In this third and final part of my Don’t Adjust series, I admit to you that there are times when you should adjust.  But you need to be able to recognize when to adjust, what to adjust, and why adjusting is the right thing to do.

The first thing to determine is if you are dealing with a common cause or a special cause.  Remember, you can’t “fix” a common cause with a corrective action.  It is a feature of the system that requires an improvement effort. Generally speaking, you need some data to determine whether this is a common or special cause.

If it is a common cause, next ask if the process is stable.  If it isn’t, don’t adjust, stabilize first.  Capability studies require a stable system.  Once stable, collect more data and see if you still need to adjust.  If the process now performs sufficiently, don’t adjust.  If performance is not sufficient, adjust for improvement.  Although you might be able to reduce variation in a process with an improvement action, you will not be able to remove all variation. 

A special cause requires an adjustment.  Remove a special cause through a problem-solving approach where a root cause is identified.  It is extremely important to make sure that you are really chasing a special cause and not a common cause.  Then use the problem-solving or corrective action approach of your choosing.

The takeaway (and the irony) here is that if you adjust a stable system to chase a specification YOU have now become the special cause!  You’ve made it worse, and the output will continue to drift around as you chase the specs.  I see Deming’s Rule 2 (see Part 1), “adjust a bit in the opposite direction”, applied often. If you use control charts and are adjusting because you are no longer within control limits due to thermal issues, tool wear, etc. that is one thing.  But if you are “tweaking” constantly after each data point, you are fighting a losing battle and will never really understand the capability of your process. I encourage you to understand and accept the variation.  Adjust a stable process (with common cause variation) only when you need to improve it.  Adjust to deal with special causes (Murphy strikes!) to correct a problem.  Yeah, it can get a little bit confusing at times, which is why my default position is don’t adjust.

Don’t adjust (unless really needed),