StATS: Unusual advice about control charts (December 18, 2006)

Someone sent me some recommended guidance on how to use a control chart and it included the following quote:

Do not correct the process if the out-of-control values can be shown to be due to chance failure when process is actually in control (special cause variation).

I'm probably taking this quote out of context, but it is a rather unusual claim.

The phrase "correct the process" could mean start a fresh chart with a new set of control limits. When should you start a fresh control chart? It depends, but I discourage a lot of tinkering. If there is a wholesale revision in a work process, definitely start with a fresh set of control limits. If you are making a minor change, perhaps it would be simpler to keep the same control limits, at least until it becomes painfully obvious that the old limits are no longer relevant.

I discourage tinkering with control limits because your time is better spent investigating issues with the work process itself. Talking to the people involved with the work process is likely to lead to more improvements to the system than fiddling with the control limits.

The phrase "chance failure" is also interesting. It could mean a point that is out of control, but where no assignable cause has been found. This can happen, because a control chart will tell you WHEN a process is out of control, but it will not tell you WHY it is out of control. If you cannot find an assignable cause, then you obviously cannot take any corrective action. But ask yourself first if you have spent enough time working on the investigation and if your co-workers are being honest with you. Many work processes have negative incentives for honesty. This encourages people to hide the truth.

Finally, the phrase "chance failure" could mean a false positive signal. This is a point that is out of control because control charts are not perfect and sometimes they produce false positive signals. This happens rarely, but it does happen. But how will the average person know what a point that is out of control represents a false positive signal? Well, one way is to be so smart and so perceptive that  you can recognize better than a control chart can what is going on with your work process. There are few of us who are that intelligent. Another way you might recognize that a signal is a false positive signal is if you discover that there is no assignable cause. But I would always worry that it was a true positive and that I just didn't investigate thoroughly enough.

There's a great website, www.skymark.com, that offers advice about how to react to signals in a control chart. I have quoted some of this advice at:

• Stats: What is a special cause of variation? and
• Stats: What is a common cause of variation?

This page was written by Steve Simon while working at Children's Mercy Hospital. Although I do not hold the copyright for this material, I am reproducing it here as a service, as it is no longer available on the Children's Mercy Hospital website. Need more information? I have a page with general help resources. You can also browse for pages similar to this one at Category: Control charts.