What This Short Cut Covers
Development of a new product requires the product development team to address many complex customer requirements during the commercialization process. Consider a situation in which a new product being developed must meet specified upper and lower specification limits based on Voice of the Customer interviews. The design team must model and understand the sources of potential variation in the new product that need to be monitored and controlled if the product is to meet the identified customer needs. The process of analyzing component variation and designing a final product that meets customer tolerance requirements is known as statistical tolerancing.
In this short cut, various Design for Six Sigma techniques for determining the impact of multiple sources of variation on a final product are examined in detail. A procedure is described for using representative models for individual product components to estimate the expected overall level of variation in the performance of a final product.
Three methods of tolerance analysis are presented and the merits of each are discussed: Worst Case Analysis, Root Sum of Squares Analysis, and Six Sigma Tolerance Analysis. A detailed case study example, involving multiple sources of variation, is employed to illustrate the application of these methods. Minitab® is used to identify the best-fitting distributions from data sets for individual components. Monte Carlo Simulation with Crystal Ball® is then employed to determine the most important individual sources of variation and the overall variation of the final product. Finally, Crystal Ball’s OptQuest® optimization feature is utilized to determine the required design value for each key parameter to meet final customer requirements.
Introduction
During the commercialization process, we often have to determine the impact of multiple sources of variation on our final product. As we develop representative models for individual product components, we can use this information to estimate the overall level of variation we expect to find in our final product. The process of analyzing component variation and designing a final product that meets customer tolerance requirements is known as statistical tolerancing.
In order to demonstrate the concept of statistical tolerancing, in this short cut we provide an Excel Case Study and Minitab data file adapted from our forthcoming book Commercializing Great Products with Design for Six Sigma (2007, http://www.prenhallprofessional.com/title/0132385996). In order to follow the step-by-step analysis provided in the short cut, it is also necessary that the reader have copies of of Minitab statistical software and Crystal Ball simulation software. A copy of Crystal Ball’s OptQuest optimization routine is also required in order to perform the optimization analysis presented. Trial versions of Minitab (http://www.minitab.com) and Crystal Ball (http://www.decisioneering.com) may be downloaded from the Web sites provided.
Let’s suppose for moment that we’ve been asked to make some candy for an upcoming family gathering. As shown in Figure 1, we have selected a candy box and must determine how many pieces of candy we can put in the box given its length. Initially, it seems pretty obvious that we can fit four pieces of candy lengthwise in the box. But as we see in the figure, there is a gap between the candy and the end of the box. Is the gap important? Does the gap vary? How do we estimate the size of the gap? To answer these questions, we introduce three methods of tolerance analysis:
- Worst Case Analysis
- Root Sum of Squares (RSS)
- Six Sigma Tolerance Analysis
Figure 1 Candy box top view
Worst Case Analysis
The first type of tolerance analysis we consider is worst case analysis. In worst case tolerancing analysis, we must have additional information about the dimensions of the candy pieces and the candy box being used. As we see in Figure 2, each piece of candy has a nominal length of 1.240 inches, while the candy box is expected to have a length of 4.976 inches. Using this information, we can determine the number of candy pieces that will fit within the length of the box. We can also develop an initial estimate of the gap dimension, as shown in Figure 3. We expect that we will be able to put four pieces of candy in each row of the box given the box length dimension. In addition, we estimate that the nominal value of the gap will be 0.016 inches. But are things really this simple? As we also see in Figure 2, both the candy pieces and the box have variation of +/–0.003 inches around their nominal dimensions.
Figure 2 Candy and box dimensions: side view
Figure 3 Nominal gap value
Using the nominal dimensions and variation estimates for the candy pieces and the box, we can estimate the nominal and worst case values for the gap. As shown in Table 1, the minimum worst case estimate of the gap is calculated by taking the lowest value estimated for the box dimension and subtracting the highest values estimated for the candy pieces. Using this set of worst case assumptions, we determine that the minimum worst case gap is estimated to be 0.001 inches.
