What is a good Cpk value?
Cpk 1.0, 1.33, 1.67 or 2.0? What the numbers mean in defects per million, typical customer requirements and how confident you can be with your sample size.
Updated 2026-10-24
Cpk in defects per million
For a centred, normally distributed process with two-sided limits:
| Cpk | Sigma level | Expected PPM | Typical use |
|---|---|---|---|
| 1.00 | 3 | 2,700 | Not capable, needs 100 % inspection |
| 1.33 | 4 | 66 | Minimum for running processes |
| 1.67 | 5 | 0.6 | New processes, PPAP, special characteristics |
| 2.00 | 6 | 0.002 | Six Sigma target, safety-critical |
Convert any value with the Cpk, PPM and sigma converter.
Typical requirements
- Cpk ≥ 1.33: the common minimum for an existing process in series production.
- Ppk ≥ 1.67: the AIAG PPAP acceptance level for the initial process study.
- Cmk ≥ 1.67 (sometimes 2.0): machine capability at machine acceptance, VDA 4.
- Customer-specific requirements override all of these: check the supplier quality manual.
Your Cpk is an estimate
A Cpk from 30 parts is much less certain than one from 300. The 95 % confidence interval (Bissell approximation) shows how much. Example: 125 values with Cpk 1.43 give a 95 % interval of 1.24 to 1.62. The lower end is below 1.33, so strictly speaking the data do not prove Cpk ≥ 1.33 even though the estimate is above it.
The capability calculator shows the interval next to every Cpk and Ppk, and the sample size calculator tells you how many parts you need to demonstrate a requirement.
When Cpk is meaningless
- The process is not stable: points outside the control limits or run patterns. Fix the special causes first.
- The data are not normal and the normal formula is used anyway. Use a distribution fit and the percentile method (ISO 22514-2).
- The measurement system is poor: with a %GRR of 30 % the gauge eats a large part of the tolerance. Check it with a Gage R&R.