Supplier Cpk data for straightness measurement

01/15/2011:

I have Cpk data from a supplier and I want to compute a Cpk index. For example, one set of data represents a straightness measurement which is bounded by zero with an upper specification limit of 1.016. The data seems to be centered within the spec limits and is more than three sigma away from the upper specification. Why does my analysis indicate the Cpk is only 1.0?

Calvin E., Quality Engineer

There are a few things going on here. First and most important, you only have 30 data points, so it can be difficult to fit a statistically relevant distribution. Furthermore, you really should never fit a distribution until you know the process is in control. An EWMA chart (lambda = 0.4) can be used to show that in this case, the process is out of control, so by definition a single distribution is not appropriate. Since the process is out of control, then the process is incapable.

A few other things I might point out:

1. For a straightness measurement, there is no lower specification limit, so that field should be left blank.

2. You should remove out of control data from the calculations of the control limits. This is standard practice. The points should be visible on the chart, but excluded from the analysis.

3. Your comment that the data is more than three sigma from the specification is only pertinent to a normal distribution. Non-normal distributions may require more sigma units on one side or the other of the median to ensure capability.

Learn more about the SPC principles and tools for process improvement in Statistical Process Control Demystified (2011, McGraw-Hill) by Paul Keller, in his online SPC Concepts short course (only $39), or his online SPC certification course ($350) or online Green Belt certification course ($499).