Table 1 Worst Case Gap Calculations
Similarly, the maximum gap can be estimated by using the maximum box dimension and the lowest candy dimensions. Using the given candy and box dimensions, we can estimate the worst case maximum gap to be 0.031 inches. A diagram representing the minimum and maximum worst case tolerance scenarios is presented in Figure 4.
Figure 4 Worst case gap estimates
We now understand that under the worst case assumption, the gap between the candy and the box is expected to be 0.016 inches, but we also understand that it could range from a minimum of 0.001 inches to a maximum of 0.031 inches. If we are in the process of commercializing a new candy product, should we proceed to production scale-up using these estimates? What are the chances that we will really have a worst case condition in a given box of candy? If we use the tolerance information of +/–0.003 inches as an estimate of +/–3 standard deviations for normally distributed data, the probability of receiving an abnormally long or abnormally short piece of candy at the extreme values of the range would be 1–0.9973 or 0.0027. For all four pieces of candy and the box to be at an extreme value the probability would be (0.0027)5 or 0.000000000000143. This is certainly a very unlikely event! Given the very low probability that either worst case scenario will occur, we turn to a more practical and useful tolerancing method, the root sum of squares analysis.
Root Sum of Squares Analysis
As we have seen, worst case tolerance analysis is interesting but the probability that a worst case scenario will actually occur is very low. To develop a more realistic tolerancing analysis we turn to the root sum of squares (RSS) technique. To demonstrate the RSS technique, let’s again examine the candy packaging arrangement presented in Figure 5.
Figure 5 Root sum of squares candy arrangement
As stated earlier, a worst case tolerance analysis occurs when each piece of candy and the box are each at their maximum levels of variation from the nominal. For our candy packaging example, the worst case value is calculated as:
A more realistic estimate of variation occurs when we take advantage of the statistical fact that variances are additive. Let’s now suppose that +/–0.003 inches represents +/–3 standard deviations of variation for each piece of candy and for the box. In this case, the standard deviation for each component is 0.001 inches. Using the additive property of variances, we can now estimate the standard deviation of the candy box assembly gap shown in Figure 6 as:
Figure 6 Candy box gap variation
If the tolerance for the gap is plus or minus 3 standard deviations, it can be calculated as:
Using our new value for root sum of squares tolerance, we can develop a new expected tolerance range for the gap, as seen in Figure 7. The nominal value for the gap is still 0.016 inches, as calculated for our worst case tolerance analysis. The root sum of squares tolerance value of 0.0067 inches is now used to calculate a minimum gap estimate of 0.0093 and a maximum estimate of 0.0227 inches.
Figure 7 Root sum of squares tolerance estimation
In Figure 8, we see that because these “tolerance” ranges have been set at +/–3 standard deviations, we would expect the gap to exceed the new minimum or maximum gap estimates 0.27% of the time.
Figure 8 Root sum of squares percent defective
The root sum of squares estimate of tolerance range is much tighter than that calculated using the worst case analysis approach. For this reason, we see that the probability of exceeding the minimum and maximum values of the gap is also much more realistic.
But should we set a tolerance range based on what our product can currently achieve in performance? The tolerance range should in fact be determined by the difference between the upper and lower specifications required to meet customer requirements. If we cannot meet those customer requirements consistently, we may be forced to seek relief from those tight specifications from the customer. In this case, the root sum of squares analysis gives us guidance to help in those specification negotiations. But our goal in design is to create Six Sigma products and processes. At best, our current product using the RSS approach is a 3 sigma level product unless we negotiate less stringent specification limits with the customer. Let’s next examine how we can use the candy box assembly information we have to design a Six Sigma product.
Six Sigma Tolerance Analysis
To develop a Six Sigma product without changing the product specs, we will have to reduce product variation. In this section, we demonstrate how the root sum of squares method can be used to design for Six Sigma product performance. We also introduce the use of Monte Carlo simulation for design in more complex tolerancing situations.
