Moving beyond Traditional Six Sigma

It is our belief that the tools, techniques, and workflows in common use with Six Sigma efforts are typically not aligned with the key idea of discovery. In the early days of Six Sigma, relevant data rarely existed, and a team was often challenged to collect data on its own. As part of the Measure phase, a team usually conducted a brainstorming session to identify which features of a process should be measured. In some sense, this brainstorming session was the team's only involvement in hypothesis generation. The data collected were precious, and hypothesis testing methods were critical in separating signals from noise.

Project teams struggling with a lack of useful data generally rely on an abundance of subjective input, and often require hypothesis testing to minimize the risk of bad decisions. This emphasis on hypothesis testing is reasonable in an environment where data are sparse. In contrast, today's Six Sigma teams often find warehouses of data that are relevant to their efforts. Their challenge is to wade through the data to discover prominent features, to separate the remarkable from the unremarkable.

These data-rich environments call for a shift in emphasis from confirmatory methods, such as hypothesis testing, to exploratory methods, with a major emphasis on the display of data to reveal prominent features that are hidden in the data. Since the human interpretation of the data context is a vital part of the discovery process, these exploratory techniques cannot be fully automated. Also, with large quantities of data, hypothesis testing itself becomes less useful—statistical significance comes easily and may have little to do with practical importance.

Of course, the simple abundance of data in a warehouse does not guarantee its relevance for improvement or problem solving. In fact, it is our experience that teams working in what they believe to be data-rich environments sometimes find that the available data are of poor quality or are largely irrelevant to their efforts. Visualization methods can be instrumental in helping teams quickly reach this conclusion. In these cases, teams need to revert to techniques such as brainstorming, cause-and-effect diagrams, and process maps, which drive efforts to collect the proper data. But, as we shall see, even in situations where only few relevant data are available, visualization techniques, supported as appropriate by confirmatory methods, prove invaluable in identifying telling features of the data.

Making Data Analysis Lean

Discovery is largely supported by the generation of hypotheses—conjectures about relationships and causality. Today's Six Sigma teams, and data analysts in the business world in general, are often trained with a heavy emphasis on hypothesis testing, with comparatively little emphasis given to hypothesis generation and discovery. They are often hampered in their problem-solving and improvement efforts by the inability to exploit exploratory methods, which could enable them to make more rapid progress, often with less effort.

In recent times, we have seen incredible advances in visualization methods, supported by phenomenal increases in computing power. We strongly believe that the approaches now allowed by these methods are underutilized in current Six Sigma practice. It is this conviction that motivated us to write the first edition of this book and, following its success, to produce a second edition that takes advantage of recent software advances. We hope you find this book useful as you shape and build your own real-world Six Sigma experience.

Requirements of the Reader

This leads to another important point, namely, that you are “part of the system.” Discovery, whether practiced as an individual or as a team sport, involves both divergent and convergent thinking; both creativity and discipline are required at different times. You should bear this in mind when forming a team or when consulting with individuals, since each person will bring his or her own skill set, perspective, and strength to the discovery process.

Given the need to be data driven, we also need to recognize one of the basic rules of using data, which is that any kind of analysis that treats data simply as a list of numbers is doomed to failure. To say it differently: All data are contextual, and it is this context and the objectives set out for the project that must shape the analysis and produce useful recommendations for action. As a practitioner, your main responsibility should always be to understand what the numbers in the data actually mean in the real world. In fact, this is the only requirement for putting the ideas in this book into practice in your workplace.