Pp and Ppk describe overall process performance against specification limits using total observed variation, often before long-term stability is proven.
Definition
Process Performance uses Pp and Ppk indices to compare observed process spread and centering to specification limits. Pp estimates potential performance using total standard deviation, while Ppk accounts for both spread and distance from the nearest specification limit.
Pp/Ppk are often contrasted with Cp/Cpk, which are associated with within-subgroup or shorter-term capability estimates in stable processes.
History
Capability and performance indices developed in quality engineering as ways to communicate process spread relative to specifications. Pp and Ppk became useful where teams needed an overall view of actual observed performance, especially during studies, launches, and customer reporting.
When to Use
Use Pp/Ppk when evaluating total observed process performance over a defined period, especially when long-term sources of variation are included. Confirm stability and distribution assumptions before treating the indices as reliable capability evidence.
Step-by-Step
- Confirm specification limits and measurement-system suitability.
- Collect representative process data over the intended period.
- Review control charts for stability and special causes.
- Check distribution shape and normality assumptions.
- Calculate Pp from total spread relative to tolerance.
- Calculate Ppk using distance to the nearest specification limit.
- Interpret indices with defect risk, centering, and process context.
- Improve variation and centering, then remeasure.
Examples
- Launch: A supplier reports Ppk after a production trial.
- Process review: Ppk shows performance is limited by off-center operation.
- Transactional: Cycle-time performance is reviewed with non-normal methods instead of normal Ppk.
Common Pitfalls
- Using Pp/Ppk on unstable data without explanation.
- Ignoring measurement error.
- Comparing Ppk to Cpk without understanding variation basis.
- Assuming normality for skewed data.
- Reporting an index without sample size and timeframe.
- Chasing index targets without reducing customer risk.