As described in our discussion of process capability in Chapter 28 of our forthcoming book, a Six Sigma process is defined as a process that has six short-term standard deviations between the process operating point and the closest specification limit. Using this definition, we can calculate the gap standard deviation required to make our candy box example a Six Sigma product. As shown in Figure 9, we calculate the required gap standard deviation by simply subtracting the lower spec limit from the upper spec limit and dividing the result by 12. The resulting gap standard deviation required for Six Sigma performance is 0.001117 inches. We can also estimate the standard deviation required for individual candy pieces and for the box itself by using the root sum of squares calculation. If the standard deviations of all components are equal, the root sum of squares calculation indicates that the standard deviation for each candy piece and for the box must be 0.0005 inches. In the short term, this Six Sigma process will produce out-of-spec conditions only 0.001 times out of a million. In the long term, assuming a 1.5 standard deviation process shift, the process will produce product outside of the tolerance range 3.4 times out of a million opportunities.
Figure 9 Producing a Six Sigma product with root sum of squares Analysis
Different Levels of Variation
While calculating the required standard deviation using root sum of squares analysis is relatively straightforward, we typically do not have equal standard deviations for each system component. In situations in which we have components with different levels of variation, tolerance analysis becomes more complex.
Let’s consider our candy box example again with a slight modification. Consider the situation in which variation is occurring in each piece of candy, the box, and the film used to wrap the candy. During product development, the Candy Wrapper Film design team has determined that a wrapper consisting of three layers of film produces the best candy taste for the customer. As shown in Figure 10, the new wrapper design consists of three film layers that unfortunately have different characteristics of variation. The design team has gathered data for the variation of each film layer’s thickness, and the data are presented in the Minitab worksheet Film Variation Data.MTW in the project file Statistical Tolerancing.MPJ.
Figure 10 Tolerance analysis with objects of differing variation
Identifying the Best Distribution Fit with Minitab
To evaluate the gap between the box and the candy with the new wrapper design, we must identify the best distribution fit for our film data. We take advantage of Minitab’s distribution identification tool and click Stat < Reliability / Survival < Distribution Analysis (Right Censoring) < Distribution ID Plot, as shown in Figure 11.
Figure 11 Identifying the best distribution fit with Minitab
We initially analyze the Film 1 data using the Weibull, Lognormal, Exponential, and Normal distributions, as described in Figure 12. Minitab offers additional distributions that we can use to further evaluate the data, if needed.
Figure 12 Film 1 distribution ID setup
In Figure 13, we see Minitab’s output from the distribution ID analysis. The best distribution fit is identified by examining both the graphical and analytical results presented. We see through visual inspection of the graphs developed by Minitab that the straight line for the Weibull distribution appears to fit the data best. Analytically, the Weibull distribution also has the highest value for the correlation coefficient, giving us verification that the Weibull distribution provides the best fit to the Film 1 data. To define the parameters for the Weibull distribution, we further analyze the data with Minitab’s Distribution Overview Analysis by clicking Stat < Reliability / Survival < Distribution Analysis (Right Censoring) < Distribution Overview Plot, as shown in Figure 14.
Figure 13 Film 1 distribution ID setup
Figure 14 Film 1 distribution overview analysis
The results of the distribution analysis are shown in Figure 15. Here, Minitab provides the information we need to define the Film 1 data distribution as we prepare to conduct a Monte Carlo simulation. Instead of mean and standard deviation, a Weibull distribution uses scale and shape parameters to define the best distribution fit for a given set of data. For the Film 1 data, we see that the shape parameter is calculated to be 0.587863 and the scale parameter is 0.156633.
Figure 15 Film 1 distribution overview analysis
A similar analysis for the Film 2 data suggests that the Film 2 data are normally distributed, as seen in Figure 16.
Figure 16 Film 2 distribution overview analysis
Further analysis of the Film 2 data with Minitab’s Distribution Overview Analysis, shown in Figure 17, indicates that the mean of the data is 0.125071 inches and the standard deviation is 0.0077098 inches.
Figure 17 Film 2 distribution overview analysis
Identifying the Best Distribution Fit with Crystal Ball
Analyzing the Film 3 data with the Distribution ID Plot does not yield an acceptable result, as shown in Figure 18. There is a significant disagreement between the data and the straight line for each distribution. We could reanalyze the data using other distributions available in Minitab, but we instead use a similar distribution fit function that is available in Crystal Ball.
