Cpk for small sample sizes?

Question

Hi everyone,

I have a JMP user making coils, which are then shipped to different customers.  They measure the thickness of each coil, and then take out a bunch to complete an order.  The customer receives the Cpk for their particular bunch of coils.  The issue is that there may be very few coils per order, rendering the Cpk essentially meaningless.  Does it make more sense to send the customers the Cpk calculated from all coils made that week, or is there a more appropriate statistic they should use?  Any help with this would be appreciated.

HadleyMyers · Answer

I definitely appreciate all the interest here.  I ended up going through the Distributions menu to show other the different ways we can visualize the variation and range of a batch of products.  They seemed to like Box Plots and I think that's the way they're going to go moving forward.

MikeD_Anderson · Answer

@Phil_Kay , @JerryFish - I wonder if (just speculation here) at some point the customer asked for the CpK of the line while their parts were being made.  Maybe over time (in the interest of being more "efficient" :face_with_rolling_eyes:) they drifted towards just calculating the CpK of the customers parts alone... ¯\_(ツ)_/¯.

Another way to read Hadley's description is that they are cherry-picking parts from the line that meet a certain customer's specification (i.e. some customers pay for more consistent parts, while others can handle a little more variation for a cheaper component.) - again, a misuse of the CpK, but who knows.

It is a weird situation.

M

Phil_Kay · Answer

Good point, @JerryFish .

What they are doing is just very weird and I don't understand why they are doing it. I've never heard of a manufacturer stating the CpK for a batch of items to their customer. I can understand why a customer might want to know the CpK for the process - to have some assurance that the supplier has control over quality.