons learned.

Summary Standardized processes for product development and its decisions must have a common understanding across an organization’s teams and management. Its expectations must integrate technical development with business development. It must facilitate the management of cross-functional activities and dependencies. The structure of the process serves as an important framework for project management plans. Excellence in the execution of the process reduces the project risks, and improves the integrity of plans and commitments. Tasks are expected to employ the best tools and methods to accomplish the right work at the right time

919

920

A Proce ss for Product Development

with efficient use of resources. These processes set high expectations for the leadership team and work groups within the project, as well as for the management team that acts as the governance body for all projects.

About the Author Bill Jewett is a consultant to businesses engaged in the development of new technologies and multi-disciplined products. With insights into important paradigms and advancements in practices, he assists improvement teams in upgrading their engineering and management processes, project management practices, cross-functional teamwork, and the governance of development programs. For many years, Bill worked for Eastman Kodak Company and Heidelberg Druckmaschinen, with a focus on the development of high-volume electrophotographic copiers and printers. Among his division-level responsibilities were the management of product development programs and of competency centers for mechanical and systems engineering. At the corporate level, he was one of the authors of the processes for advanced product planning, technology development, product commercialization, and their related governance. For over a decade, he taught the processes and coached teams in their adaptation and implementation. As the process steward, he evolved the models to incorporate lessons learned from internal practice and external benchmarking. Currently, Bill and an associate are writing a book to integrate strategies and methods for the development of increased robustness and higher reliability in products. They expect their book to be available in the last half of 2007. Bill can be reached by telephone at 585-705-3100 or by email at [email protected].

Selecting Project Portfolios Using Monte Carlo Simulation and Optimization By Lawrence Goldman and Karl Luce As a Six Sigma implementation matures, Champions are faced with the increasingly challenging task of selecting a portfolio of high-quality projects with a strong financial return. Selecting Six Sigma projects is a one-time decision, and comparing and selecting Six Sigma projects is a difficult task when large numbers of projects are dissimilar, complex, and potentially risky. Each project will be either financially successful or not, and the Six Sigma focus on savings and improved income remains a key consideration for every Champion and Six Sigma executive. If a project is not successful, the company runs the risk of incurring the loss of the initial investment and maybe even its faith in the benefits of Six Sigma. Existing project selection techniques can help align projects to company strategies, shareholder value, and management metrics, but these approaches generally lack an assessment of the monetary risks related to individual projects. The goal of this article is to describe Monte Carlo simulation and stochastic optimization, an increasingly popular alternative technique for the financial justification of projects. The combination of simulation and optimization is a unique approach that provides Champions with a more granular, finance-based solution to portfolio analysis. Managers can enhance project portfolios to include the effects of uncertain costs, revenues, and resource usage, which most acknowledge but lack the ability to calculate. This article uses Crystal Ball® software, published by Decisioneering®, Inc., to simulate the uncertainties in a project portfolio spreadsheet and to optimize the portfolio so as to increase value and decrease risk. The paper includes an aerospace success story and an extended case study example, which demonstrates in more detail the application of these techniques.

Portfolio Selection Methods Among the companies, organizations, and institutions that implement Six Sigma, there is no established standard when it comes to a project selection method. Practitioners are nowhere close to a consensus on which specific approach is best, nor does there appear to be a standard in the making. The goal of this article is to describe an increasingly popular alternative technique for the financial justification of projects: the combination of Monte Carlo simulation and stochastic optimization. 921

922

Sele cting Proje ct Portfolios Using Monte Carlo Simulation and Optimization

Most Six Sigma projects are selected because they have a positive impact on the customer.1 In companies just beginning their Six Sigma journey, many projects are selected because they are the proverbial “lowhanging fruit:” quick-win opportunities with a strong financial return. Once a company has practiced Six Sigma for several years and exhausted its supply of obvious and lucrative projects, the hard work of project selection begins. At this stage, with the quick-win projects resolved, Champions find themselves faced with a larger pool of mixed and potentially marginal projects. Many of these projects, while highly desirable in terms of impact on the customer and bottom line, have been delayed due to questions about their feasibility or financial risk. Champions must contemplate a new slate of issues: What selection methods can help you to cope with a large portfolio of recommended projects? How do you compare dissimilar projects? Is it possible to forecast which projects will provide the best return given uncertainties around scheduling, investment, and success? How you decide to prioritize and select projects is quite often dependent upon the practices and preferences of your organization. Listed next, in no particular order, are just a few of the project selection techniques cited in the literature. Each approach has been employed with a varying degree of success, and all have their strengths and weaknesses. • Pareto Analysis2—A strong benefit of Pareto analysis is that it is

usually based upon data (warranty costs, rework costs, production costs, defect measurements) that have been tracked (historical data) and accepted by the management community. Additionally, a Pareto chart can be compiled by an analyst or Master Black Belt (MBB) and so does not tie up substantial management time. The downside of this approach is that the problems (and projects to solve those problems) are based only upon one or a few measured criteria that may not incorporate the bigger picture of customer satisfaction. Also, the Pareto Priority Index (PPI), when calculated, may not provide clear direction as to which projects are truly the best ones to pursue. • Cost of Poor Quality3—Like Pareto analysis, Cost of Poor Quality

(COPQ) uses data that indicates where costs are incurred due to defective product in prevention controls, testing controls, or warranty/scrap/rework activities. However, it focuses solely on the 1

Pyzdek, T. The Six Sigma Handbook: Revised and Expanded. New York: McGraw Hill, Inc., 2003. p. 188.

2

Ibid, p. 198.

3

Ibid, p. 219.

Portfolio Sele ction Methods

cost side of the product, which may or may not reflect customer satisfaction levels. A primary disadvantage of COPQ is that the majority of these costs are hidden costs not considered with traditional accounting systems. • Project Prioritization Matrix4—This is a top-down approach that

defines the candidate projects such that they line up directly with company-wide strategic initiatives. This methodology, based on the Cause-and-Effect (C and E) matrix, offers ease-of-use and the ability of management to understand how to use it. On the downside, this approach requires upper-level management and Six Sigma leadership (Champions and/or MBBs) to devote a block of time together, and the inability to schedule this team activity may become an obstacle to success. • Project Clustering5—In this top-down technique, projects are “clus-

tered” around company-wide strategic initiatives. Like the Project Prioritization Matrix, this approach should be performed with management and Six Sigma leadership to ensure the top-down nature of the decision making, which means a concurrent block of time must be devoted for the team activity. • Value-based Management6—This technique, which relies on an

understanding of where value lies within an organization, prioritizes based on a flow-down from the top-level strategy and shareholder needs through to business units and value streams to individual projects. The advantages of this approach include that it requires a thorough assessment of the competition and that the company never loses sight of where economic value resides. The disadvantage is that such a value-based process does not account for projects that cannot easily be applied to bottom-line shareholder value. • Quality7,8 Function Deployment (QFD) and Strategy Deployment

Matrix —This method creates a matrix around customer needs that helps to quantify and analyze the relationships between the corporate goals and the individual projects. Like the Project Prioritization (C and E) Matrix, this methodology is a top-down approach that should involve upper-level management and Six Sigma leadership. More advanced Six Sigma organizations use QFD for product development (DFSS) and readily accept this technique. One

4

Zinkgraf, S.A. Six Sigma: The First 90 Days. New Jersey: Prentice Hall, 2006. p. 157.

5

Ibid, p. 152.

6

George, M.L. Lean Six Sigma for Service. New York: McGraw-Hill, Inc., 2003. p.106.

7

Pyzdek, T. The Six Sigma Handbook: Revised and Expanded. New York: McGraw Hill, Inc., 2003. p. 188.

8

Muir, A. Lean Six Sigma Statistics. New York: McGraw-Hill Professional, 2006. p. 30.

923

924

Sele cting Proje ct Portfolios Using Monte Carlo Simulation and Optimization

disadvantage of this approach is that it requires experienced facilitators and more training than for C&E approaches. It most likely will require multiple sessions and substantially more time for participants. • Theory of Constraints9—The Theory of Constraints (TOC) approach

uses the concept of resource constraints to help prioritize where to focus improvement projects. It is superior to traditional Total Quality Management (TQM) approaches in that TOC can quickly identify opportunities at a below-the-macro-operations level. The primary disadvantages of this method are that the analyst identifying projects must be trained and practiced in the technique and that the time required to perform the analysis may be longer than management desires. It is not widely used in Six Sigma programs. • Political Choices—In this pernicious “method,” the structure of a

company and Six Sigma initiative is such that certain individuals or groups of individuals have the greatest sway over which projects are selected and which are rejected. Sometimes, these individuals are simply the ones who offer the loudest or glitziest presentations. The lack of an objective, consistent selection system has mostly negative consequences for morale and a team-based culture. In general, the preceding methods are popular because they align Six Sigma projects with an organization’s strategic and customer concerns. However, none of these approaches directly addresses the kinds of metrics, such as probability of project completion or potential financial return, which are deemed critical to the long-term success of the Six Sigma program. Although these metrics are more difficult to forecast, decision makers can use one relatively new selection technique—Monte Carlo simulation and optimization—to help define a project portfolio that accounts for project cost and success. The following section discusses the details of this methodology and where it can be applied.

Monte Carlo Simulation in Six Sigma The term “Monte Carlo simulation” is used to describe a group of related probabilistic techniques that help answer the question of “what if…?” Simply put, Monte Carlo simulation is a sampling experiment where the inputs are defined as probability distributions rather than estimated or average values. Monte Carlo simulation randomly samples from these distributions to generate multiple input values and calculate how a “system” responds to these random inputs. A system is defined as a c ombination of inputs (X’s) and outputs (Y’s) that represents a realistic mathematical model such as an engineering design, a market demand forecast, a transactional process, or a portfolio of products. 9

Pyzdek, T. The Six Sigma Handbook: Revised and Expanded. New York: McGraw Hill, Inc., 2003. p. 201.

