I had the opportunity to attend a training programme conducted by one of our customer where the lecturer / facilitator told us something that was in contradiction to what I learnt earlier.

They said that Cp,Cpk is related to short term capability and Pp,Ppk is related to long term capability.What do the forum users feel about it?

You'll typically find, in Automotive:

Cpk= Long term---due to estimated std dev
Ppk= Short term---due to calculated std dev

And, MHO, if the long-term process is stable, this is the way to go. If not a stable process, Cpk is always worthless.

In Software, I have typically seen just the opposite.

I don't see, whether or not it's car parts or CD's, where an estimated sigma (Cpk) is good for any short term capability.

Though, very interested in other opinions as always.

Hope this helps ! Have a day ! :smokin:

I agree with noboxwine.

Stability is a prerequisite for Process Capability. This can be established using a control chart. A long term study would imply that the process has had a chance to run thorugh all possible cycles and therefore all the common and special causes of variation have been covered. In such cases Cpk is calculated using sigma=Rbar/d2.

Process Performance Ppk involves short span runs of may be 100 consecutive parts. Some of the special causes of variation may not have ocurred at all.

Here are two earlier threads on the same topic:
http://Elsmar.com/Forums/showthread.php?t=1994
http://Elsmar.com/Forums/showthread.php?t=4048

Now I'm confused. I don't have a copy of the AIAG SPC manual here, so I pulled out Ford's CSRs because I remember it being talked about in there. But they contradict themselves! In 4.37 it says...

The choice of the capability index used for initial process studies - Cpk (predictive), or Ppk (historical) – shall be based solely on the nature of the process data collected
But in Table B it says...

After process stability has been demonstrated and capability has been calculated, the most recent point on the control chart and the historical process capability indices (Cpk/Cp) may be used to determine appropriate actions.

It sounds like Ford is confused also. My understanding has always been as stated above - that Cpk was historical and Ppk was short-term.

So the answer is???

So the answer is???ASQ's Body of Knowledge for the Quality Manager certification test describes
Cpk as long-term, stable (Cpk > 1.3 = capable)
and
Ppk [Potential Process Capability] as short-term, stability undetermined (Ppk > 1.67 = capable) The difference in ratio of being acceptable to be called "capable" is just a fudge factor.

The difference in what is capable ratio is that because the run is short, Ppk does not include the normal variability that will be seen in a full production run.

Now ask me, "What do I really believe?"

I really believe that reliance on such ratios for short runs (Ppk) is really not as informative as pure control charts.

There are links to interesting articles on the topic at
http://www.qualityadvisor.com/library/capability/capability-menu.htm

I've stumbled across this debate on this forum, which is a coincidence because I was having the same arguement with one of my Quality Engineers.

He also states that his training told him that Cp/Cpk related to Long term, and Pp/Ppk related to short term.

I am a certified Six Sigma Blackbelt, and my training pushed me in the other direction. After a heated debate, I reviewed the training files from the Six Sigma Institute which states the following:

When evaluating process capability…
- A short term study is conducted to see how good a process can be. Data is collected over a short period of time during which the process is influenced only by random causes.Cp, Cpk
- A long term study is conducted to see the process’ actual long-term performance. Data is collected over a long enough period of time such that the process is influenced by both random and non-random causes.Pp, Ppk
- The Process Sigma Level is the z value corresponding to a short term study (centered distribution, when using variable data).
- The Short Term Process Sigma Level can be estimated from a long term study by the adjustment: zST = zLT + 1.5.
- The 1.5 Sigma adjustment accounts for the shift and drift of the short term distribution over time.
- When variable data is used, the capability indices Pp and Ppk are computed:
Cp indicates the process’ potential capability if it were perfectly centered between specification limits.
Cpk takes into account off-centering.
- For a Six Sigma process, Cp = 2.0 and Cpk = 1.5
- When attribute data is used in a long term study, the Long Term Process sigma Level is the z value corresponding to the proportion defective or DPO. The Short Term Process Sigma Level (reported) = zST = zLT + 1.5.


I Hope this helps!

Smithy

Smithy,
Thanks and Welcome to the Cove.

Just make sure you understand the sources of variability in the surrogate process.

If you have a capability index of one and only one person looks at it one time then one report will probably do until one addresses management commitment.

When evaluating process capability…
- A short term study is conducted to see how good a process can be. Data is collected over a short period of time during which the process is influenced only by random causes.Cp, Cpk
- For a Six Sigma process, Cp = 2.0 and Cpk = 1.5

I Hope this helps!

Smithy
Welcome, Smithy.

I thought a Cpk of > 2 was required for a "six sigma" process. Can you double-check?

Here is another opinion on the debate, FWIW, from an article I have by Dr. Mehernosh Kapadia:
____________
In 1991, ASQ / AIAG task force published the "Statistical Process Control" reference manual, which presented the calculations for capability indices ( Cp, Cpk ) as well as process performance indices ( Pp, Ppk ).

The difference between the two indices is the way the process standard deviation ( s ) is calculated.

Cpk uses s which is estimated using ( R-Bar / d2 ) or ( S-Bar / C2 ) .

Ppk uses the calculated standard deviation from individual data where s is calculated by the formula :


So the next question is which metric is best to report Cpk or Ppk ?
In other words, which standard deviation to use - estimated or calculated ?
Although both indices show similar information, they have slightly different uses.

