Process Capability: Cp, Cpk, Pp, and Ppk Explained for the CQE

A process can be perfectly stable and still produce scrap. Statistical control tells you a process is predictable; capability tells you whether what it predictably produces actually meets the customer’s requirements. Process capability indices—Cp, Cpk, Pp, and Ppk—are how quality engineers quantify that, and they appear throughout the quantitative methods section of the CQE exam.

I am John Lee of Alpha Training and Consulting. Let me demystify these four indices, because the differences between them are exactly what the exam likes to test.

The core idea: process spread vs. tolerance

Capability is a ratio. The numerator is the tolerance width—the distance between the upper and lower specification limits set by the customer. The denominator is the process spread, usually six standard deviations. If the tolerance is wide relative to the spread, the process is capable. If the spread is wide relative to the tolerance, defects are inevitable.

Cp vs. Cpk: potential vs. actual centering

Cp answers a hypothetical: if the process were perfectly centered between the specification limits, could its spread fit? It ignores where the process is actually running. Cpk closes that gap by measuring the distance from the process mean to the nearer specification limit. A process can have an excellent Cp but a poor Cpk if it drifts toward one edge of the tolerance.

Cp/Cpk vs. Pp/Ppk: short-term vs. long-term

The second distinction is about which standard deviation you use. Cp and Cpk use within-subgroup variation, capturing short-term potential when only common causes are active. Pp and Ppk use the overall standard deviation of all the data, capturing long-term performance including the shifts and drifts that accumulate over hours, shifts, and material lots. A large gap between Cpk and Ppk signals that the process is being disturbed by special causes over time.

What is a good capability value?

A Cpk of 1.00 means the process spread just fits within the specifications—roughly 2,700 defects per million with a centered process. Most industries target at least 1.33, which provides a comfortable margin, and many automotive and aerospace applications require 1.67 or higher. The higher the index, the more room the process has to absorb ordinary variation without producing defects.

One caution the exam loves: capability numbers are only valid for a process in statistical control and with a trustworthy measurement system. If your gage variation is large, your capability estimate is contaminated—which is why measurement systems analysis is the natural companion to capability study.