Monte Carlo Simulation in Six Sigma

While its roots can be traced back as far as the 17th century, Monte Carlo analysis was most prominently used in the Manhattan Project, where scientists at the Los Alamos National Laboratory used it to predict the possible effects of nuclear explosions. Today, Monte Carlo simulation is a proven, efficient technique requiring only a computerized random number generator.10 A Monte Carlo simulation can measure thousands or millions of what-if scenarios in a very short time, and most anyone with a reasonably fast PC can run a Monte Carlo analysis. In Six Sigma or DFSS implementations, this form of simulation is commonly used to test process cycle time or parts tolerance, where the probability distributions represent the variation in the input factors. Because these problems are mathematical in nature, they can be described in the format of Y = f(x) and modeled within a spreadsheet or other analytical software tool. Simulation is especially useful in non-intuitive situations where equations are non-linear, there are large numbers of inputs, and the data is incomplete or impractical to collect. In contrast, business problems like project selection do not initially seem to be good candidates for simulation. How can you model a strategic business decision? Where is the variability? How do you include subjective or less-tangible information? While most organizations employ non-probabilistic forecasting models like discounted cash flow, capital budgeting, and process flow-charting, fewer actually use these models to quantify the risks in their strategies and investments. One solution is to recognize that probability distributions can also describe unknown inputs, such as expected revenues or resource usage, as shown in Figure 1. These inputs, whether estimated or based on historical data, represent an uncertain future rather than known variation around a target value. The simulation outputs are business metrics and optimized portfolios. When would Monte Carlo simulation be an ideal tool for project portfolio selection? Under these conditions: • Project has financial uncertainty—When the financial returns and

risks of a product or process are uncertain, and the financial estimations or forecasts can be described in a spreadsheet model. • Project has schedule uncertainty—When a project schedule

includes a critical path with tasks that have uncertain durations or efforts. • Project has cost controls—When it is prohibitively expensive to

acquire data for an input or output of a process or product, and simulation can create realistic virtual data. 10

Kelton, W. D., and A. Law. Simulation Modeling & Analysis. New York: McGraw Hill, Inc., 1991.

925

926

Sele cting Proje ct Portfolios Using Monte Carlo Simulation and Optimization Outputs EVA Forecast

Inputs Candidate Project List Expected Revenues Optimal Project Portfolio

MCS Resource Usage

2

X3

1 Initial Investment

4

X5

6

X X 7

8

Figure 1: Monte Carlo Simulation (MCS) Uses Probability Inputs to Simulate Models and Determine the Inputs’ Effects on Strategic Outputs

• Project is high-risk—When making decisions about a high-risk proj-

ect—for example, one with a potentially huge negative revenue impact—simulation helps to estimate the certainty of success, magnitude of risk, and key drivers influencing success. Portfolio simulation and optimization are popular tools outside of Six Sigma.11,12,13,14,15 These tools have yet to earn the same level of recognition and respect within the Six Sigma field of study. But the same techniques 11

Faulder, D.D., and F.L. Moseley. 2004. “A ‘Top Down’ Approach for Modern Portfolio Theory to Oil and Gas Property Investment.” Proceedings of the 2004 Crystal Ball User Conference [online]. Available online at http://www.crystalball.com/cbuc/2004/papers/ CBUC04-Moseley.pdf.

12

Rodriguez, J., and K. Pádua. 2005. “An Application of Portfolio Optimization with Risk Assessment to E&P Projects.” Proceedings of the 2005 Crystal Ball User Conference [online]. Available online at http://www.crystalball.com/cbuc/2005/papers/cbuc05-rodriguez.pdf.

13

de Lange, T. 2005. “Optimizing the Growth Portfolio of a Diversified Mining Company.” Proceedings of the 2005 Crystal Ball User Conference [online]. Available online at http://www.crystalball.com/cbuc/2005/papers/cbuc05-delange.pdf.

14

Letzelter, J.C. 2005. “Finding the Efficient Frontier: Power Plant Portfolio Assessment.” Proceedings of the 2005 Crystal Ball User Conference [online]. Available online at http://www.crystalball.com/cbuc/2005/papers/cbuc05-letzelter.pdf.

15

Hill, C. 2006. “Portfolio Optimization Applied to Acquisition Evaluation.” Proceedings of the 2006 Crystal Ball User Conference [online]. Available online at http://www.crystalball.com/ cbuc/2006/papers/cbuc06-hill.pdf.

Succe ss Stor y: Aerospace R&D Proje ct Portfolio

used to optimize a portfolio of financial assets can be applied to optimizing a portfolio of potential DMAIC, DFSS, or Lean Six Sigma projects. In practice, these techniques can provide valuable insights that lead to more informed strategic decisions and more successful project portfolios. As with any tool, portfolio simulation and optimization should be used with a proper amount of caution and skepticism. The accuracy of a simulation cannot be better than the accuracy of its model in representing the real system. All simulation and optimization models should be tested for accuracy and applicability, and Champions and managers should be able to defend their reasoning for applying specific probability distributions within a model. Inexperienced modelers and those with particularly narrow agendas can easily bias models and misrepresent and misuse simulation results. Yet, when performed with skill, these modeling techniques prove to be powerfully effective decision-making tools. In the next two sections of this article, a success story from the aerospace industry and a more generalized Six Sigma example demonstrate the effectiveness of simulation and optimization to project portfolio selection.

Success Story: Aerospace R&D Project Portfolio The following story was relayed to us by the business development division of a leader in defense and aerospace. Each year, this business division must decide which projects to fund from a possible portfolio of over 300 new marketing or R&D projects. The goal is to select the optimal portfolio of projects to fund such that (1) returns are maximized, (2) budget and resource constraints respected, (3) risk is minimized, and (4) growth rate remains at or above market norms. Prior to implementing the new project selection process a little over two years ago, decisions on which projects to fund were often based on prior sales for the division requesting project funding. The project’s opportunity, its associated risk, and correlation to the success of other projects were not an integrated part of the decision-making process. The division also found it difficult to collect and use metrics on past performance in order to improve future performance. The division implemented a new process that included a Web-based application to capture project information, calculate its return, and pass the data to Microsoft® Excel®. Crystal Ball® software was then used to analyze the various projects and determine which projects met the required returns, complied with budget constraints, and produced the necessary growth. The simulation and optimization model also accounted for positive and negative correlation between the projects.

927

928

Sele cting Proje ct Portfolios Using Monte Carlo Simulation and Optimization

Although the business was already growing at a rate that exceeded the market norm, with the new project selection process, this division achieved a multi-million dollar positive shift in performance to budget in the first year, roughly equivalent to 10% of the overall budget. This money is now used to fund additional projects, further increasing the growth rate. Another important, yet less quantifiable, benefit was a substantial reduction in management resources needed to make final project selection decisions, leading to overall management support for the new project selection process.

Project Selection Example Now let’s see how this type of analytical selection technique would work in practice. As a Six Sigma Champion, you have been presented with eight possible projects for the upcoming year. For each project, your Six Sigma experts have computed: (1) the expected change in revenue for each project, (2) the expected cost savings, or change in expenses, and (3) the initial investment required for each project. Using these figures, the finance manager has created a spreadsheet model to compute the gross profit and the Economic Value Added (EVA), or economic profit, for each project, as illustrated in Figure 2.

Figure 2: Spreadsheet Model Comparing Eight Six Sigma Projects

In the best of all worlds, you would run all eight projects; in reality, you have budget and labor limitations. If you select all eight projects, then you are $2,400,000 over budget and need 14 more team members on the job than are available. Additionally, many of the variables, including the Project Revenues, Cost Savings, Investments, and Staff Requirements, are highly uncertain. Thus, the problem is to determine, based on financial considerations, which Six Sigma projects to select to maximize the Total EVA while staying within the budget and labor limitations.

Proje ct Sele ction Example

Complicating your job is the fact that while several of the eight projects appear to offer similar EVAs, each offers a different financial story. For some projects, the Expected Revenue is low or nonexistent, while the Cost Savings are high. These could be Six Sigma process improvement projects. Other projects show the reverse, with higher Expected Revenue and no Cost Savings. These could be Design for Six Sigma (DFSS) projects. And several of the projects fall between these extremes, showing both Expected Revenue and Cost Savings. How can you compare such dissimilar projects? One method is to compare each project’s EVA and to select the subset of projects that leads to the highest, or optimal, EVA. This model was constructed such that, by entering a 1 or 0 for the Investment Decision column on the right, you can turn a project “on” or “off.” If you enter a 0 for Projects 3 and 8, then Excel recalculates, and you end up with a breakeven budget at an appropriate staffing level. You also see a relatively small drop in Total EVA. In a world without uncertainty, this would be an acceptable solution, but you know that many of the estimates in this model are uncertain, including the Project Revenues, Cost Savings, Investments, and Staff Requirements. Thus, your goal is to incorporate risk analysis within the context of the project selection, and to select the best projects that satisfy your constraints while optimizing your EVA. Next, you will begin to use Crystal Ball software, Decisioneering®’s spreadsheet-based suite of tools that includes a simulation program and a stochastic optimizer; together, these software tools provide the power to optimize a project portfolio while accounting for uncertainty in the inputs.

