Cpk vs Ppk - Part 2

In fact I m not so interested in Cpk or Ppk value because I feel the number can be very misleading if we don’t really understand how the data behave. Understand data behavior is more important because we may end up drawing a wrong conclusion if solely make decision based on Cpk or Ppk value as what my managers is doing.

Usually, when I receive a set of 10Lot data, I will straight dump the data into Minitab and lot the I-MR chart as shown below;





I like to analyze my data this way because it provides me a quick understanding of data behavior either within or between the lots. From the data, I can see that the overall process is stable and in control in general. Any abnormality also can easily identify.

After assessing the I-MR chart, only I check for the process capability. I will check on the Cpk and Ppk value and the distribution of the data, again, Cpk and Ppk is not the only criteria use for measuring the capability of a process.

Look at the chart below, Cpk and Ppk value is so low, can say that the process is not capable for mass production ? Yes, the process is not capable for mass production, if I make my conclusion solely based on the number. But if look at the I-MR chart and the overall data distribution, it shows that the process is stable enough for production release. What I m more particular is mean of the data distribution must be on target.



The low Cpk or Ppk in this case mainly because by the process variation, and don’t forget we are now considering the overall process variation. Of course we expect a bigger variation compare to the variation within individual sub group. This is because there are many variations have to factors in when we are in the actual production mode, such as machine to machine, people to people and batch to batch variation.

Since I mentioned the main contributor to the low Cpk and Ppk value is the process variation, what happen if the process shift, then the process will out of control!

Ok, in fact I m not so worry about the process variation in 10 Lot monitoring, this is because the data were collected from several machine, several set up and several shift over quite along period. The spread of the data shows above is good enough to represent the future production condition. In other word, variation of the long term process should not be different too much from the 10 lots data, provided the machine condition, and set up is well controlled.

Last update : 06/06/08