I guess the same should hold for the customers that require Cpk on characteristics that are not applicable. There is a lot of training needs to go around....
Me: I can show you a run chart that indicates that the measurements are well-centered around the nominal, but Cpk wouldn't make sense in this instance because we're not working with a normal distribution.
Him: What are you doing to correct that?
Me: Correct what?
Him: The fact that the distribution is out of control.
Me: It's not out of control; it's just not a normal distribution.
Him (peeved): Well what are you doing to make it normal?
It went downhill from there, if you can imagine such a thing. Sometimes you just have to give the customer what he asks for, good and hard.
My guess is that at least 75% of all Cpk results are functionally meaningless for one reason or another, the two main ones being non-normal distributions and no evidence of control when the distribution is normal.