3.5. Visual Six Sigma Data Analysis Process and Roadmap

As this quick tour of functionality may suggest, JMP has a very diverse set of features for both EDA and CDA. The earlier section "Visual Displays and Analyses Featured in the Case Studies" gave a preview of those parts of JMP that you will see again in later chapters. But, there is a danger that the section's techniques may have appeared as a laundry list, perhaps at odds with our contention in Chapter 2 that Visual Six Sigma is a way of combining such techniques to get value from data. Let's attempt to put the list of techniques into some context.

In Chapter 2, we discussed the outcome of interest to us, represented by Y, and the causes, or inputs that affect Y, represented by Xs. As we saw, Six Sigma practitioners often refer to the critical inputs, resources, or controls that determine Y as Hot Xs. Although many Xs have the potential to affect an outcome, Y, the data may show that only certain of these Xs actually have an impact on the variation in Y. In our credit card example from Chapter 2, whether a person is an only child or not may have practically no impact on whether that person responds to a credit card offer. In other words, the number of siblings is not a Hot X. However, an individual's income level may well be a Hot X.

Consider the Visual Six Sigma Data Analysis Process, illustrated in Exhibit 3.29, which was first presented in Chapter 2. In your Six Sigma projects, first you determine the Y or Ys of interest during the Frame Problem step. Usually, these are explicit in the project charter or they follow as process outputs from the process map. In Design for Six Sigma (DFSS) projects, the Ys are usually the Critical to Quality Characteristics (CTQs).

Figure 3.29. Visual Six Sigma Data Analysis Process

The Xs of potential interest are identified in the Collect Data step. To identify Xs that are potential drivers of the Ys, a team uses process maps, contextual knowledge, brainstorming sessions, cause-and-effect diagrams, cause-and-effect matrices, and other techniques. Once the Xs have been listed, you seek data that relate these Xs to the Ys. Sometimes these observational data exist in databases. Sometimes you have to begin data collection efforts to obtain the required information.

Once the data have been obtained, you face the issue of identifying the Hot Xs. This is part of the Uncover Relationships step. Once you have identified the Hot Xs, you may or may not need to develop an empirical model of how they affect the Ys. Developing this more detailed understanding is part of the Model Relationships step, which brings us back to the signal function, described in Chapter 2. You may need to develop a model that expresses the signal function for each Y in terms of the Hot Xs. Here we illustrate with r Hot Xs:

Only by understanding this relationship at an appropriate level can you set the Xs correctly to best manage the variation in Y.

Identifying the Hot Xs and modeling their relationship to Y (for each Y) is to a large extent the crux of the Analyze phase of a DMAIC project, and a large part of the Analyze and Design phases of a DMADV project.

Exhibit 3.30 shows an expansion of the Visual Six Sigma Roadmap that was presented in Chapter 2. Recall that this Roadmap focuses on the three Visual Six Sigma Data Analysis Process steps that most benefit from dynamic visualization: Uncover Relationships, Model Relationships, and Revise Knowledge. In this expanded version, we show how a subset of the techniques listed earlier in this chapter can be used in a coordinated way to accomplish the goals of Visual Six Sigma. In our experience, this represents an excellent how-to guide for green belts who are starting the Analyze phase of a traditional Six Sigma project, or anyone who is simply faced with the need to understand the relationship between some set of inputs and some outcome measure outside a DMAIC or DMADV framework.

Figure 3.30. Visual Six Sigma Roadmap

However, remember that the EDA approach to uncovering relationships requires an unfolding style of analysis where your next step is determined by your interpretation of the previous result. So, although it is a great starting point to guide your usage of JMP in many situations, Exhibit 3.30 should never be followed slavishly or without thought. As you gain more familiarity with your data and with JMP, you may well develop your own Visual Six Sigma style that works better for you and your business.