Ppk vs Cpk: Big Differences - What conclusion can I reach about my process?

[ricardo.qre]

After using the data from our process we got:
cpk=2.42
ppk=0.76

we used a I-MR chart to get the data.

I understand ppk should be used as long term and cpk short term, since the differences on how the sigma is calculated, but

What conclusion can I reach about my process?
I mean, by having ppk vs cpk big differences is there any typical conclusion we can obtaine from this?

and what about if cpk = ppk (or very close)

waht does that mean ?

 

[Miner]

Cp and Cpk are calculated using a standard deviation that reflects short term variation that ideally represents only common cause variation. For sub-grouped data, it is calculated using only the variation within subgroups. For I-MR data it is calculated using the variation between successive measurements.

Pp and Ppk are calculated using a standard deviation that reflects long term variation that represents the inclusion of any special cause variation that might exist in the data. This is calculated using an overall variation rather than the within sub-group or between successive data.

See this Minitab Technical Support Document for the actual equations for these standard deviations, and this Technical Support Document for the calculation of Pp and Ppk.

Based on a Cpk of 2.42 and a Ppk of 0.76, you process has little common cause variation, but a lot of special cause variation. If you were to eliminate the special cause variation, your process would be highly capable, but under the current conditions the process is incapable.

 

[Bev D]

You can draw even better conclusions if you plot your data in time series and look at it.

You probably have a lot of small shifts or slow trends/cycles.

If you post your data we can provide better advice.
And tell us something about the process

 

[bobdoering]

Also, having some idea of the process is important to answer this question. You can answer the question "what is the difference between Cpk and Ppk?" without that, but you cannot answer why they are different. It can go to a larger question, such as: are these indices correct for your process variation distribution? That is why time series data taken with an adequate measurement system is critical.