Ford CSR Cpk/Ppk Requirements

Bev D

Heretical Statistician
Leader
Super Moderator
We have written about this Cpk/Ppk Ford thing before.
First the formulas are not the same* In the latest installment of definitions (a couple of decades old) Cpk uses the within subgroup standard deviation (hence why it is sometimes called short term). Ppk uses the standard deviation of all of the individual values. So Ford uses Ppk of 1.67 for the initial (short term) capability because in pre-production and early production there is not a lot of material and process variation. Which makes sense… Once a process is up and running and normal variation is now occurring they accept the ‘long term’ variation of 1.33.

*NOW for the confusion. In the very beginning There was no Ppk. There was no short term or long term. There was only Cpk and it used the standard deviation of all of the individual values. There also was no conflating it with a defect rate…but companies have to be different and put their own spin on things…so some companies hav e given up on the whole Cpk/Ppk thing and have reverted to the original definition of Cpk. (Which Ford still calls Ppk).

Of course the whole Cpk/Ppk thing is stupid and useless but what the heck if your Customer requires it then you gotat do it.
Just be sure you KNOW which formula for standard deviation is used…
 

MOester

Starting to get Involved
The worst thing we ever did was introduce the long term and short term into the definitions.

Cp/Cpk by definition is the process CAPABILITY (what it COULD do)
Pp/Ppk by definition is the process PERFORMANCE (what it actually did)

Long and short term have nothing to do with it.

The formulas behind it ALSO support the definitions of CAPABILITY and PERFORMANCE.
When calculating Pp/Ppk you are using the standard deviation of the entire data set.
When calculating Cp/Cpk you are using the pooled standard deviation. This is basically looking at the subgroups individually then averaging them all together. It eliminates the overall process drift and just focuses on the within subgroup variation.

If you consider a bunch of subgroups where the subgroup variation is small, but long term they are drifting all through the tolerance range, Cp/Cpk will be better than Pp/Ppk. Because the pooled standard deviation basically isolates out the change beteween subgroups and only considers the variation within subgroups. It performed poorly, given all the variation through the tolerance. (Actual performance, Pp/Ppk). But it is CAPABLE of better performance. If an operator looked at each subgroup and made adjustments the center the process, it would bring the subgroups in line and reduce variation overall - the swing. And standard deviation would approach pooled standard deviation and thus Pp/Ppk would approach Cp/Cpk. It would begin to perform to it's capability.

Long term/short term is best reserved for the length of the study. Did I do it over a shift or a week? If you research those terms on the interscreens, half the sites say Pp/Ppk is short term, and the other half say it's Cp/Cpk. Referring to the metrics as either long or short term adds to the confusion when the definitions they currently have are MORE than adequate.

Here they are again:

Cp/Cpk by definition is the process CAPABILITY (what it COULD do)
Pp/Ppk by definition is the process PERFORMANCE (what it actually did)
 

Bev D

Heretical Statistician
Leader
Super Moderator
Or better yet: stop the Cpk/Ppk madness. Plot your raw data in time series, look at the data dispersion vs the specifications. Is it within specifications? Is it stable? That’s how we did it for decades before the Cpk insanity in the eighties.

The manipulation of mathematical formulas is no substitute for thinking…
 

MOester

Starting to get Involved
Or better yet: stop the Cpk/Ppk madness. Plot your raw data in time series, look at the data dispersion vs the specifications. Is it within specifications? Is it stable? That’s how we did it for decades before the Cpk insanity in the eighties.

The manipulation of mathematical formulas is no substitute for thinking…
100%
We WANT a number and a hurdle because we don't have to think. There is no substitute for graphing the data - I constantly tell my reports this and to ask "Does what you are seeing make sense?"

Unfortunately, the madness of Cpk/Ppk has rooted deeply in my world (automotive).
 
Top Bottom