Statistical Thinking: Improving Business Performance, Second Edition

Chapter 7

Using Process Experimentation to Build Models

Experiment, make it your motto day and night.

—Cole Porter

The focus of Chapter 6 was on building models using existing process data to understand how a process works and to better manage and improve it. Building models is an iterative process in which we move back and forth between hypotheses about how the process works and data that confirm or deny them. This strategy was illustrated in Chapter 2 in the statistical thinking model, and it is quite useful. However, this strategy is limited by the quality of the existing process data used to build the model.

In this chapter, we examine a second strategy, which is to proactively experiment with the process to obtain high-quality data. Using this approach, we can test our ideas about cause-and-effect relationships. The technique used to do this is the statistical design of experiments. Using this methodology, we define how the experiments should be conducted, what specific process changes should be made, what data should be collected, and how the data should be analyzed to build the model. The thought process is the same as that used in Chapter 6, but the data used to develop the model are collected by experimenting with the process rather than by observing the process as it operates normally.

Design of experiments was also shown in Chapter 4 to be a key tool in the process improvement and DMAIC frameworks when trying to understand cause-and-effect relationships. Design of experiments ...