Which of the following indicators is important for Operations - Ppk, Cpk, or Cpm?

R

refractoryman

Which on is "most important" ... Ppk, Cpk, or Cpm? In other words, which one is the most practical to use for an operation that has yet to use any of these? Or are all three equally important? Please advise.

Thanks,
CUFF
 

antoine.dias

Quite Involved in Discussions
I would start with Cpk (based on spotchecks) and use the data for steering the process.
After a while - considerable number of spotchecks - you can calculate Ppk (based on all measured values). Actions should be taken to get both "in control"
Then I would start using Cpm to get the process more and more centered thus improving the process more.

Hope this helps
 
D

Darius

Which on is "most important" ... Ppk, Cpk, or Cpm? In other words, which one is the most practical to use for an operation that has yet to use any of these? Or are all three equally important? Please advise.

Thanks,
CUFF

You can have a long discusion about this question, :2cents:, any indicator is a simplification of the real thing, there are some things that you can't see on any of them, as trends, outliers, cyclical trends, etc.

IMHO, any of them could be a good thing to follow, if there is no outliers on the data. Most of the people could say that Cpk is better than ppk because, ppk is for non-stable process and being non-stable, you can't trust on it to predict the process capability (as Don Wheeler, Advanced topics on SPC), even they say that you shouldn't calculate it if the stability has not been proved.

Some say that Cpmk, is good if your process is not centered, but the good point is that it take in account the centering to the target, with cpk or ppk you can have the same or better index if you are farther from the target but with less variation, and that is wrong, cpmk is more robust to this kind of event. The problem with Cpmk/Cpm is that just a few use it.

The question of ppk or cpk is ... when is stable?
 

Bev D

Heretical Statistician
Leader
Super Moderator
If your customer does not require you to report these - just plot your data. these indices are far more trouble than they are worth.

If you are interested in capability, you should be plotting your data in time series against the specification limits. Look up multi-vari charts. this is the best graphical format for capability and the best place to start understanding and visualizing your process variation without the confusion of summarized indices. (I do use these before implementing SPC)
 
Last edited:

RoxaneB

Change Agent and Data Storyteller
Super Moderator
Which on is "most important" ... Ppk, Cpk, or Cpm? In other words, which one is the most practical to use for an operation that has yet to use any of these? Or are all three equally important? Please advise.

Thanks,
CUFF

Most important to whom? Who is the data for and what is the purpose for collecting it? Is there value to calculating these values for any particular Stakeholder?

The answers to those questions might help your organization determine what is "most important".
 
R

refractoryman

Steve:

Please expound on "none of the above."

Thanks
 
Last edited by a moderator:

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
One of the biggest problems with Cpk, etc. is it is one number used to represent a very dynamic issue - process capability or performance. The only person that is impressed with the power of one number is a lottery winner.

Also, there are strict limitations as to when these values are applicable, such as the variable must be bilateral with target in the center, the process distribution must be normal (and not normal because poor gaging and measurement are masking the true process distribution). The target in the center must be supported by an accurate specification with a narrow loss function. Otherwise, chasing the center is a waste of resources.

It is a requirement often rubber stamped by people looking for a cheesy 10,000 foot feel for the process. I hold very little decision making power, unless the process is horrible (in which case you likely don't need this index to tell you that).
 
R

refractoryman

bobdoering:

Please explain "bilateral with the target in the center" and "specification with a narrow loss function" for me please

I want to be absolutely sure that I am understanding all of this in detail. I appreciate your patience with me.

 

TPMB4

Quite Involved in Discussions
I'm also curious what cheesy 10,000 ft feel means!

You Americans with your quaint sayings!!!!;)
 
Top Bottom