Figure 18 Film 3 distribution overview analysis
To use Crystal Ball’s distribution fit capability we begin with the Distribution Gallery. At the bottom of the gallery, as shown in Figure 19, is the Fit button, which we click. We are then asked to identify the location of the data in the analysis worksheet. In Figure 20, we have indicated that the Film 3 data we wish to analyze are located in the cell range D5 to D104.
Figure 19 Fitting a distribution with Crystal Ball
Figure 20 Selecting the data range
The results of Crystal Ball’s analysis are shown in Figure 21. Crystal Ball has suggested that the data actually follow a uniform distribution. In a uniform distribution, any value between a specified minimum value and a specified maximum value occurs with equal probability. After further investigation, the Candy Wrapper Film design team has determined that the Film 3 data does indeed follow a uniform distribution between a minimum value of 0.050525 and a maximum of 0.603475 inches.
Figure 21 Crystal Ball Film 3 distribution analysis results
Monte Carlo Simulation in Statistical Tolerancing
We are now ready to perform a Monte Carlo simulation. We begin by setting up the assumptions for the simulation in Crystal Ball. In Figure 22, we see that we have defined Film 1 values as having a Weibull distribution with a scale parameter of 0.156633 and a shape parameter of 0.587863, Film 2 values have a normal distribution with a mean of 0.125071 and a standard deviation of 0.007623, and Film 3 values have a uniform distribution with a minimum value of 0.050525 and a maximum of 0.603475 inches.
Figure 22 Tolerancing Monte Carlo simulation setup
We must similarly define simulation assumptions for the candy pieces and the candy box. In Figure 23, we see that the candy lengths have been defined as having a normal distribution, with a mean of 0.865 inches and a standard deviation of 0.003 inches. The candy box length is also normally distributed, with a mean of 4.976 inches and a standard deviation of 0.005 inches.
Figure 23 Candy and box simulation assumptions
In Figure 24, we see the simulation worksheet model that we will use for the simulation. The model is contained in the DFSS Case Study Excel file in the tab labeled Statistical Tolerancing Initial. As discussed earlier, the box is long enough to hold four wrapped pieces of candy. The initial values in the worksheet are not really important, because they will change according to the defined simulation assumptions. As a general rule, the initial values should represent a reasonable set of conditions given our knowledge of the parameters being modeled.
Figure 24 Worksheet model setup
The model is arranged according to the physical setup of the candy box system. We have one box, four pieces of candy, and three layers of film on each candy piece. The model is constructed to add the candy and film values and to then subtract this sum from the box length. The resulting value is the gap we are investigating. A negative gap means that the candy and wrapper combination is too long for the box and will not fit without being reworked by an operator. A positive gap means that the sum of the lengths of the candy and wrapper is shorter than the gap. A large positive gap may result in possible damage to the candy in shipping. Damaged candy may be returned by the customer. Negative and positive gaps both have potential costs associated with them. In Figure 25, we see the cost equations associated with different gap values. Forecasts of both the gap value and the total cost associated with the gap are generated in the simulation.
Figure 25 Gap cost function
In Figure 26, we see the results of our initial gap and cost Monte Carlo simulation. According to Crystal Ball, we have a rather wide-ranging and costly gap. The gap appears to range from –2.5 inches to +0.5 inches. The median cost of the current design is approximately $1.325 per box of candy. Given the constraints of the design, we now use our model to determine whether we can achieve Six Sigma product performance. We also determine how much we can lower the cost of the new product with a Six Sigma tolerance design.