Defining Simulation Inputs As stated earlier, a Monte Carlo simulation requires that each uncertain variable in a model is represented by a fully specified distribution (that is, Normal, Lognormal, Poisson, Binomial). While it may initially seem to be a challenge to determine where to start, you can help yourself by asking: “Which of these inputs is just an estimate?” In the completed spreadsheet model in Figure 2, each value for Project Revenue, Cost Savings, Project Investment, and Staff Requirement is a single-point estimate or possibly an average value. These inputs are your uncertain factors. The next logical question is: “Which distribution do I use and how do I determine the parameters?” The answer depends on how carefully each of the project sponsors did their homework. How did they determine the value for their inputs? Did they select the center of a range of values? Did they use historic data from similar projects? You must understand these details to help guide your selection of the proper probability distributions.

929

930

Sele cting Proje ct Portfolios Using Monte Carlo Simulation and Optimization

For example, the given Investment for Project 1 is shown as $1,250,000. A discussion with the project sponsor reveals that this value is indeed the likeliest one, but that investment may be as low as $1,125,000 or as high as $1,562,500. You can use these three parameters to define a triangular distribution, as seen in Figure 3. You also inquire about the revenues for Project 1. The owner of this process held a strong opinion on mean for the revenues, $2,000,000, but when asked about uncertainty around the mean, she was less forthcoming. After discussion, you both agreed to implement the variable as a normal distribution where the range was defined using the 5th and 95th percentile around the mean, illustrated in Figure 4.

Figure 3: Triangular Distribution Representing the Investment for Project 1

Figure 4: Normal Distribution Representing the Expected Revenue of Project 1

You repeat this process until you have defined all of the uncertain inputs in your model. The input distribution types and parameters are not hard coded; you can always adjust them as you simulate and analyze the model.

Defining Outputs and Running a Simulation After specifying distribution models for all of the uncertain inputs in the model, your next step is to define one or more output variables. These outputs, or forecasts, are the responses or effects of the system

Proje ct Sele ction Example

represented by the Y in Y = f(x). The most obvious outputs in this model are the amount over budget, the number of staff over available labor, and the Total EVA of the portfolio. Running a simulation is a simple task compared to creating the probabilistic model. In most simulation programs, you select a number of trials and then click on a run button. For each simulation trial, the software randomly selects a value from a defined distribution and enters that value into the spreadsheet, which then recalculates the affected formulas. The software then saves the forecast values from each simulation trial, and, when done with the simulation, calculates descriptive statistics for the entire group of trials. In Crystal Ball®, the forecast values are compiled on graphs and tables that help in the interpretation of the simulation results.

Analyzing a Simulation You may first want to test run a single potential project portfolio. As shown earlier, you can deselect projects 3 and 8 to forecast an attractive Total EVA of $11,025,000 that is within budget and within labor constraints. A 1000-trial simulation of this portfolio, however, reveals a less rosy picture: the mean Total EVA would be just less than $6,500,000, with a 1% certainty of exceeding the original estimate of $11,025,000, as shown in Figure 5. In the same project portfolio scenario, you are 52% certain to need more than the available staffing and 87% certain to be over budget.

Figure 5: Simulated Forecast for a Portfolio of Projects 1, 2, 4, 5, 6, and 7

With a forecast range of over $15,000,000, this simulated portfolio has become an unattractive, high-risk alternative. To determine what is driving the variation, you can use tools such as sensitivity analysis.

931

932

Sele cting Proje ct Portfolios Using Monte Carlo Simulation and Optimization

Crystal Ball® software calculates sensitivity by computing rank correlation coefficients between every assumption and every forecast, and these normalized coefficients provide a meaningful measure of the degree to which assumptions and forecasts change together. This critical tool helps modelers to quickly and easily judge the influence each distribution input has on a particular output. Sensitivity analysis ranks the assumptions according to their importance to each forecast cell. Similar to a Pareto chart, the sensitivity chart, Figure 6: Sensitivity Chart for the Total EVA illustrated in Figure 6, displays these rankings and indi- Forecast cates which of the critical inputs in the analysis cause the predominance of variation in the output of interest. For the Total EVA forecast, the Cost Savings and Investment for Project 2 and the Expected Revenue and Cost Savings for Project 5 contributed most to the variation in the forecast. This insight can help to improve the quality and accuracy of the model, help you understand what’s driving cost and staff overruns, and perhaps guide you to an early elimination of certain projects from the overall portfolio.

Stochastic Optimization: Discovering the Best Portfolio with the Least Risk The next obvious step would be to run repeated simulations with different portfolios. But imagine how many consecutive simulations of project subsets you would have to run to eventually find the best portfolio, say the one with minimal risk (lowest standard deviation) and the highest mean value. At most, you could have 2n project portfolios, where n = 8 (the number of projects). Constraint equations would reduce this number of portfolios but still leave you with a daunting number of potential portfolios. With a manual approach, you would probably never find the best of all portfolios.

Proje ct Sele ction Example

Because simulation alone is insufficient to identify the best solution, your project portfolio will require a combination of simulation and optimization, referred to as stochastic optimization. Stochastic optimization helps modelers find the controllable variable settings that result in the best statistical parameters of forecast variables. For example, stochastic optimization can find ways to optimize an inventory system to minimize costs while ensuring enough inventory to meet uncertain future product demand. Other optimization examples include production scheduling, project/strategy selection and prioritization, and workforce planning. Using an optimizer in conjunction with a simulation tool, you can run consecutive simulations very quickly while tasking the optimizer to return only the best portfolios (as described previously). OptQuest® global optimization software is an add-in created specifically for Crystal Ball software. The optimizer is built with multiple search methods and intelligence features that allow it to train on good and bad results and then to converge more quickly on best results. The numerical methods used in OptQuest algorithms are collectively known as “metaheuristic optimization.”16 In a Crystal Ball model, the controllable inputs are referred to as decision variables. Each decision variable is defined by upper and lower bounds and value type, which is either discrete or continuous. A single OptQuest run comprises many Crystal Ball simulations, with perhaps 1000 trials per simulation, at different settings of the decision variables. The optimizer runs sequences of Monte Carlo simulations to find the right combination of decision variables for the best possible results. In the portfolio model, the decisions variables are the Investment Decision for each project. If the optimizer selects a “0,” then the project is excluded from the portfolio. A “1” means that the project is included in the portfolio. Combinations of 1’s and 0’s define each portfolio that the optimizer simulates. The objective of the optimization is to maximize the mean value of the Total EVA forecast. Still, the highest Total EVA may not be associated with a satisfactory portfolio given your budget and staffing restrictions. To meet these restrictions, you need to set requirements after each optimization to sort out which solutions are feasible or infeasible. For a more reasonable result, you decide to include two requirements that describe an acceptable management: (1) that the Over/Under Budget forecast is that 95% or more of the simulation trials are below $0 M (the budget constraint) and (2) that the Over/Under Staffing forecast is that 95% or more of the simulation trials are below 0 personnel (the labor constraint). With the objective and two requirements, the optimizer will search for solutions where it will select the best subset of projects that keep very close to the budget and labor constraints but yield the highest possible Total EVA. 16

Laguna, M. 1997. “Optimization of Complex Systems with OptQuest.” Available online at http://www.crystalball.com/optquest/complexsystems.html.

933

934

Sele cting Proje ct Portfolios Using Monte Carlo Simulation and Optimization

After analyzing 256 simulations (the equivalent of all possible 256 potential portfolios), the optimizer has converged on the best solution (in Figure 7), which is a portfolio that includes Projects 2, 4, and 7. The mean value for Total EVA is now over $9.1 million, with a standard deviation of less than $2 million (in Figure 8). In this portfolio, you are 98% certain to be under budget, and you will always use less staff than is available.

Figure 7: Performance Graph of Portfolio Optimization

One issue you may want to consider is that the mean forecast for staffing is 37 Figure 8: Forecast Chart for Optimized Portfolio people who are not assigned to a project. This would be a waste of resources, so you could perform a second optimization with an additional staffing requirement where the 50th percentile of the staffing forecast would be at (18) staff. You can also use the Solution Analysis tool to review details of other robust project portfolios (Figure 9), in case politics or logistics require an alternative solution. Projects 2 and 4 were always included in the most profitable portfolios, but Projects 7, 6, 1, and 8 all resulted in EVAs greater than $8,250,000. At this point, your analytical work is done. What remains is the more sensitive task of reporting your financially based conclusions to the other implementation leaders. For any such report to succeed, it must include a description of the model assumptions and optimization parameters. While doing so may invite challenges to the validity of the inputs, this level of transparency can expose the economic risks of certain projects and can help build stronger support for the final selected portfolio. The ability to propose alternative portfolios with high returns (Figure 9) also lends flexibility to the management decision.

Conclusions

Figure 9: Solution Analysis Tool Compares Alternative Robust Portfolio Solutions

Conclusions The health of a Six Sigma implementation is highly vulnerable to factors such as changes in leadership, the quality of Belt training, the support of the company culture, and the success of the projects a company chooses to address. Although many project selection methods help align projects to company strategies, shareholder value, and management metrics, these approaches generally lack an assessment of the monetary risks related to individual projects. Monte Carlo simulation and stochastic optimization allow Champions a more granular and finance-based approach to portfolio analysis. With the continuing development of faster personal computers and sophisticated software products, Monte Carlo simulation is fast becoming a staple in the desktop analytic toolkit. When applied correctly, Monte Carlo simulation and stochastic optimization provide valuable insights not available through non-finance–based methods and can result in higher quality and more successful portfolios of Six Sigma projects.

Crystal Ball Software Crystal Ball, a Microsoft Excel-based suite of software tools, provides Monte Carlo simulation, optimization, and forecasting techniques that can help you predict capability, pinpoint critical-to-quality factors, and explore design alternatives. The Crystal Ball web site (http://www.crystalball.com) offers Six Sigma papers, example models (including the Project Selection–EVA example described previously), and recorded web seminars.

