Skip to content

Learn

Cp, Cpk, Pp and Ppk explained

What the four process capability indices measure, how they differ, how to calculate them with a worked example, and what a gap between Cpk and Ppk tells you.

Published · Updated · 6 min read

What process capability measures

Process capability compares how much a process varies with how much variation the specification allows. For a characteristic with a lower specification limit (LSL) and an upper specification limit (USL), the allowed spread is USL − LSL. A capable process fits comfortably inside it.

Cp and Cpk

Cp = (USL − LSL) / (6σ) describes the potential of the process: it ignores where the process is centred. Cpk = min((USL − μ) / 3σ, (μ − LSL) / 3σ) takes the mean μ into account and equals the distance to the nearest limit measured in units of 3σ. Cpk is never larger than Cp, and they are equal only when the process is perfectly centred. For a one-sided specification, report the relevant upper or lower capability index. Cp and Pp require two limits.

In Cp and Cpk, σ is a short-term, within-subgroup estimate. For individual measurements it is commonly estimated from the average moving range divided by 1.128.

A worked example

A diameter is specified as 10.00 ± 0.06 mm, so LSL = 9.94 and USL = 10.06. The process has a mean of 10.02 mm and a short-term standard deviation of 0.01 mm.

Cp = 0.12 / 0.06 = 2.00. Cpk = min((10.06 − 10.02) / 0.03, (10.02 − 9.94) / 0.03) = min(1.33, 2.67) = 1.33. The process has plenty of potential, but because it is running off-centre toward the upper limit, its capability is much lower.

Pp and Ppk

Pp and Ppk use the same formulas but with the overall standard deviation of all data, so they include drift and shifts between subgroups. They describe the performance actually delivered over the period rather than the short-term potential.

Keep the mean and limits from the diameter example, but suppose the overall standard deviation is 0.02 mm. Pp is 0.12 / (6 × 0.02) = 1.00 and Ppk is 0.04 / (3 × 0.02) = 0.67. Now you can see how much the short-term estimate misses.

Why the gap matters

Plot the measurements in production order and mark tool changes and material lots. A steady climb within each tool run suggests wear. Two tight clusters from different machines suggest a difference between setups. Both can produce the same Cpk–Ppk gap.

Qynetic computes these indices with deterministic, automatically tested code. They are not produced by a language model; AI is used to reason about the results.

Limits of the indices

The indices assume a stable process and, in their simplest form, roughly normal data. Check stability with a control chart before trusting a capability number, and use enough data: small samples give uncertain estimates.

Standards and further reading

Book a call about your factory

Tell us what you make and which machines you run, and we will walk through how Qynetic fits. If you would rather get news and early access, join the waitlist.

Learn