Running both the short-term and long-term calculations on the same study gives you a diagnostic tool. Cpk uses within-subgroup sigma; Ppk uses overall sigma. If the process behaved the same way every day, those two sigmas will be close and Cpk and Ppk will nearly match. If the average moved between lots, shifts or setups, the overall sigma grows while the within-subgroup sigma stays small, and Ppk falls well below Cpk. That gap is between-subgroup drift made visible.
Which index to report depends on the question. When a customer wants to know what they received over a production run, Ppk is the honest number because it includes everything that happened. When the production team wants to know what the press is inherently able to do once the drift is removed, Cpk is the right target. Reporting only the higher of the two is not analysis; it is marketing.
Once a Cpk-to-Ppk gap is confirmed, go looking for the time-based cause. Plot the subgroup averages against the resin lot number, the shift and the setup date. The source that lines up with the jumps is where the work is.
Standard Procedure
- 1Calculate Cpk from within-subgroup sigma and Ppk from overall sigma on the same dataset.
- 2If the two are close, the process is stable over time; focus on inherent variation or centring.
- 3If Ppk is much lower than Cpk, plot subgroup averages against lots, shifts and setups to find the drift source.
- 4Report Ppk to describe what was delivered and Cpk to describe what the process could deliver once stabilised.
- 5Re-run the study after corrective action and confirm the gap has closed.
Lesson Summary
- ✓Cpk close to Ppk means little drift; Ppk far below Cpk means the average wanders between subgroups.
- ✓Choose the index by the question being asked, and never report only the flattering one.