By Jennifer Turvey,
Staying In Control
What if you get to the end of an expensive Six Sigma project, and you discover the improvements have not significantly impacted the problem? Believe it or not this does happen, and it’s likely because of an outside influence on the process that was not detected during the project.
Using Statistical Process Control (SPC), which employs a combination of statistical and graphical methods to give you a look at what’s happening inside the process can help ensure you never have that experience.
Its hallmark visual is the control chart, which shows the variation present in the process.
When all the variation belongs to the process it’s said to be in control. When variation external to the process is influencing it, that means the process is out of control. External variation must be eliminated for an improvement project to be worthwhile, which is the message of the scenario at the beginning of this article.
What is SPC?
Statistical Process Control refers to a set of graphical and statistical tools which collect, display, and analyze variation to assess the performance of a process and to identify issues requiring attention.
SPC is part of the Six Sigma toolkit and supports the Six Sigma project manager in the effective use of statistics to manage improvement projects across all types of processes in all types of organizations.
Some ways SPC can be used in a Six Sigma project:
- To show how well a process is performing
- To compare process of performance before and after an improvement
- To indicate the presence of issues requiring investigation and pinpoint their location in the process so they can be removed
SPC’s graphical elements are the run chart and the control chart. They visually summarize what’s happening during a given time period by analyzing data collected on a key process metric.
To be analyzed with SPC, data must have a structure which repeats and be expressed in time increments.
- Sales by month
- Customer complaints per week
- Spoiled food by day
- Wait times by hour
- Cases produced per minute
The control chart is a run chart that visualizes statistically generated observations about the process. Control charts are much more useful and more frequently used. There are many types, and the one you use depends on the type of data you are analyzing.
Features of the Control Chart:
- Time increments (e.g. days) are on the horizontal axis, the values of the data points (e.g. sales totals) are on the vertical axis
- Values are plotted in time order from left to right
- A centerline represents the average (mean) of all data points in the chart
- The lines above and below the average are the control limits — upper control limit (UCL) and lower control limit (LCL)
The control limits are calculated from the standard deviation and average of the data in the chart.
Being Out of Control
There are 5 common categories of variation to watch for in your control charts.
Processes with only typical variation are considered in a state of statistical control, or in-control or stable. The presence of non-typical variation signals something that requires investigation and removal. Here’s an example of non-typical variation and its significance in process analysis for a Six Sigma project:
Say you are the project manager for Six Sigma project focused on improving the customer service provided by your organization’s call center. During the Analyze phase of the project, you create a control chart of the key project metric, the number of calls answered per day. You discover patterns in your control chart, and spend some time with call center mangers investigating calls received during the times corresponding to the patterns. The managers discover one of the following: an untrained team member, a software bug in a call center’s system, or a call agent who regularly takes a 4-hour nap during his shift. At this point, there’s an opportunity to remove the non-typical variation: the guy gets his training, or that bug is fixed, or the agent’s sleepy behavior is addressed.
You can’t dependably analyze and improve a process that is not in control. Why? You risk solving the wrong problem. Here’s an example:
Now, instead imagine that you didn’t do a control chart and remained unaware that there’s a system bug AND a sleeping call center agent. The project moves forward full steam ahead and toward the control phase. You and the project team are blissfully unaware of these unaddressed issues. You decide to run a pilot of the newly improved process.
Worst case scenario? Data collected in the pilot appears relatively non responsive to those improvements the team. Why? Because the project wasn’t addressing all the issues. It’s even possible that the project focused on issues of relative unimportance compared to that software bug. This is the scenario the article opened with.
This an example of why the effectiveness of improvement efforts cannot be guaranteed when non-typical variation is present. In fact, experts will tell you that attempting to improve a project that is not in control is a fool’s errand.
Fortunately a control chart can show where to look for the issues: the point in the process that correspond to the patterns in the chart.
Statistical Process Control makes it possible to observe two types of variation: typical and non-typical. A control chart can be used to visualize the variation in a process and detect non-typical variation.
When non-typical variation is present the process is considered not in control or not stable.
Attempting to statistically analyze and improve a process that’s not in control puts an improvement project at risk. For this reason, non-typical variation should be investigated and removed.
For projects with repetitive data, using a control chart to identify, pinpoint and remove non-typical variation during the Control phase of a project makes a lot of sense. Doing so early in the project as well makes even more sense.
“I took the one less traveled and that has made all the difference.” – Robert Frost
About Jennifer Turvey
Jennifer Turvey is
Read What is Lean?
click here to create your own account.