Figure 26 Initial gap and cost forecast
Optimizing Variation to Meet Customer Requirements
To perform Six Sigma tolerancing, we need to determine the required gap specifications for the design. We also must identify which parameters in the candy box system can be adjusted and within what limits during our tolerancing study. The design team has determined that the primary spec must be to keep the gap positive but below a value of 0.5 inches. It has been determined that positive gaps greater than this value result in significant levels of customer dissatisfaction, product returns, and lost market share. The team has also determined that there are three parameters in the design that may be adjusted as decision variables during our simulation, as shown in Figure 27. A decision variable is a variable that the design team feels can be controlled as needed in the product design. As shown in Figure 28, the team feels that the Film 1 Weibull shape factor can be set anywhere between 0.4 and 1.0 for the new product. Similarly, the team has identified ranges for the Film 2 standard deviation and the candy box mean.
Figure 27 Six Sigma tolerancing decision variables
Figure 28 Decision variable ranges
As shown in Figure 29, we have run Crystal Ball’s OptQuest program to select our optimization decision variables that can be changed and controlled in the final product. In the forecast section, we have asked OptQuest to minimize the cost associated with the resulting gap while requiring that the 99.9999966 percentile of the gap for the new design be less than 0.5 inches. This is equivalent to asking OptQuest to allow the gap to exceed our 0.5 spec only 3.4 times per million opportunities, which would be the expected long-term performance of a Six Sigma product relative to the gap upper spec.
Figure 29 Six Sigma tolerancing OptQuest setup
As shown in Figure 30, OptQuest is able to find several feasible improvements to the current product design during multiple iterations of the optimization.
Figure 30 Six Sigma tolerancing progress
In Figure 31, we see that a Six Sigma product design solution has been identified that significantly reduces the gap variation and the product cost. The cost associated with the gap has been reduced from $1.325 per box of candy to $0.9245. To accomplish these results, the Film 1 shape parameter must be set to a value of 0.907472, Film 2’s standard deviation set at 0.0108 inches, and the mean length of the candy box at 4.95958 inches. Using Monte Carlo simulation for Six Sigma tolerancing, we have been able to significantly improve our design for improved customer satisfaction while minimizing operating costs.
Figure 31 Final gap and cost forecast
What’s in the Book Commercializing Great Products with Design for Six Sigma?
Commercializing Great Products with Design for Six Sigma (www.prenhallprofessional.com/title/0132385996) is a unique book that demonstrates the business value of DFSS in today’s highly competitive business environment. Any business that strives for greatness must offer its customers a portfolio of great products. Successful development and commercialization of new products is required of all companies—not only for their growth, but for their mere survival. Because all products are subject to a product life cycle, companies not continuously updating product lines to meet the changing needs of key markets are faced with stagnation, diminished profits, and bankruptcy.
Commercializing Great Products with Design for Six Sigma is a complete look at the steps companies must follow in order to successfully bring new products to market. The book answers the following three fundamental questions:
- Why should I use Design for Six Sigma (DFSS) in new product commercialization?
- What steps and tools are required to commercialize products with DFSS and in what sequence should they be executed?
- How do I properly use the DFSS tools required to develop and bring new products to market?
Using the tools of DFSS, the book presents step-by-step instructions for business case development, market analysis, product concept development, product design, manufacturing scale-up, and product launch. This book will help business managers and design teams to identify the product concepts that are important to their customers and to efficiently translate those concepts into high-impact sources of new income. Along with a step-by-step discussion of key DFSS tools and roadmaps, the book contains a detailed case study example that illustrates tool execution and linkages. Supplementary materials accompanying the book include tool application examples in a complete Excel-based commercialization case study and data sets used to perform statistical analysis in Minitab and Crystal Ball.
Why We Wrote This Book
Having worked in industry developing new products for many years, we passionately believe that companies must stay on the cutting edge of product design in order to remain competitive in today’s global business environment. We wrote this book not only to inspire senior business leaders, marketing staff, and technical staff to expect great results from their new product development programs, but also to demonstrate how these results can be achieved. Through a detailed case study example, we demonstrate to leaders and practitioners alike how to apply the principles of DFSS in the identification and development of new products and services. In the text, we give step-by-step instructions along with easy-to-use templates and examples for the uses of required tools. We discuss and demonstrate the use of each tool in sequence, as shown in the DFSS commercialization roadmap presented in the book.
