Cpk vs Ppk: what is the difference?
Cpk and Ppk use exactly the same formula. The only difference is which standard deviation goes into it, and that small difference answers two different questions about your process.
| Cpk | Ppk | |
|---|---|---|
| Sigma used | Within (short-term): MR̄/d2 or R̄/d2 | Overall (long-term): sample standard deviation |
| Question it answers | What the process could do if it stayed where it is | What the process actually delivered over the study period |
| Sensitive to drift, shifts, tool wear | No — they are hidden in the within estimate | Yes — every shift widens the overall spread |
| Typical use | Ongoing production of a stable process | Initial process studies, PPAP, customer reports |
Why the two numbers differ
σwithin is estimated from differences between neighbouring points (moving ranges) or from ranges inside each subgroup. Slow changes in the process mean — a new material lot, tool wear, a temperature cycle — barely affect it. σoverall is calculated from all the data at once, so those changes are included.
Here are two simulated processes with the same specification (9.90–10.10), the same short-term noise and 50 individual measurements each. In the second one the mean shifts by 0.06 halfway through the run:
| Process | σwithin | σoverall | Cpk | Ppk |
|---|---|---|---|---|
| Stable | 0.0197 | 0.0194 | 1.68 | 1.70 |
| Mean shifts halfway | 0.0208 | 0.0331 | 1.59 | 1.00 |
For the stable process Cpk and Ppk are practically equal. For the shifting process Cpk still looks excellent, because each pair of neighbouring points is close together — but Ppk shows that the parts the customer actually received were only just inside the specification.
How to read the gap
- Ppk ≈ Cpk: the process is stable; the short-term capability is what you deliver.
- Ppk clearly below Cpk: there is variation between subgroups or over time. Look at the control chart for shifts, trends or cycles — fixing them raises Ppk toward Cpk without touching the machine's inherent precision.
- Ppk above Cpk: unusual. It often means the data are alternating (consecutive points systematically differ), or subgroups are too small or not rational. Check how the data were collected.
Which one should you report?
Many customers follow the AIAG PPAP manual, which uses a process performance (Ppk) index for the initial process study: above 1.67 meets the acceptance criteria, 1.33 to 1.67 may be acceptable, and below 1.33 does not. For ongoing production of a stable process, Cpk ≥ 1.33 is a common requirement. When in doubt, report both and show the control chart — the pair tells the whole story.
Get both numbers in one paste
The free calculator reports Cp, Cpk, Pp and Ppk side by side, with the control chart that explains any gap.
Open the free Cpk calculator →