Overall Cpk for Different Parts to Same Customer

[PigPiG]

Hi all,

I am working in a sub-con company. Hence our company will produces different parts to same customer. And different spec limits will be used for each part. Now we have a problem is customer request us to send an Overall Cpk report to them in monthly basis. How to get the overall cpk from this different parts since they having different spec limits?
Anyone can help on it? i had checked some forum, some of them, their customer will request them to provide an average Cpks as overall cpk. Is it valid statistically?

 

[milind dhakad]

Re: Overall Cpk for different parts

Dear Friend
as per my knowedge, if you have PPM data for the all component then you can calulate the 6 level and overall Cp & CPk

 

[bobdoering]

i had checked some forum, some of them, their customer will request them to provide an average Cpks as overall cpk. Is it valid statistically?

You can find a grain of validity in the statistics (as in you can calculate the data in a variety of pathetic ways, such as averaging all of your individual Cpks), but it lacks any real value except for those who are inordinately lazy in their evaluation of vendor performance.

Cpks have very little real analytical value - much less value than some organizations place on them. Often they are even used when they hold no statistical relevance, just to feed a customer's thirst for rubber stamping. They can even lead to overadjusting. So, the grand overall value of an index of little value to begin with is busywork at best.

 

[Darius]

check this thread

http://elsmar.com/Forums/showthread.php?t=38557&#4

also found some time ago, in Bologna University site an article from Michele Scagliarini and Raffaele Vermiglio, it looks good (at least as a lay man, and they can share an Excel file with the calculations): Computing Multivariate Process Capability Indices With Microsoft Excel

There is also another posibility, not taken the calculation as a total Cpk but as way to evaluate your process, the use of Composite Indicators is recommended by the European Union for evaluation of the Economic wellfare. It's like a weighed average, where every variable has it's weight according to the importance of it to the global index.

 

[Statistical Steven]

Hi all,

I am working in a sub-con company. Hence our company will produces different parts to same customer. And different spec limits will be used for each part. Now we have a problem is customer request us to send an Overall Cpk report to them in monthly basis. How to get the overall cpk from this different parts since they having different spec limits?
Anyone can help on it? i had checked some forum, some of them, their customer will request them to provide an average Cpks as overall cpk. Is it valid statistically?

Never average Cpk values. Remember the relationship of 3*Cpk is a standard normal z-score when a process is perfectly centered. If you want a composite Cpk, you can convert each Cpk to a PPM and average, but then again as has been mentioned, Cpk only talks about how close the mean is to one of the specifications. I am attaching an Excel spreadsheet that gives what I believe is the best estimate for an overall Cpk.

 

[bobdoering]

Before going to deep over the edge, you may want to ask what algorithm your customer suggests to develop "Total Cpk". If they look at it simplistically (especially if they do average, even though as Steve suggests it is not mathematically correct), rather than as a multivariate function, then you know it is simply a rubber stamp application and not to worry too much about its statistical validity. You will have evidence that they are not concerned about that. You can even use Steve' averages of estimates - all they will look for is a "magic number", the one number that describes all of your processes! I still say only a lottery winner would place that much importance in one number.

 

[Bev D]

OMG. what has become of us? have we sunk so low in our laziness? this is just pathetic. I'm sorry but your Customer doesn't care about Quality nor do they understand it. This is nothing but voodoo.

Steven is correct: averaging your Cpks is just a fake number (oh wait they are all fake numbers). The mathematically correct way is to get your overall defect rate and back calculate the Cpk value thru the Z table. It's still but maybe tehn you can have a productive converstaion with your Customer or your organization on hwoto really understand your capability and improve your processes...