In Commercializing Great Products with Design for Six Sigma, we provide a practical, “how to” guide for the use of DFSS in product commercialization. The product development techniques and roadmaps presented in this book have evolved throughout our combined 65 years of experience in product commercialization. Many of the fundamental concepts presented were learned, developed, and enhanced during the courses of our individual careers. Randy Perry has worked in product commercialization for 25 years, including 18 years at AlliedSignal (now Honeywell), where, under the leadership of CEO Larry Bossidy, Six Sigma became a weapon to drive growth and productivity improvement. David Bacon, inspired as a graduate student by his former research supervisor George Box, has more than 40 years of experience as an engineering professor and industrial consultant. The tools and roadmaps described in this book continue to be expanded, refined, and improved through work with a diverse array of corporate clients and fellow consultants.
An Overview of the Content
Commercializing Great Products with Design for Six Sigma consists of five sections: (1) Getting Started, (2) Preparing the Business Plan, (3) The Voice of the Customer, (4) Product/Process Development, and (5) Product/Process Launch. Within these sections, the book contains 38 chapters and follows the development of a new product or service from business concept through final product launch. This section provides a brief description of each section and the chapters within it.
Section 1: Getting Started
Section Overview
In this section, we begin by summarizing the history of Six Sigma and of Design for Six Sigma before quickly moving into a discussion of key business infrastructure needed to support a successful commercialization program.
The section begins with a brief overview of how companies, markets, and products are constantly changing, and how these forces of change drive the need for new products. After a detailed discussion of how financial metrics are used to measure the value of DFSS, the first section concludes with a discussion of how to select new projects and manage the company’s new-product portfolio.
Chapter Overview
In Chapter 1, we begin with the overview, “What Is Design for Six Sigma?” In this chapter, we trace the history of Six Sigma and discuss various DFSS roadmaps in use for product commercialization today.
In Chapter 2, “The Business Case for DFSS,” we discuss why business management should aggressively work to implement DFSS in the company’s new product development processes. In this chapter, we demonstrate and discuss the devastating consequences of failing to continually replenish company’s pipeline of new products.
In Chapter 3, “Six Sigma Financial Metrics,” we present a detailed look at how to place a value on Design for Six Sigma projects. Assessing the financial value of DFSS projects is critical as we track the benefits realized by improving our knowledge of customer needs and reducing product development rework. In this chapter, we introduce the Candy Wrapper Film Case Study, which is used throughout the remainder of the book to illustrate precisely how and when required DFSS tools are to be executed.
In Chapter 4, “Project Identification and Portfolio Management,” we discuss the critical need for a dynamic project selection process. The commercialization pipeline of new products represents a company’s future. Careful tracking and management of this product portfolio using the methods discussed in this chapter are essential.
In Chapter 5, “Stage-Gate Processes,” we discuss the general concept behind the use of Stage-Gate in product commercialization. The benefits of using Stage-Gate to minimize the risk of using people, time, and money inefficiently on projects are examined.
In Chapter 6, “Project Management,” we discuss the need for project management discipline to complete the Stage-Gate deliverables. A review of good project management techniques is presented.
Section 2: Preparing the Business Plan
Section Overview
In Section 2, we deal with the preparation of a business plan for a new product. We discuss various key components of a business plan in detail, including performing market segmentation, identifying market opportunities, defining product value, and estimating the financial value of a project. We end this section with a discussion of how to best position a new product for success in the marketplace.
Chapter Overview
In Chapter 7, “Business Plan Overview,” the concept of developing a business plan to describe the business, marketing, and operating strategy for a new product is introduced. The contents of a good business plan are presented and reviewed.
In Chapter 8, “Market Segmentation,” the value of strategically grouping customers having similar characteristics and needs with the goal of improving overall business profitability is discussed. Methods and techniques for segmenting markets are presented.
In Chapter 9, “Identifying Market Opportunities,” two specific tools for examining new market opportunities—the Strengths, Weaknesses, Opportunities, and Threats (SWOT) analysis and the Market Failure Modes and Effects analysis (FMEA)—are discussed. Specific instructions and an example for execution of each of these tools are presented.