Ppk attempts to answer the question "does my current production sample meet specification ?" Process performance indices should only be used when statistical control cannot be evaluated.

On the other hand, Cpk attempts to answer the question "does my process in the long run meet specification?" Process capability evaluation can only be done after the process is brought into statistical control. The reason is simple: Cpk is a prediction, and one can only predict something that is stable.
____________
Dr. Mehernosh Kapadia is General Manager - Quality Engineering, Supply Chain Management Business Group at Tata AutoComp Systems Limited.

He has a Master's Degree in Mechanical Engineering from Penn State University, USA and MBA from NTU, USA. He is a Six-Sigma Master Black Belt and an ASQ certified CQE & CRE. He obtained his Doctorate from The University of Bombay

Welcome, Smithy.

I thought a Cpk of > 2 was required for a "six sigma" process. Can you double-check?


Mike,
I have doubled checked the information I quoted, and it seems to be accurate, the document I have clearly states that for a Six Sigma Process, Cpk = 1.5

Whilst doing the training, I wrote the following notes:

If Cpk = 2 and process is centered then we have plus and minus 6 sigmas (std dev) on each side of the mean. If we look at one side from the mean then we could see that the tail would take up half or 3 sigmas (std dev) with 3 sigmas open to the spec. Cpk is focus from mean to spec - 6 std dev / 3 = 2
Cpk = (spec - mean)/3s or Zmin/3

Thus for a Six Sigma process - long term - Cpk = (6 - 1.5shift) / 3 = 1.5 Cpk
Long term is always in terms of the shift - taking in all the factors of variation.

Smithy.

Thus for a Six Sigma process - long term - Cpk = (6 - 1.5shift) / 3 = 1.5 Cpk
Long term is always in terms of the shift - taking in all the factors of variation.

Welcome to the Cove Smithy. Just a quick comment on this if I may.
Although I certainly recognize that theoretically there is a process shift of 1.5 sigma over time and that it should be considered in any process fallout analysis, I don't think it is appropriate to a Cpk calculation. The method of calculation is clearly defined and accepted. Under this accepted method, Mike is clearly correct - 6 sigma = 2 Cpk (or as you said 6 std dev / 3 = 2).
I assure you, I understand where you are coming from but I would hate to see everybody running out to buy new software because the calculations didn't work out to 1.5 Cpk.

Dave

Welcome QeSmithy. :bigwave:

IMHO :frust:

Juran's QC Handbook (4th & 5th ed.) calls Cpk a "performance index". The AIAG SPC Manual (1995) calls Ppk a "performance index", and Cpk a "capability index for a stable process". So it gets a little muddied after a while (and let me say, not all that important).

Cp and Cpk are at least as old as Ishikawa's 1982 "Guide to Quality Control" and are simply the inverse of capability ratios. They applied to short and long term measures of capability. Ppk was first used in the context of new product submissions (ISIR/PPAP) and still exists on some company's warrant forms for their initial (and obviously short term) estimates of process potential. Ford Motor company was one of the first. Ford did not invent Cpk & Ppk, but they DID champion their use - with the same understanding that Cpk = long term and Ppk = short term. Ppk was used also in other earlier PPAPs for Short Term estimates of capability. However, somehow, in the last decade the roles have reversed (especially by the 6 sigma people). This is partly due to the AIAG's SPC Manual defining Ppk as a nebulous "performance index". Does that definition imply performance over time? Who knows. Somehow the issue got confused.

Interestingly, the 1999 printing of the AIAG PPAP Manual speaks of using Cpk when historical (long term) data is available - enough to divide into subgroups, and using Ppk when it is not. In any event, the general calculations are the same (sure, the estimate or calculation for standard deviation varies, to which I say, blah blah blah :bonk: ), and since it is all really a matter of semantics, if everyone agrees (especially you and your customer) then that is all that matters.

I stand by the older Cpk = long term, Ppk = short term understanding. Cpk is a more valid index for on-going process capability measurement.

PS: You'll find tons of debate on this if you do a Google search on "cpk vs ppk".

Hi all,

Just to share what i understand about Cpk and ppk (correct me if my understanding is not right.

The difference between Cpk and Ppk lies on the estimate of Sixgma.
Cpk used Rbar/d2 whereas Ppk use the the overall std dev
[Sqrt(x-xbar)2/n-1].

This mean that Cpk is only meaningful only if the process is stable because it only consider within sample variation as your process variation. As long as your process is stable, it is reasonable to say that your between sample variation (long term) will be approximately the same as your within sample variation (short term). That's why some say "Cpk" as "long term" estimate of a process capability .

Imagine, if cpk is used for a unstable proess (out of control process ), then the between samples variation (due to out of control or special cause) is not being considered during your estimate of process capability.
(As long as you had a very small within sample varaition and if these variation is stable, you still get a high Cpk even though your "Xbar" control chart is out of control)

Ppk consider the overall variation. Normally, It is used as a actual performance index (so call short term). If your process is not stable, you will get a low ppk value as compare to cpk value.

In summry, if a process is stable with only common cause variation only, then ppk and cpk should not be quite far apart.

Hope this help.

There are links to interesting articles on the topic at
http://www.qualityadvisor.com
Nice link! Good stuff! Well organized! Thanks!

Cpk takes into account off-centering.
See: http://Elsmar.com/Cp_vs_Cpk.html