Learn · updated 2026-10-28
Cpk vs Ppk: Which One Does Your Customer Mean?
Cpk and Ppk are both the distance from the process mean to the nearer spec limit in units of three standard deviations: Cpk uses the short-term (within) σ and shows what the process can do, Ppk uses the overall σ of all the data and shows what it actually delivered.
A large, lasting gap between the two usually means the process moves around between readings, so the gap is a useful diagnosis, as long as there are enough readings to trust it. Below: one data set worked both ways, what the gap tells you, and which number a customer usually means.
Cpkmin(USL − x̄, x̄ − LSL) ÷ (3 × σ within)Ppkmin(USL − x̄, x̄ − LSL) ÷ (3 × σ overall)σ withinMR̄ ÷ 1.128 (or R̄ ÷ d2 for subgroups)σ overallsample standard deviation s of all the data
Worked Example: The Same Ten Readings, Both Ways
The NIST flow-rate data used in the I-MR chart guide, with spec limits 45 to 57. The mean is 50.81, so the nearer limit is the LSL: 50.81 − 45 = 5.81.
| Short term (within) | Long term (overall) | |
|---|---|---|
| σ | MR̄ ÷ 1.128 = 1.8778 ÷ 1.128 = 1.665 | s = 2.034 |
| Spread index | Cp = 12 ÷ (6 × 1.665) = 1.20 | Pp = 12 ÷ (6 × 2.034) = 0.98 |
| Centered index | Cpk = 5.81 ÷ (3 × 1.665) = 1.16 | Ppk = 5.81 ÷ (3 × 2.034) = 0.95 |
Same data, same spec, same formula: Cpk says the process could run at 1.16, Ppk says the parts it actually made ran at 0.95. Here the gap hints at the drift you can see on the chart, but with only 10 readings it is not proven (all ten are inside the control limits); collect more data before acting on it.
What the Gap Tells You
- Ppk about equal to Cpk: the process is steady; short-term and long-term variation are the same. Improving capability means reducing the common-cause spread.
- Ppk well below Cpk (as here): the process shifts or drifts between readings, between shifts or between lots. Find what moves the mean before you try to tighten the spread; a run chart or the Nelson rules show when it moved.
- Ppk above Cpk: usually a small sample, or readings that zig-zag (each one pulled away from the last), which inflates the moving range. With fewer than about 30 parts both numbers are rough; see how many parts a Cpk needs.
Which One Does Your Customer Mean?
Ask, because the same target means different things. In automotive PPAP (AIAG, 4th edition), the initial process study is judged on Ppk: above 1.67 meets the criteria, 1.33 to 1.67 may be accepted with the customer's approval, and below 1.33 does not; ongoing production is commonly held to Cpk 1.33 or more. Many supplier quality manuals follow the same split: Ppk to approve a new process, Cpk to show it stays capable.
The AIAG SPC manual adds the condition that matters most: Cpk is only meaningful when the process is in statistical control. If the control chart shows special causes, σ within doesn't describe the process, so report Ppk (a description of what was made, not a prediction) and fix the instability first.
Three Common Mistakes
- Calling s-based results "Cpk". Many spreadsheets compute Cpk with the overall standard deviation, which is Ppk. Say which σ you used.
- Reporting Cpk from an unstable process. Check the control chart first; out-of-control points make Cpk optimistic.
- Comparing a 10-part Cpk with a target as if it were exact. With 10 parts the 95% interval on this Cpk runs from about 0.59 to 1.74.
Sources
- NIST/SEMATECH e-Handbook of Statistical Methods, §6.1.6 What is Process Capability?
- NIST/SEMATECH e-Handbook, §6.3.2.2 Individuals control charts (the flow-rate data)
- AIAG (2005), Statistical Process Control (SPC) Reference Manual, 2nd edition, Automotive Industry Action Group
- AIAG (2006), Production Part Approval Process (PPAP), 4th edition, Automotive Industry Action Group
The LimitLine Newsletter
About one email a month: new guides, tool updates and one practical SPC tip. Unsubscribe in one click.