RE: REQUEST: Histograms and process capability

[Rip Stauffer <rstauffer@bluefirepartners.com>]

Although I agree with Steve Prevette that Cp and Cpk without a stable
process are useless, I have found the indices (when based on a stable
process) to be useful in convincing management that there is a danger in
complacency. If developed properly, they have a bit more theoretical
predictive power than a histogram. A manager looking at a histogram of a
stable process in which there has been nothing produced outside the specs to
date may assume that the process "is not broke, therefore doesn't need
fixin'." To demonstrate the idea that the process still has the potential to
produce defective material sometimes helps convince them that we have to
bring the average back to the nominal and reduce the variation.

The Taguchi Loss Function is similar; calculation of any particular loss
function for any particular set of data is a snapshot and an estimate.
However, if you can find a clear loss function, you can sometimes use it to
convince managers to pursue improvement of processes they might otherwise
ignore. We were pretty successful with it in the Navy, using an example of
an enemy tank in a road with Marines hidden in the bushes on both sides,
calling for Naval Gunfire or artillery support. I'll be glad to elaborate if
anyone's interested.

Best regards,

Rip