935

936

Sele cting Proje ct Portfolios Using Monte Carlo Simulation and Optimization

You can also download free trial versions of Crystal Ball and software tutorials. For specific inquiries, contact the authors at [email protected]. “Decisioneering” and “Crystal Ball” are registered trademarks of Decisioneering, Inc. All Rights Reserved. Microsoft and Excel are registered trademarks of Microsoft Corporation in the U.S. and other countries. OptQuest is a registered trademark of OptTek Systems, Inc.

About the Authors Lawrence Goldman is the Director of Six Sigma Marketing at Decisioneering in Denver, CO. He joined the company in June 1997 as a program manager with the product development group and soon moved into the training and marketing groups. Lawrence has published several papers on the topic of simulation and optimization and taught courses and seminars at academic institutions and corporations in the U.S., Canada, and the U.K. His current focus is developing educational and training materials and seminars for Crystal Ball for the Six Sigma market. He completed his Six Sigma Black Belt certification in August 2006. Prior to working for Decisioneering, Lawrence was a trainer and technical support assistant for a company specializing in mining and engineering database software. He received an M.S. in Geology from the University of Cincinnati in 1993 and a B.A. from Cornell University in 1988. Karl Luce is a Senior Consultant and Master Black Belt at Decisioneering, Inc. His primary activities include the development and delivery of training courses for Decisioneering. Additionally, Karl delivers consulting services and support for the analysis, design, and implementation of spreadsheet models using Crystal Ball. Prior to his current role, he was Design for Six Sigma Master Black Belt at Lear Corporation, where he mentored DFSS project teams, taught DFSS principles, and oversaw creation and confirmation of predictive models. He has applied DFSS on a wide range of automotive interior products such as seats, instrument panels, door panels, and headliners. His past assignments include managing CAE groups in Sweden and U.S.A. and serving as a Transactional Six Sigma Black Belt. He received a B.S. in Aeronautical/Astronautical Engineering from M.I.T. in 1985 and has a broad range of engineering knowledge within the aerospace and automotive industries.

Pa r t I V Appendixes Appendix A

Statistical Distribution Tables

Appendix B

Glossary

Appendix C

References

937

This page intentionally left blank

A Statistical Distribution Tables Table A-1: t Distribution Critical Values (tα for a one-tail test; divide alpha by 2 for two-tailed tests – tα/2) d.f.

t0.100

t0.050

t0.025

t0.010

t0.005

d.f.

1

3.078

6.314

12.706

31.821

63.657

1

2

1.886

2.920

4.303

6.965

9.925

2

3

1.638

2.353

3.182

4.541

5.841

3

4

1.533

2.132

2.776

3.747

4.604

4

5

1.476

2.015

2.571

3.365

4.032

5

6

1.440

1.943

2.447

3.143

3.707

6

7

1.415

1.895

2.365

2.998

3.499

7

8

1.397

1.860

2.306

2.896

3.355

8

9

1.383

1.833

2.262

2.821

3.250

9

10

1.372

1.812

2.228

2.764

3.169

10

11

1.363

1.796

2.201

2.718

3.106

11

12

1.356

1.782

2.179

2.681

3.055

12

13

1.350

1.771

2.160

2.650

3.012

13

14

1.345

1.761

2.145

2.624

2.977

14

15

1.341

1.753

2.131

2.602

2.947

15

16

1.337

1.746

2.120

2.583

2.921

16

17

1.333

1.740

2.110

2.567

2.898

17

continues 939

940

Appendix A Table A-1: Continued d.f.

t0.100

t0.050

t0.025

t0.010

t0.005

d.f.

18

1.330

1.734

2.101

2.552

2.878

18

19

1.328

1.729

2.093

2.539

2.861

19

20

1.325

1.725

2.086

2.528

2.845

20

21

1.323

1.721

2.080

2.518

2.831

21

22

1.321

1.717

2.074

2.508

2.819

22

23

1.319

1.714

2.069

2.500

2.807

23

24

1.318

1.711

2.064

2.492

2.797

24

25

1.316

1.708

2.060

2.485

2.787

25

26

1.315

1.706

2.056

2.479

2.779

26

27

1.314

1.703

2.052

2.473

2.771

27

28

1.313

1.701

2.048

2.467

2.763

28

29

1.311

1.699

2.045

2.462

2.756

29

infinity

1.282

1.645

1.960

2.326

2.576

infinity

Statistical Distribution Table s Table A-2: Chi-Square (χ2) Distribution Critical Values Left-tail Test (1 - α) 2 χ20.90 0.99 χ 0.95

Right-tail Test (α) χ20.05 χ20.01 0.10

d.f.

χ2

χ2

1

0.00016

0.0039

0.0158

2.71

3.84

6.63

1

2

0.0201

0.1026

0.2107

4.61

5.99

9.21

2

3

0.115

0.352

0.584

6.25

7.81

11.34

3

4

0.297

0.711

1.064

7.78

9.49

13.28

4

5

0.554

1.15

1.61

9.24

11.07

15.09

5

6

0.872

1.64

2.20

10.64

12.59

16.81

6

7

1.24

2.17

2.83

12.02

14.07

18.48

7

8

1.65

2.73

3.49

13.36

15.51

20.09

8

9

2.09

3.33

4.17

14.68

16.92

21.67

9

10

2.56

3.94

4.87

15.99

18.31

23.21

10

11

3.05

4.57

5.58

17.28

19.68

24.73

11

12

3.57

5.23

6.30

18.55

21.03

26.22

12

13

4.11

5.89

7.04

19.81

22.36

27.69

13

14

4.66

6.57

7.79

21.06

23.68

29.14

14

15

5.23

7.26

8.55

22.31

25.00

30.58

15

16

5.81

7.96

9.31

23.54

26.30

32.00

16

18

7.01

9.39

10.86

25.99

28.87

34.81

18

20

8.26

10.85

12.44

28.41

31.41

37.57

20

24

10.86

13.85

15.66

33.20

36.42

42.98

24

30

14.95

18.49

20.60

40.26

43.77

50.89

30

40

22.16

26.51

29.05

51.81

55.76

63.69

40

60

37.48

43.19

46.46

74.40

79.08

88.38

60

120

86.92

95.70

100.62

140.23

146.57

158.95

120

d.f.

941

942

Table A-3: F Distribution Critical Values (ANOVA) for α = 0.05 Degrees of Freedom (d.f.) in Numerator for α = 0.05

d.f. in 1

2

3

4

5

6

7

8

9

10

12

15

1

161.4

199.5

215.7

224.6

230.2

234.0

236.8

238.9

240.5

241.9

243.9

245.9

2

18.51

19.0

19.16

19.25

19.30

19.33

19.35

19.37

19.38

19.40

19.41

19.43

3

10.13

9.55

9.28

9.12

9.01

8.94

8.89

8.85

8.81

8.79

8.74

8.70

4

7.71

6.94

6.59

6.39

6.26

6.16

6.09

6.04

6.00

5.96

5.91

5.86

5

6.61

5.79

5.41

5.19

5.05

4.95

4.88

4.82

4.77

4.74

4.68

4.62

6

5.99

5.14

4.76

4.53

4.39

4.28

4.21

4.15

4.10

4.06

4.00

3.94

7

5.59

4.74

4.35

4.12

3.97

3.87

3.79

3.73

3.68

3.64

3.57

3.51

8

5.32

4.46

4.07

3.84

3.69

3.58

3.50

3.44

3.39

3.35

3.28

3.22

9

5.12

4.26

3.86

3.63

3.48

3.37

3.29

3.23

3.18

3.14

3.07

3.01

10

4.96

4.10

3.71

3.48

3.33

3.22

3.14

3.07

3.02

2.98

2.91

2.85

11

4.84

3.98

3.59

3.36

3.20

3.09

3.01

2.95

2.90

2.85

2.79

2.72

12

4.75

3.89

3.49

3.26

3.11

3.00

2.91

2.85

2.80

2.75

2.69

2.62

13

4.67

3.81

3.41

3.18

3.03

2.92

2.83

2.77

2.71

2.67

2.60

2.53

14

4.60

3.74

3.34

3.11

2.96

2.85

2.76

2.70

2.65

2.60

2.53

2.46

15

4.54

3.68

3.29

3.06

2.90

2.79

2.71

2.64

2.59

2.54

2.48

2.40

Appendix A

Denominator

Degrees of Freedom (d.f.) in Numerator for α = 0.05

d.f. in Denominator

20

30

40

50

60

Infinity

20

2.12

2.04

1.99

1.96

1.95

1.84

30

1.93

1.84

1.79

1.76

1.74

1.62

40

1.84

1.74

1.69

1.66

1.64

1.51

50

1.78

1.69

1.63

1.60

1.58

1.44

60

1.75

1.65

1.59

1.56

1.53

1.39

Infinity

1.57

1.46

1.39

1.35

1.32

1.00

Statistical Distribution Table s 943

944

Table A-4: F Distribution Critical Values (ANOVA) for α = 0.025 Degrees of Freedom (d.f.) in Numerator for α = 0.025