In Chapter 10, “Defining Product Value,” the concept of customer value is introduced. In this chapter, we discuss how customers buy products based on value, not based on price. A discussion of value chain mapping techniques and how this information can be used in making strategic decisions is presented.
In Chapter 11, “Estimating Financial Value,” methods to estimate the financial value for a product under development are discussed. Financial Excel models are constructed and sensitivity analyses using Crystal Ball are conducted.
In Chapter 12, “Product Positioning,” two primary tools for product positioning are discussed—the Market Perceived Quality Profile and the Product Positioning Map. The purpose of these tools is to establish what major product and service attributes most influence a customer’s decision to purchase products and then to define how our current products are positioned compared to those of competitors in these key requirement areas.
Section 3: The Voice of the Customer
Section Overview
In Section 3, we provide an in-depth discussion of how to gather and analyze “The Voice of the Customer.” In this section, we emphasize techniques to identify the business-critical needs of key customers, and then we explore the use of interview techniques that allow us to examine these needs more deeply. We continue our discussion in Section 3 with a detailed look at the use of KJ Analysis to determine which needs identified during customer interviews are most important. Later in the section, we examine new product ideation and concept generation/selection techniques. We end Section 3 with a detailed discussion of Quality Function Deployment (QFD) and how this key tool is used to develop key product and process specifications.
Chapter Overview
In Chapter 13, “Concept Development,” we discuss a series of specific tools tied together in a roadmap format with the intent of developing the best product to meet the needs of a given market. Concept development is a unique approach to product or service development and provides a structured methodology for dealing with the “fuzzy front end” of product development.
In Chapter 14, “Developing the Interview Guide,” we discuss a well-defined process for developing an interview guide to be used in interviewing customers.
In Chapter 15, “Conducting Customer Interviews,” specific techniques are presented for interviewing customers and collecting needed Voice of the Customer information.
In Chapter 16, “KJ Analysis,” we discuss the KJ process for analyzing Voice of the Customer interview results in order to capture the most important customer requirements for our new product or process.
In Chapter 17, “Relative Importance Survey,” we review the importance of a follow-up customer survey to confirm or modify the importance ratings of customer requirements resulting from the KJ Analysis. Specific examples of surveys and survey analysis techniques are presented.
In Chapter 18, “Ideation,” a method for developing innovative product solution ideas is discussed and demonstrated using the Candy Wrapper Film Case Study.
In Chapter 19, “Pugh Concept Selection,” the Pugh Concept method for selecting the best overall product concept is presented. A detailed example of how the Pugh method is executed is discussed.
In Chapter 20, “QFD,” the Quality Function Deployment (QFD) tool is reviewed in detail. Specific execution details for QFD are presented and the flow-down nature of QFD is demonstrated.
In Chapter 21, “TRIZ,” the use of the TRIZ (pronounced “TREEZ”) methodology—developed by the Russian engineer and scientist Genrich Altshuller to resolve significant technical conflicts identified in the QFD roof—is discussed.
In Chapter 22, “Critical Parameter Management,” the development and use of critical parameter scorecards to ensure that critical parameters identified through the QFD process meet process capability requirements are presented.
Section 4: Product/Process Development
Section Overview
Section 4 covers the fundamental technical tools needed for product and process development. This section begins with a discussion of Process Mapping and continues with detailed examination of the use of the Cause and Effects Matrix, Failure Modes and Effects Analysis, basic statistical tools, measurement systems analysis, process capability, tools for data analysis, design of experiments, robust design, mixture experiments, and multiple response optimization. The section ends with a review of how to scale up a process from pilot scale to full-scale production with a well-defined control plan.
Chapter Overview
In Chapter 23, “Process Mapping,” we demonstrate the techniques required to develop good process maps. We also demonstrate how process mapping interfaces with the QFD analysis.
In Chapter 24, “Cause and Effects Matrix,” the tools and techniques for development of the C&E Matrix are presented. In this chapter, we show how the C&E Matrix links to the QFD process.
