Blog

Special Cause Variation

I belong to several groups on the Linked In web site. I find some of the discussions both interesting a informative. Not long ago, on one of the sites a member named Qaisar posted the following topic for discussion: "How to address special cause variation... Manufacturing produced 100 doors. 60 doors were rejected because they were outside the specification limit. Reasons known were inaccurate dimension, excessive knots, cracks, etc. Thanks for your comments."

I posted the following response to Qaisar's query:

Qaisar, were the 60 rejected doors due to special cause variation, or merely out of spec? If it was a significantly high reject rate in comparison to past rejects, this would indicate the presence of a special cause. The appropriate corrective strategy would be to conduct systematic root cause analysis to find, remove and prevent the reoccurrence of the special cause; i.e.,

1. Name the problem
2. Describe the problem (construct detailed problem specification)
3. List possible causes
4. Test possible causes (against the facts noted in the problem specification)
5. Identify most likely cause
6. Verify true cause (singular -- because you're chasing a single, smoking gun, special cause)

It is possible, however, that the singular special cause is the result of an interaction between two contributing causes. For example, a new material might work with some of your doors but not be able to handle current pressure settings for another door design. The culprit would not be the material alone, but the interaction between the new material and pressure.

On the other hand, if your 60 percent reject rate is not significantly higher than past results, there is no special cause. Rather, the source of the defective doors would be multiple common causes of variation -- and interactions between and among those common causes of variation -- from within the same process that produced the 40 good doors. Trying to identify a special cause would be a costly mistake, wasting time and defeating your purpose, because there would be no single, smoking gun, special cause.

Instead, you should deploy a team to study the current process and develop recommendations for changes you can make to the process to reduce future reject rates. You must develop recommendations (plural) and make changes (plural) because you're up against the effects of common causes (plural) of variation. In such a case, Design of Experiments (DOE) or some other technique that can handle multiple variables would be more appropriate than root cause analysis.