d.f. in 1

2

3

4

5

6

7

8

9

10

12

15

1

647.8

799.5

864.2

899.6

921.8

937.1

948.2

956.7

963.3

968.6

976.7

984.9

2

38.51

39.00

39.17

39.25

39.30

39.33

39.36

39.37

39.39

39.40

39.41

39.43

3

17.44

16.04

15.44

15.10

14.88

14.73

14.62

14.54

14.47

14.42

14.34

14.25

4

12.22

10.65

9.98

9.60

9.36

9.20

9.07

8.98

8.90

8.84

8.75

8.66

5

10.01

8.43

7.76

7.39

7.15

6.98

6.85

6.76

6.68

6.62

6.52

6.43

6

8.81

7.26

6.60

6.23

5.99

5.82

5.70

5.60

5.52

5.46

5.37

5.27

7

8.07

6.54

5.89

5.52

5.29

5.12

4.99

4.90

4.82

4.76

4.67

4.57

8

7.57

6.06

5.42

5.05

4.82

4.65

4.53

4.43

4.36

4.30

4.20

4.10

9

7.21

5.71

5.08

4.72

4.48

4.32

4.20

4.10

4.03

3.96

3.87

3.77

10

6.94

5.46

4.83

4.47

4.24

4.07

3.95

3.85

3.78

3.72

3.62

3.52

11

6.72

5.26

4.63

4.28

4.04

3.88

3.76

3.66

3.59

3.53

3.43

3.33

12

6.55

5.10

4.47

4.12

3.89

3.73

3.61

3.51

3.44

3.37

3.28

3.18

13

6.41

4.97

4.35

4.00

3.77

3.60

3.48

3.39

3.31

3.25

3.15

3.05

14

6.30

4.86

4.24

3.89

3.66

3.50

3.38

3.29

3.21

3.15

3.05

2.95

15

6.20

4.77

4.15

3.80

3.58

3.41

3.29

3.20

3.12

3.06

2.96

2.86

Appendix A

Denominator

Degrees of Freedom (d.f.) in Numerator for α = 0.025

d.f. in Denominator

20

30

40

50

60

Infinity

20

2.46

2.35

2.29

2.25

2.22

2.09

30

2.20

2.07

2.01

1.97

1.94

1.79

40

2.07

1.94

1.88

1.83

1.80

1.64

50

1.99

1.87

1.80

1.76

1.72

1.55

60

1.94

1.82

1.74

1.70

1.67

1.48

Infinity

1.71

1.57

1.48

1.43

1.39

1.00

The following table is used to calculate area or probability for a left-tail test of the normal curve at the different Z values. For the area under the curve to the right of the Z-critical value, subtract the table value from 1. α

-Z critical

Statistical Distribution Table s

Left-tailed test

TIP

945

946

Table A-5: Standard Normal Table (Z values) X.X0

X.X1

X.X2

X.X3

X.X4

X.X5

X.X6

X.X7

X.X8

X.X9

-3.0

.00135

.00131

.00126

.00122

.00118

.00114

.00111

.00107

.00104

.00010

-2.9

.0019

.0018

.0017

.0017

.0016

.0016

.0015

.0015

.0014

.0014

-2.8

.0026

.0025

.0024

.0023

.0023

.0022

.0021

.0021

.0020

.0019

-2.7

.0035

.0034

.0033

.0032

.0031

.0030

.0029

.0028

.0027

.0026

-2.6

.0047

.0045

.0044

.0043

.0041

.0040

.0039

.0038

.0037

.0036

-2.5

.0062

.0060

.0059

.0057

.0055

.0054

.0052

.0051

.0049

.0048

-2.4

.0082

.0080

.0078

.0075

.0073

.0071

.0069

.0068

.0066

.0064

-2.3

.0107

.0104

.0102

.0099

.0096

.0094

.0091

.0089

.0087

.0084

-2.2

.0139

.0136

.0132

.0129

.0125

.0122

.0119

.0116

.0113

.0110

-2.1

.0179

.0174

.0170

.0166

.0162

.0158

.0154

.0150

.0146

.0143

-2.0

.0228

.0222

.0217

.0212

.0207

.0202

.0197

.0192

.0188

.0183

-1.9

.0287

.0281

.0274

.0268

.0262

.0256

.0250

.0244

.0239

.0233

-1.8

.0359

.0351

.0344

.0336

.0329

.0322

.0314

.0307

.0301

.0294

-1.7

.0446

.0436

.0427

.0418

.0409

.0401

.0392

.0384

.0375

.0367

-1.6

.0548

.0537

.0526

.0516

.0505

.0495

.0485

.0475

.0465

.0455

Appendix A

Z

X.X0

X.X1

X.X2

X.X3

X.X4

X.X5

X.X6

X.X7

X.X8

X.X9

-1.5

.0668

.0655

.0643

.0630

.0618

.0606

.0594

.0582

.0571

.0559

-1.4

.0808

.0793

.0778

.0764

.0749

.0735

.0721

.0708

.0694

.0681

-1.3

.0968

.0951

.0934

.0918

.0901

.0885

.0869

.0853

.0838

.0823

-1.2

.1151

.1131

.1112

.1093

.1075

.1057

.1038

.1020

.1003

.0985

-1.1

.1357

.1335

.1314

.1292

.1271

.1251

.1230

.1210

.1190

.1170

-1.0

.1587

.1562

.1539

.1515

.1492

.1469

.1446

.1423

.1401

.1379

-0.9

.1841

.1814

.1788

.1762

.1736

.1711

.1685

.1660

.1635

.1611

-0.8

.2119

.2090

.2061

.2033

.2005

.1977

.1949

.1922

.1894

.1867

-0.7

.2420

.2389

.2358

.2327

.2297

.2266

.2236

.2207

.2177

.2148

-0.6

.2743

.2709

.2676

.2643

.2611

.2578

.2546

.2514

.2483

.2451

-0.5

.3085

.3050

.3015

.2981

.2946

.2912

.2877

.2843

.2810

.2776

-0.4

.3446

.3409

.3372

.3336

.3300

.3264

.3228

.3192

.3156

.3121

-0.3

.3821

.3783

.3745

.3707

.3669

.3632

.3594

.3557

.3520

.3483

-0.2

.4207

.4168

.4129

.4090

.4052

.4013

.3974

.3936

.3897

.3859

-0.1

.4602

.4562

.4522

.4483

.4443

.4404

.4364

.4325

.4286

.4247

0.0

.5000

.4960

.4920

.4880

.4840

.4801

.4761

.4721

.4681

.4641

Statistical Distribution Table s

Z

947

948 Appendix A

Table A-6: Standard Normal Table (Z values)

Right-tailed test

TIP The following table is used to calculate area or probability for a right-tail test of the normal curve at the different Z values. For the area under the curve to the left of the Z-critical value, subtract the table value from 1.

α

Z critical

Z

X.X0

X.X1

X.X2

X.X3

X.X4

X.X5

X.X6

X.X7

X.X8

X.X9

0.0

.5000

.5040

.5080

.5120

.5160

.5199

.5239

.5279

.5319

.5359

0.1

.5398

.5438

.5437

.5517

.5557

.5596

.5636

.5675

.5714

.5753

0.2

.5793

.5832

.5871

.5910

.5948

.5987

.6026

.6064

.6103

.6141

0.3

.6179

.6217

.6255

.6293

.6331

.6368

.6406

.6443

.6480

.6517

0.4

.6554

.6591

.6628

.6664

.6700

.6736

.6772

.6808

.6844

.6879

0.5

.6915

.6950

.6985

.7019

.7054

.7088

.7123

.7157

.7190

.7224

X.X0

X.X1

X.X2

X.X3

X.X4

X.X5

X.X6

X.X7

X.X8

X.X9

0.6

.7257

.7291

.7324

.7357

.7389

.7422

.7454

.7486

.7517

.7549

0.7

.7580

.7611

.7642

.7673

.7704

.7734

.7764

.7794

.7823

.7852

0.8

.7881

.7910

.7939

.7967

.7995

.8023

.8051

.8079

.8106

.8133

0.9

.8159

.8186

.8212

.8238

.8264

.8289

.8315

.8340

.8365

.8389

1.0

.8413

.8438

.8461

.8485

.8508

.8531

.8554

.8577

.8599

.8621

1.1

.8643

.8665

.8686

.8708

.8729

.8749

.8770

.8790

.8810

.8830

1.2

.8849

.8869

.8888

.8907

.8925

.8944

.8962

.8990

.8997

.9015

1.3

.9032

.9049

.9066

.9082

.9099

.9115

.9131

.9147

.9162

.9177

1.4

.9192

.9207

.9222

.9236

.9251

.9265

.9279

.9292

.9306

.9319

1.5

.9332

.9345

.9357

.9370

.9382

.9394

.9406

.9418

.9429

.9441

1.6

.9452

.9463

.9474

.9484

.9495

.9505

.9515

.9525

.9535

.9545

1.7

.9554

.9564

.9573

.9582

.9591

.9599

.9608

.9616

.9625

.9633

1.8

.9641

.9649

.9658

.9664

.9671

.9678

.9686

.9693

.9699

.9706

1.9

.9713

.9719

.9726

.9732

.9738

.9744

.9750

.9756

.9761

.9767

continues

Statistical Distribution Table s

Z

949

950

Z

X.X0

X.X1

X.X2

X.X3

X.X4

X.X5

X.X6

X.X7

X.X8

X.X9

2.0

.9773

.9778

.9783

.9788

.9793

.9798

.9803

.9808

.9812

.9817

2.1

.9821

.9826

.9830

.9834

.9838

.9842

.9846

.9850

.9854

.9857

2.2

.9861

.9864

.9868

.9871

.9875

.9878

.9881

.9884

.9887

.9890

2.3

.9893

.9896

.9898

.9901

.9904

.9906

.9909

.9911

.9913

.9916

2.4

.9918

.9920

.9922

.9925

.9927

.9929

.9931

.9932

.9934

.9936

2.5

.9938

.9940

.9941

.9943

.9945

.9946

.9948

.9949

.9951

.9952

2.6

.9953

.9955

.9956

.9957

.9959

.9960

.9961

.9962

.9963

.9964

2.7

.9965

.9966

.9967

.9968

.9969

.9970

.9971

.9972

.9973

.9974

2.8

.9974

.9975

.9976

.9977

.9977

.9978

.9979

.9979

.9980

.9981

2.9

.9981

.9982

.9983

.9983

.9983

.9984

.9985

.9985

.9986

.9986

3.0

.99865

.99869

.99874

.99878

.99882

.99886

.99889

.99893

.99896

.99900

Appendix A

Table A-6: Continued

B Glossary

Accuracy: The precision with which a target is reached. Accuracy is measured as the difference between the outcome and a target or standard value. Affinity Diagram: A tool used to gather and group ideas; usually depicted as a Tree diagram. Alias: A synonym for “confounded”; a common DOE term. Alternative hypothesis (Ha): A statement of what is desired to be concluded through sample data. The alternative statement is dependent on the Null hypothesis. ANOM: (Analysis of Mean) An analytical process that quantifies the mean response for each individual Control Factor level. ANOM can be performed on data that is in regular engineering units or data that has been transformed into some form of signal-to-noise ratio or other data transform. Main effects and interaction plots are created from ANOM data. ANOVA: (Analysis of Variance) An analytical process that decomposes the contribution each individual control factor has on the overall experimental response. The ANOVA process is also capable of accounting for the contribution of interactive effects between control factors and experimental error in the response provided enough degrees of freedom are established in the experimental array. The value of Epsilon Squared (% contribution to overall CFR variation) is calculated using data from ANOVA. Array: An arithmetically derived matrix or table of rows and columns that is used to impose an order for efficient experimentation. The rows contain the individual experiments. The columns contain the experimental factors and their individual levels or set points. 951