In Chapter 25, “Failure Modes and Effects Analysis,” we discuss the process for identifying critical failure modes and their causes for both process design and manufacturing.
In Chapter 26, “Statistical Analysis Tools Overview,” we explore key fundamental statistical analysis techniques. Graphical and numerical analysis approaches using detailed Minitab instructions and output are presented.
In Chapter 27, “Measurement Systems Analysis,” we discuss the importance of good measurement systems in product development. In this chapter, we present step-by-step instructions and examples of how assessments of measurement systems are conducted using Minitab.
In Chapter 28, “Process Capability,” we discuss methods for determining how well product or process performance satisfies specifications. We present commonly used formulas for process capability and demonstrate how process capability analysis is conducted using Minitab.
In Chapter 29, “Tools for Data Analysis,” we demonstrate in detail techniques for identifying underlying relationships in data. Using Minitab and the Candy Wrapper Film Case Study, detailed instructions are given for a variety of statistical analysis techniques, including t tests, ANOVA, correlation, regression, and nonparametric statistical analysis. Discussions of confidence intervals, sample size calculation, and control charting are also presented.
In Chapter 30, “Design of Experiments,” we discuss techniques for conducting commonly used designed experiments. Full Factorial, Fractional Factorial, and Response Surface designs are discussed in detail.
In Chapter 31, “Robust Design,” we discuss concepts and methods for designing a product or process to resist the impact of noise. Specific robust design approaches and examples are presented.
In Chapter 32, “Mixture Experiments,” we discuss the use of experimental design techniques to determine the optimum formulation for a product that contains multiple components.
In Chapter 33, “Seeking an Optimal Solution,” approaches are presented for simultaneously optimizing multiple performance characteristics in product development. Techniques using Minitab®, Excel, and Crystal Ball are demonstrated using the Candy Wrapper Film Case Study.
In Chapter 34, “Design for Reliability (DfR),” we discuss techniques to test, analyze, and improve product reliability.
In Chapter 35, “Statistical Tolerancing,” we discuss methods to ensure that multiple components in an assembly or composite product are designed to meet assembled product specifications.
In Chapter 36, “Production Scale-Up,” we discuss techniques to ensure that a product meets Design for Manufacturability requirements.
In Chapter 37, “Control Plans,” we discuss the process for developing procedures to ensure that optimum product or process performance will be sustained as we move forward.
Section 5: Product Launch and Project Post-Mortem Analysis
Section Overview
The book ends with Section 5, in which several tools are described for execution after Product/Process Launch is completed. In this section, we discuss the generation of a post-launch follow-up report with key customers to ensure that the new product meets their requirements, and the need for a review of production yields compared to project targets. We conclude with a review of the post-mortem analysis process to capture improvement opportunities for future new product development projects.
Chapter Overview
In Chapter 38, “Product Launch and Project Post-Mortem Analysis,” we review the need to track the launch of a product in order to ensure successful commercialization with targeted customers. We also demonstrate techniques for conducting post-mortem project follow-up to ensure that project learnings are captured for use in future projects.
Summary
In summary, Commercializing Great Products with Design for Six Sigma contains a broad spectrum of valuable insights for improving the product commercialization process. The book is intended to
- Appeal to business management by providing a discussion of the business value of DFSS
- Address both marketing and technology activities in an integrated DFSS roadmap
- Provide a detailed step-by-step discussion of how to use each key DFSS tool
- Demonstrate tool usage with a complete case study utilized throughout the book
- Provide an easy-to-use DFSS tool template in Excel format for each key tool
By applying the methods presented in this book and illustrated by the case study examples, significant improvement in a company’s product development process can be quickly achieved.
Case Study
Commercializing Great Products with Design for Six Sigma demonstrates the product development process through the use of a detailed step-by-step case study. The case study begins with the identification of a new Candy Wrapper Film product idea. The case study is then used to illustrate detailed instructions for assessing the business opportunity, gathering the Voice of the Customer, and technically designing and manufacturing the product. The case study contains over 100 easy-to-use design templates and analysis files that can be modified for use in the development of any product.