What is the purpose of using the Six Sigma Methodology? Have you ever heard Six Sigma referred to as a problem-solving methodology? This is a disservice to both problem solving and Six Sigma. There are numerous problem-solving methodologies around, but they are focused on finding the cause of some exceptional event that has occurred. Six Sigma, on the other hand, is a process improvement methodology and goes far beyond problem solving.

Process improvement is intended to be a permanent change to the way a process works by reducing the impact of sources of variation on critical process characteristics. Just how to improve a process depends on the nature of the sources of variation in the process. The unique tool for categorizing this variation and thereby taking the most effective path to improvement is the process behavior chart, or a control chart.

For the majority of Six Sigma practitioners, process behavior charts are used only in the Control Phase of the DMAIC (Define–Measure–Analyze–Improve–Control) methodology. However, using the process behavior chart in the Measure and Analyze Phases provides critical process behavior information as a foundation for the use of the other statistical methods. The insight that these charts supply can lead to effective and sustainable process improvement.

From its inception, the Six Sigma DMAIC Methodology has placed an emphasis on using a variety of statistical techniques to determine how best to improve any process in an organization that is not operating to its potential or as expected. The focus of the techniques is the use of data for critical-to-quality characteristics (CTQs or Ys) to identify sources of variation in the process (Xs) and then eliminate or minimize their impact. This is summarized in the familiar equation Y = f(X), which is read as “Y is a function of X.”

In the Measure Phase of the methodology, a variety of graphical and numerical summary tools are recommended to use with baseline data as well as a determination of the Sigma Level of the process. These summaries provide a global picture of the variation in a process, but fall short of identifying the actual sources. Run charts are the best technique to observe patterns in data over time and connect the patterns with specific sources of variation in the process.

The primary emphasis in the Analyze and Improve Phases of the methodology is the use of statistical techniques that will highlight a cause-and-effect relationship between the CTQs of interest and the process variables (Xs). Statistical techniques such as correlation and many versions of regression, as well as a variety of types of designed experiments, are used in these phases.

Finally, in the Control Phase, process behavior charts are used to ensure that the gains from the newly improved process are sustained-while monitoring the process behavior for changes.

If this traditional approach to identifying and minimizing the impact of sources variation in a process has worked well for many organizations, why start using process behavior charts in the Measure and Analyze Phases? There are two main arguments to do so:

• In any process that is repetitive over time, the time order of the data may well contain the most valuable information for identifying specific sources of variation that impact the CTQs.
  
  
• The behavior of a process, predictable or unpredictable, is critical to the approach and ability to sustain improvement. A predictable process has only routine, or common cause, variation while an unpredictable process has both routine and exceptional or special cause variation.

Time Order of Data is Critical

The run chart traditionally used in the Measure Phase captures the patterns and process behavior for CTQs. However, without the control limits calculated from the data, there is a risk of misunderstanding the source of the patterns. Some perceived patterns may well be a result of only routine process variation instead of something exceptional that can be connected to a particular source of variation. Even with this potential pitfall, the run chart can be a more useful graphical technique than the histogram when working with baseline data.

An example of the benefits of the process behavior chart with the baseline data from a process came from a Six Sigma project, “On Time to Process a Loan Application.” The Black Belt for this project interpreted the histogram of the baseline data for 35 loan applications to be a skewed distribution and was considering how to transform the data to obtain a normal distribution. See the histogram, “Histogram of Loan Application Times.”

However, adding the control limits to a run chart of the data in time order shows that there are definitely three loans that have exceptional variation. An examination of the specifics of these three loans can reveal a source of exceptional variation that might be either removed from the process or minimized. Also, if the data for these three loans were not used in the calculation of the control limits, another loan application time would fall outside the revised limits. Further investigation of these specific loans can be beneficial in identifying sources of variation in the process. See the process behavior chart, “Process Behavior for Loan Application Times.”

Process Behavior: Predictable or Unpredictable

Why the fuss about predictable vs. unpredictable behavior in a process? As Walter Shewhart clearly pointed out in the early development of process behavior charts, the focus on process improvement is different for sources of variation that are unpredictable (exceptional or special causes) vs. those that are predictable (routine or common causes). Unpredictable variation is best identified using a process behavior chart and the sources of this variation will need root cause analysis for identification that leads to actions to eliminate or minimize these causes. Continued use of the process behavior chart will provide the verification as to whether or not the changes are being sustained. Predictable variation is inherent to a process and requires fundamental process change to reduce variation for successful improvement. If fundamental process changes are made prior to learning the true process behavior, those fundamental changes may not address causes of exceptional variation, which will continue to compromise and keep your process in an unpredictable state.

In particular, when exceptional variation can easily be seen using a process behavior chart, there is no need for designed experiments or regression analysis at this stage. Design of experiments is a powerhouse that can be used to study the routine variation in a process and determine which sources of variation contribute most to the CTQs.

For the project “On Time to Process a Loan Application,” the sources of variation for the exceptions can be investigated early on in the DMAIC methodology with the process behavior chart. After the sources of variation for these exceptions have been eliminated from the process, the remaining variation comes from the fundamental design of the process. This routine or common cause variation is best studied through design of experiments.

Another example comes from a Six Sigma project to improve the variation in the capacity of glass containers. Two characteristics of interest to the team are capacity and weight of the containers. Weight is determined in the process when a “gob” of glass is cut and goes to a mold. Capacity is determined in the process at the molding operation. In addition to the specific mold involved, the weight of the glass gob does have an impact on the capacity of the finished container. In this situation, capacity is a function of weight plus other process variables. The team obtained two sets of data from the production process for two different time periods and made histograms. The two sets of data had very different standard deviations and correlation coefficients. This information is summarized in the table, “Capacity of Glass Containers.”

The team made scatter diagrams and calculated the correlation coefficients as well as regression equations to determine why the standard deviation and correlation coefficient in data set 2 were much smaller than the corresponding values for data set 1.

At the suggestion of a coach, they made process behavior charts of both sets of data and discovered that data set 1 came from a period of time when the process was unpredictable while data set 2 came from a period of time when the process was predictable. The sources of variation that caused the unpredictable behavior could more easily be identified from the process behavior chart than from a scatter diagram or regression analysis.

Six Sigma is designed to provide a structured process improvement methodology. The benefits of using process behavior charts in the Measure and Analyze Phases of the DMAIC methodology are twofold. First, exceptional variation can be seen easily and the sources of this variation can be identified and removed from the process. Secondly, after the process has only routine variation, the techniques of design of experiments and regression can be used more effectively.

By gaining process insight early using process behavior charts, the Six Sigma methodology and its accompanying tools become an even more powerful framework for making effective, sustainable improvements.