952

Glossar y

ASQ: American Society for Quality, an associate recognized for its work in promoting quality and six sigma; www.asq.org/. Attribute (or discrete) data: Has clear boundaries between values. There are several types of discrete data—count (for example, number of defects), nominal with more than two categories (e.g. people’s names, colors), binary with a rank order (yes/no, good/better/best), and binary without rank order (male/female, heads/tails). Balanced design: A classic DOE term that describes when the test design involves an equal number of runs for each combination of each setting (or level) for each factor. Benchmarking: The process of comparative analysis between two or more concepts, components, subassemblies, subsystems, products or processes. The goal of Benchmarking is to qualitatively and quantitatively identify a superior subject within the competing choices. Often the benchmark is used as a standard to meet or surpass. Benchmarks are used in building Houses of Quality, Concept Generation, and the Pugh Concept Selection Process. Best Practice: A preferred and repeatable action or set of actions completed to fulfill a specific requirement or set of requirements. Often used during planning phases, including within a product development process. Beta (β): The Greek letter, β, is used to represent the slope of a best fit line. It indicates the linear relationship between the signal factor(s) (Critical Adjustment Parameters) and the measured Critical Functional Response in a dynamic robustness optimization experiment. Bias: The difference between a standard value and the average of multiple measurements taken of different items. Black Belt: A job title or role indicating that the person has been certified as having mastered the Six Sigma DMAIC (Define-Measure-AnalyzeImprove-Control) content and demonstrated expertise in leading one or more projects. Title usually designates the team leader of a Six Sigma project and is often a coach of Green Belts. Blocking: A technique used in classical DOE to remove the effects of unwanted, assignable cause noise or variability from the experimental response so that only the effects from the control factors are present in the response data. Blocking is a data purification process used to help assure the integrity of the experimental data used in constructing a statistically significant math model. Capability Growth Index (CGI): The calculated percentage between 0% and 100% that a group of System, Subsystem, or Subassembly CFRs have attained in getting their Cp Indices to equal a value of 2 (indicating how

Confidence Inter val

well their Critical Functional Responses have attained Six Sigma performance during Product Development). The CPI for critical functions is a metric often found on an Executive Gate Review scorecard. Capability Index: Cp and Cpk Indices that calculate the ratio of the Voice of the Customer versus the Voice of the Product or Process. Cp is a measure of capability based on short-term or small samples of data—usually what is available during Product development. Cpk is a measure of longterm or large samples of data that include not only variation about the mean, but also the shifting of the mean itself—usually available during steady state production. Checklist: A simple list of action items, steps, or elements needed to complete a task. Each item is “checked off” as it is completed. Classical Design of Experiments (DOE): Experimental methods employed to construct math models relating a dependent variable (the measured Critical Functional Response) to the set points of any number of independent variables (the experimental control factors). DOE is used sequentially to build knowledge of fundamental functional relationships (Ideal/Transfer Functions) between various factors and a response variable. Coefficient of Determination (r2): Determines how well the scatter plot’s best fit line fits the data; sometimes called R-square. See Regression Analysis. Commercialization: A business process that harnesses the resources of a company in the endeavor of conceiving, developing, designing, optimizing, certifying design and process capability, producing, selling, distributing, and servicing a product. Compensation: The use of feedforward or feedback control mechanisms to intervene when certain noise effects are present in a product or process. The use of compensation is only done when insensitivity to noise cannot be attained through robustness optimization. Component: A single part in a subassembly, subsystem or system. An example would be a stamped metal part prior to having anything assembled to it. Component Requirements Document: The document that contains all the requirements for a given component. They are often converted into a Quality Plan that is given to the production supply chain to set the targets and constrain the variation allowed in the incoming components. Confidence Interval: The probability range within which the population data (parameters) exists. This region can be described two ways, either a single limit (an upper or lower) or a two-sided limit (both an upper and lower). Outside these limits is defined as the “rejection region,” wherein the “test variable” is said to not be part of the population.

953

954

Glossar y

Confidence Level: Determined by the formula one minus alpha (or 1- α), to connote a value with a degree of statistical significance or certainty. Confounded: A classic DOE term describing when the effects of two or more factors are indistinguishable from one another (also called “alias”). Conjoint Analysis: A structured technique that reveals the underlying customer preferences that guide purchasing decisions, based on trade-off decisions among combined alternatives (or bundles). Contingency table: A matrix used to organize categorical data in preparation for analysis such as a Chi-Square hypothesis test. The columns and rows represent two different ways to analyze the same set of data and are used to calculate degrees of freedom (i.e. df = [(# rows–1) x (# columns–1)]). Continuous (or variable) data: A type of data that has no boundaries between values, hence the number can be divided by two and still make sense. Continuous data in non-counting intervals and ratios such as time, height, weight, length, and temperature. Control Charts: A tool set used to monitor and control a process for variation over time. Varies with the type of data it monitors. Control Factor: The factors or parameters (CFP or CTF Spec.) in a design or process that the engineer can control and specify to define the optimum combination of set points for satisfying the Voice of the Customer. As defined by DOE, it is an input (X) of a process having an effect on a response and whose value can be easily selected. Correlation: A metric that measures the linear relationship between two process variables. Correlation describes the X and Y relationship with a single number (the Pearson’s Correlation Coefficient (r)), whereas regression summarizes the relationship with a line—the regression line. See Regression Analysis. Critical Adjustment Parameter (CAP): A specific type of CFP that controls the mean of a CFR. They are identified using sequential DOE and engineering analysis. They are the input parameters for Response Surface Methods for the optimization of mean performance optimization after Robust Design is completed. They enable Cpk to be set equal to Cp, thus enabling entitlement to be approached if not attained. Critical Functional Parameter (CFP): An input variable (usually an engineered additivity grouping) at the subassembly or subsystem level that controls the mean or variation of a CFR. Critical Functional Response (CFR): A measured scalar or vector (complete, fundamental, continuous engineering variable) output variable that is critical to the fulfillment of a critical (highly important) customer

Dashboard

requirement. Some refer to these critical customer requirements as CTQs. A metric often found on an Executive Gate Review scorecard. Critical-to-Function Specification (CTF): A dimension, surface or bulk characteristic (typically a scalar) that is critical to a component’s contribution to a subassembly, subsystem, or system-level CFR. Critical Parameter Management (CPM): The process that develops critical requirements and measures critical functional responses to design, optimize, and certify the capability of a product and its supporting network of manufacturing and service processes. Critical Path: The sequence of tasks in a project that take the greatest amount of time for completion. Critical Value: For hypothesis testing, the critical value demarcates the point at which the Null hypothesis (HO) can be rejected or not. It is determined by the given type of hypothesis test, its distribution, the confidence level (or significance value), and often the degrees of freedom. If the test statistic (or calculated statistic, based on the data) falls within the “non-rejection” region, then the Null hypothesis is not rejected, as shown in Figure G-1.

Test Distribution

Fail to Reject Ho: Calculated Calculated < Stastistic Stastistic For this One-tail Test

Non-Rejection Region

Reject Region

X Calculated Statistic

Critical Value

Figure G-1: Critical Value Illustration

Criticality: A measurable requirement or functional response that is highly important to a customer. All requirements are important, but only a few are truly critical. Cross-functional Team: A group of people representing multiple functional disciplines and possessing a wide variety of technical and experiential background and skills working together. Particularly applicable in the product commercialization process. See Multi-functional teams. CTQ: Critical to Quality translates the customer’s generic requirements into specific, actionable, and measurable definition that describes how a process player’s work helps to fulfill those requirements. Often displayed in either a tree or matrix format. Dashboard: A summary and reporting tool of critical success metrics and information about a process and/or product performance. Usually viewed as more complex than scorecards and depicts the critical parameters necessary to run the business.

955

956

Glossar y

Degrees of Freedom (df): A statistical term that represents the amount of freedom (or “float”) the data has to “represent” the population. The number of measurements that are independently available to estimate a population parameter. As data is collected, degrees of freedom (df) are earned. As statistical information is calculated (or a parameter is estimated), degrees of freedom are “spent” on describing the population from the sample data. The mean is a calculated statistic, it uses up one degree of freedom, resulting in (n-1) degrees of freedom, where “n” represents the sample size. If a population average is represented by four numbers, there is freedom for the first three numbers to be whatever they want to be, but the fourth number must be “dedicated” (as a calculation) to achieve the same population average. Hence, that fourth number does not have the freedom to be whatever it wants to be; so a four-number sample (n=4) has three degrees of freedom to describe the average (e.g. df = n–1; 4–1 = 3). Deliverable: A tangible, measurable output completed as an outcome of a task or series of tasks. Design Capability (Cpd): The Cp Index for a design’s Critical Functional Response in ratio to its Upper and Lower Specification Limits (VOCbased Tolerance Limits). Design of Experiments (DOE): A process for generating data that utilizes a mathematically derived matrix to methodically gather and evaluate the effect of numerous parameters on a response variable. Designed experiments, when properly used, efficiently produce useful data for model building or engineering optimization activities. It examines the interactions of multiple variables and is a preferred approach to One Factor At a Time (OFAT) experimentation. Deterioration Noise Factor: A source of variability that results in some form of physical deterioration or degradation of a product or process. This is also referred to as an “inner noise” because it refers to variation inside the controllable factor levels. DFSS: A Six Sigma concept used by the engineering technical community to design and develop a product. The acronym represents Design-For-SixSigma. Discrete (or attribute) data: A type of data that has clear, distinct boundaries between values. There are several types of discrete data—count (for example, number of defects), nominal with more than two categories (people’s names, colors), binary with a rank order (yes/no, good/better/best), and binary without rank order (male/female, heads/tails). DMADV: A 5-step Six Sigma method used primarily to re-design a broken process, as well as to solve problems and/or improve processes or products of defects. The acronym stands for Define-Measure-AnalyzeDesign-Validate.

Environmental Noise Factors

DMAIC: A 5-step Six Sigma method used to solve problems and/or improve processes or products of defects. The acronym stands for DefineMeasure-Analyze-Improve-Control. DMEDI : A 5-step method combining classic Six Sigma and Lean concepts to re-design a broken process, as well as to solve problems and/or improve processes or products of defects. The acronym stands for DefineMeasure-Explore-Develop-Implement. DPMO: Defects per Million Opportunities measurement, a calculation to count failures within a process and corresponds with a sigma level used in Six Sigma. Economic Coefficient: The economic coefficient is used in the Quality Loss Function. It represents the proportionality constant in the loss function of the average dollars lost (A0) due to a customer reaction to off-target performance and the square of the deviation from the target response (∆02). This is typically, but not exclusively, calculated when approximately 50% of the customers are motivated to take some course of economic action due to poor performance (but not necessarily out-right functional failure). This is often referred to as the LD 50 point in the literature. ECV: Expected Commercial Value; a financial metric often found on an Executive Gate Review scorecard. Energy Flow Map: A representation of an engineering system that shows the paths of energy divided into productive and non-productive work. Analogous to a free body diagram from an energy perspective. This accounts for the law of conservation of energy and is used in preparation to math modeling and design of experiments. Energy Transformation: The physical process a design or product system uses to convert some form of input energy into various other forms of energy that ultimately produce a measurable response. The measurable response may itself be a form of energy or the consequence of energy transformations, which have taken place within the design. Engineering Metrics: A scalar or vector that is usually called a CTF Spec., CFP, CAP, or a CFR. They are greatly preferred as opposed to Quality Metrics (yield, defects, and so on) in DFSS. Engineering Process: A set of disciplined, planned, and interrelated activities that are employed by engineers to conceive, develop, design, optimize, and certify the capability of a new product or process design. Environmental Noise Factors: Sources of variability that are due to effects that are external to the design or product, also referred to as “outer

957

958

Glossar y

noise.” They can also be sources of variability that one neighboring subsystem imposes on another neighboring subsystem or component. Examples include vibration, heat, contamination, misuse, overloading, and so on. Evolutionary Operation (EVOP): A type of DOE that describes a sequence of experimental designs wherein the cumulative learning (often small and incremental) helps to improve the prediction of future treatments to arrive at a better response. Experiment: An evaluation or series of evaluations that explore, define, quantify, and build data that can be used to model or predict functional performance in a component, subassembly, subsystem or product. Experiments can be used to build fundamental knowledge for scientific research or they can be used to design and optimize product or process performance in the engineering context of a specific commercialization process. Experimental Efficiency: This is a process-related activity that is facilitated by intelligent application of engineering knowledge and the proper use of designed experimental techniques. Examples include the use of fractional factorial arrays, control factors that are engineered for additivity and compounded noise factors. Experimental Error: The variability present in experimental data that is caused by meter error and drift, human inconsistency in taking data, random variability taking place in numerous noise factors not included in the noise array and control factors that have not been included in the inner array. In the Taguchi approach, variability in the data due to interactive effects is often but not always included as experimental error. Experimental Factors: Independent parameters that are studied in an orthogonal array experiment. Robust Design classifies experimental factors as either control factors or noise factors. Experimental Space: The combination of the entire control factor, noise factor and signal factor (CAP) levels that produce the range of measured response values in an experiment. 5Ms and P: Potential root cause memory trigger acronym used primarily for despersion problems, and represents Machines, Methods, Materials, Measurements, Mother Nature, and People. 4Ps: Potential root cause memory trigger acronym used primarily for services problems and represents Policies, Procedures, People, and Plant. F-Ratio: The ratio formed in the ANOVA process by dividing the mean square of each experimental factor effect by the MS of the error variance. This is the ratio of variation occurring between each of the experimental

Gate Reviews

factors in comparison to the variation occurring within all the experimental factors being evaluated in the entire experiment. It is a form of signalto-noise ratio in a statistical sense. The noise in this case is random experimental error—not variability due to the assignable cause noise factors in the Taguchi noise array. Feedback Control System: A method of compensating for the variability in a process or product by sampling output response and sending a feedback signal that changes a Critical Adjustment Parameter to put the mean of the response back on its intended target. FMEA: Failure Modes & Effects Analysis. A risk analysis technique that identifies and ranks the potential failure modes of a design or process and then prioritizes improvement actions. Fractional Factorial Design: A DOE design describing fewer combinations than a full design is tested to reduce the number of experimental runs needed, but at the cost of examining fewer interactions among the factors. Full Factorial Design: A DOE design wherein all the possible combinations of all levels of all factors are tested; none are omitted. Two- and three-level orthogonal arrays that include every possible combination between the experimental factors. Full factorial experimental designs use degrees of freedom to account for all the main effects and all interactions between factors included in the experimental array. Basically all of the interactions beyond two way interactions are likely to be of negligible consequence, thus there is little need to use large arrays to rigorously evaluate such rare and unlikely 3-way interactions and above. Fundamental: The property of a Critical Functional Response that expresses the basic or elemental physical activity that is ultimately responsible for delivering customer satisfaction. A response is fundamental if it does not mix mechanisms together and is uninfluenced by factors outside of the component, subassembly, subsystem, and system design or production process being optimized. Gantt Chart: A horizontal bar chart used for project planning and control that lists the necessary project activities as row headings against horizontal lines showing the dates and duration of each activity. Gate: A short period of time during a process when the team reviews and reacts to the results against requirements from the previous Phase and proactively plans for the smooth execution of the next Phase. Gate Reviews: Meetings with the project team and sponsors to inspect completed deliverables. Focus is on the results from specific tools and best practices and manages the associated risks and problems. They also make sure the team has everything they need to apply the tools and best

959

960

Glossar y

practices for the next Phase with discipline and rigor. A Gate Review’s time should be 20% reactive and 80% proactive. Goal Post Mentality: A philosophy about quality that accepts anything within the tolerance band (USL-LSL) as equally good and anything that falls outside of the tolerance band as equally bad. See Soccer, Hockey, LaCrosse, and Football rulebooks. Goal Statement: Identifies the critical parameters (including timeframe) for a targeted improvement. (Use SMART technique to ensure completeness.) GOSPA: Goals, Objectives, Strategies, Plans & Actions planning methodology. Grand Total Sum of Squares: The value obtained when squaring the response of each experimental run from a matrix experiment and then adding the squared terms together. Green Belt: A job title or role indicating that the person has been certified as having demonstrated an understanding of the basic Six Sigma DMAIC (Define-Measure-Analyze-Improve-Control) concepts. This role may support a Black Belt on a Six Sigma project, or in some companies work on a small-scale project directly related to their respective job. Histogram: A graphical display of the frequency distribution of a set of data. Histograms display the shape, dispersion, and central tendency of the distribution of a data set. House of Quality: An input—output relationship matrix used in the process of Quality Function Deployment. Hypothesis Testing: A statistical evaluation that checks the validity of a statement to a specified degree of certainty. These tests are done using well-known and quantified statistical distributions. IDEA: A 4-step Six Sigma method used by strategic marketing to define, develop, manage and refresh a portfolio of offerings (products and services). The acronym represents Identify-Define-Evaluate-Activate. Ideal/Transfer Function: Fundamental functional relationships between various engineering control factors and a measured critical functional response variable. The math model of Y = f(x) that represents the customer focused response that would be measured if there were no noise or only random noise acting on the design or process. Inbound Marketing: Marketing activities that are focused on providing deliverables for “internal consumption,” as opposed to deliverables intended for the marketplace. Independent Effect: The nature of an experimental factor’s effect on the measured response when it is acting independently with respect to any

Lagging Indicator

other experimental factor. When all control factors are producing independent effects, the design is said to be exhibiting an additive response. Inference: Drawing some form of conclusion about a measurable functional response based on representative or sample experimental data. Sample size, uncertainty, and the laws of probability play a major role in making inferences. Inner Array: An orthogonal matrix that is used for the control factors in a designed experiment and that is crossed with some form of outer noise array during Robust Design. Inspection: The process of examining a component, subassembly, subsystem, or product for off-target performance, variability, and defects either during product development or manufacturing. The focus is typically on whether the item under inspection is within the allowable tolerances or not. Like all processes, inspection itself is subject to variability, and outof-spec parts or functions may pass inspection inadvertently. Interaction: The dependence of one experimental DOE factor on the level set point of another experimental factor for its contribution to the measured response. There are two types of interaction: synergistic (mild to moderate and useful in its effect) and anti-synergistic (strong and disruptive in its effect). Interaction Graph: A plot of the interactive relationship between two experimental factors as they affect a measured response. The ordinate (vertical axis) represents the response being measured, and the abscissa (horizontal axis) represents one of the two factors being evaluated. The average response value for the various combinations of the two experimental factors is plotted. The points representing the second factor’s low level are connected by a line. Similarly, the points representing the second factor’s next higher level are connected by a line. IRR: Internal Rate of Return (%IRR); a financial metric often found on an Executive Gate Review scorecard. KJ Analysis: Stands for “Jiro Kawakita,” a Japanese anthropologist who treated attributes (or language) data in a similar manner as variables data by grouping and prioritizing it. A KJ diagram (similar to an Affinity diagram) focuses on the unique and different output, linking the critical customer priorities to the project team’s understanding and consensus. Kurtosis: A term that refers to a graphical distribution peakedness; it describes the amount that the curve’s middle is flattened or spiked. Lagging Indicator: An indicator that follows the occurrence of something; hence used to determine the performance of an occurrence or an event. By tracking lagging indicators, one reacts to the results. For example, the high and low temperature, precipitation, and humidity of a given day.

961

962

Glossar y

Leading Indicator: An indicator that precedes the occurrence of something; hence used to “signal” the upcoming occurrence of an event. By tracking leading indicators, one can prepare or anticipate the subsequent event and be proactive. For example, barometric pressure and doplar radar of surrounding region gives indication of ensuing weather. Lean Six Sigma: Modified Six Sigma approach to emphasize improving speed of a process by making “lean” its “non-value-add” steps. Started in a manufacturing environment, its concepts have expanded to multiple industries and applications. Its common metrics include zero wait time, zero inventory, line balancing, cutting batch sizes to improve flow through, and reducing overall process time. Least Squares method: A regression analysis technique that selects the minimum sum of the squared deviations about the regression line. The approach uses the smallest sum of all the squared residual values calculated for each data point to determine the regression line. Level: The DOE settings or value for a given factor. The set point at which a control factor, signal factor (CAP), or noise factor is placed during a designed experiment. Life Cycle Cost: The costs associated with making, supporting, and servicing a product or process over its intended life. Linear Combination: This term has a general mathematical definition and a specific mathematical definition associated with the dynamic robustness case. In general, a linear combination is the simple summation of terms. In the dynamic case it is the specific summation of the product of the signal level and its corresponding response. Linear Graph: A graphical aid used to assign experimental factors to specific columns when evaluating or avoiding specific interactions. Linear Regression: A quantitative model building tool that relates one or more independent variables (Xs) to a single dependent variable (Y). See “Regression Analysis.” Linearity: 1) A term that describes a measurement device’s inherent bias across its operating range, which is defined by its measurement scale. For example, a device may introduce variability at either extreme of its scale—bias relative to a true value. 2) The relationship between a dependent variable (the response) and an independent variable (such as the signal or control factor) that is graphically expressed as a straight line. Linearity is typically a topic within the dynamic cases of the robustness process and in Linear Regression analysis. LMAD: A 4-step Six Sigma method used by marketing to manage the ongoing operations of a portfolio of launched offerings (products and services) across the value chain. The acronym represents LaunchManage-Adapt-Discontinue.

Monte Carlo Simulaion

Loss to Society: The economic loss that society incurs when a product’s functional performance deviates from its targeted value. The loss is often due to economic action taken by the consumer reacting to poor product performance but can also be due to the effects that spread out through society when products fail to perform as expected. For example, a new car breaks down in a busy intersection due to a transmission defect, and 14 people are 15 minutes late for work (cascading loss to many points in society). Lower Specification Limit: The lowest functional performance set point that a design or component can attain before functional performance is considered unacceptable. Main Effect: The effect of a single DOE factor that is independent of any other factors. The contribution an experimental factor makes to the measured response independent of experimental error and interactive effects. The sum of the half effects for a factor is equal to the main effect. Manufacturing Process Capability (Cpm): The ratio of the manufacturing tolerances to the measured performance of the manufacturing process. Matrix: An array of experimental set points that is derived mathematically. The matrix is composed of rows (containing experimental runs) and columns (containing experimental factors). Matrix Experiment: A series of evaluations that are conducted under the constraint of a matrix. Mean: The average value of a sample of data that is typically gathered in a matrix experiment. Mean Square Deviation (MSD): A mathematical calculation that quantifies the average variation a response has with respect to a target value. Mean Square Error: A mathematical calculation that quantifies the variance within a set of data. Measured Response: The quality characteristic that is a direct measure of functional performance. Measurement Error: The variability in a data set that is due to poorly calibrated meters and transducers, human error in reading and recording data and normal, random effects that exist in any measurement system used to quantify data. Meter: A measurement device usually connected to some sort of transducer. The meter supplies a numerical value to quantify functional performance. Monte Carlo Simulation: A computer simulation technique that uses sampling from a random number sequence to simulate characteristics or events or outcomes with multiple possible values.

963

964

Glossar y

MSA: Measurement System Analysis tool to understand the level of reproducibility and repeatability. MTBF: Mean Time Between Failure measurement of the lapsed time from one failure to the next. Multi-disciplined Team: A group of people possessing a wide variety of technical and experiential background and skills working together. Particularly applicable in the product commercialization process. See Crossfunctional teams. Noise: Any source of variability. Typically noise is either external to the product (such as environmental effects), a function of unit to unit variability due to manufacturing, or it may be associated with the effects of deterioration. In this context, noise is an assignable, non-random cause of variation. Noise Factor: An input (X) to a process having an effect on a response but is not easily managed as defined in terms of DOE. Noise Directionality: A distinct upward or downward trend in the measured response depending on the level at which the noises are set. Noise factor set points can be compounded depending upon the directional effect on the response. Noise Experiment: An experiment designed to evaluate the strength and directionality of noise factors on a product or process response. Noise Factor: Any factor that promotes variability in a product or process. Normal Distribution: The symmetric distribution of data about an average point. The normal distribution takes on the form of a bell shaped curve. It is a graphic illustration of how randomly selected data points from a product or process response will mostly fall close to the average response with fewer and fewer data points falling farther and farther away from the mean. The normal distribution can also be expressed as a mathematical function and is often called a Gaussian Distribution. NPV: Net Present Value; a financial metric often found on an Executive Gate Review scorecard. NUD: Acronym representing New, Unique and Difficult customer requirements, that if fulfilled will delight them and help to outpace competition. Null Hypothesis (H0): A statement of what typically is desired to be disproved through sample data. Off-Line Quality Control: The processes included in pre-production commercialization activities. The processes of concept design, parameter

Outbound Marketing

design, and tolerance design make up the elements of off-line quality control. It is often viewed as the area where quality is designed into the product or process. One Factor at a Time (OFAT) Experiment: An experimental technique that examines one factor at a time, determines the best operational set point, locks in on that factor level, and then moves on to repeat the process for the remaining factors. This technique is widely practiced in scientific circles but lacks the circumspection and discipline provided by full and fractional factorial designed experimentation. Sometimes one factor at a time experiments are used to build knowledge prior to the design of a formal factorial experiment. On-line Quality Control: The processes included in the production phase of commercialization. The processes of statistical process control (loss function based and traditional), inspection and evolutionary operation (EVOP) are examples of on-line quality control. Operating Income: Calculated as gross profit minus operating expenses. A financial metric often found on an Executive Gate Review scorecard. Operational Definition: Defines what good looks like, often used for attribute data, and describes a standard criteria to measure against. Needs to be concise and easily understood by the full range of its users. It might include prose (or text) descriptions, pictures, photographs, models, or samples. Operational Marketing: Pertains to marketing’s activities in support of launching and managing an offering (product and/or service) or set of offerings across the value chain. Optimize: Finding and setting control factor levels at the point where their mean, standard deviation, or S/N (signal-to-noise) ratios are at the desired or maximum value. Optimized performance means the control factors are set such that the design is least sensitive to the effects of noise and the mean is adjusted to be right on the desired target. Orthogonal: The property of an array or matrix that gives it balance and the capability of producing data that allow for the independent quantification of independent or interactive factor effects. Orthogonal Array: A balanced matrix that is used to lay out an experimental plan that avoids any confounding of the main effects or any interactions; wherein each factor level appears the same number of times. A full factorial is described as balanced and orthogonal because the design represents an equal number of runs for each level and combinations of all the factors. Outbound Marketing: Marketing activities that are focused on providing deliverables for the customers as opposed to deliverables intended for “internal consumption.”

965

966

Glossar y

Outer Array: The orthogonal array used in dynamic robust design that contains the noise factors and signal factors. Each treatment combination of the control factors specified in the inner array is repeated using each of the treatment combinations specified by the outer array. p-Value: In hypothesis testing, the p-value represents the probability of the Null hypothesis (H0) being true, based on chance alone. It is a “goodness of fit” statistic. Typically the Null hypothesis states that the factors of interest are equal, or from the same population. If the p-value is large (>0.05 for a 95% confidence level), then there is insufficient evidence to reject the Null hypothesis of equality; therefore, fail to reject the Null hypothesis; therefore, presume they are the same. If the calculated pvalue < 0.05, for a 95% confidence test and the test statistic is beyond the critical value that determines the hypothesis is true, then there is sufficient evidence to reject the Null hypothesis; therefore, they are different. A small p-value indicates that it is very unlikely the Sample A Sample B items are the same (by chance alone), and the distance from the means is far apart. Figure G-2 Reject Region illustrates a small pvalue, where the Null hypothesis that Sample B +95%. A B is from the same populawhere p