Cp and Cpk: Reading Process Capability the Right Way

Process capability indices answer a question every quality team faces: is this process tight enough, and well-centred enough, to reliably meet its specification? Two indices do the work. Cp compares the width of the specification to the natural spread of the process — six standard deviations of variation — and tells you whether the process could fit inside the limits if it were perfectly centred. Cpk goes further by accounting for where the process is actually centred, so it reflects how close you are to whichever specification limit you are nearest. The two are easy to confuse, and confusing them leads to confident, wrong conclusions.

Both indices rest on assumptions that matter: the process should be stable and in control, and the data roughly normal. Run the capability study on an out-of-control process — which a lot of teams did while recovering shaky supply and staffing in early 2021 — and the numbers describe nothing real. Capability is a statement about a predictable process, not a wish for an unpredictable one.

Reading the two indices well

1. Confirm control first. Use a control chart to show the process is stable before you compute any capability index. Capability on an unstable process is meaningless.

2. Read Cp as potential. A high Cp says the spread is narrow enough to fit the spec window — if the process were centred. It says nothing about where the process actually sits.

3. Read Cpk as reality. Cpk accounts for off-centre running. When Cp is high but Cpk is low, the process is capable in principle but drifting toward one limit.

4. Compare the two to diagnose. Cp and Cpk close together means a centred process; a big gap between them points straight at a centring problem you can correct.

What good interpretation looks like

• The team verifies the process is in control before trusting any index.

• A gap between Cp and Cpk is read as a centring issue and triggers an adjustment of the process mean, not a panic about variation.

• Targets are set in context — a common rule of thumb is Cpk of 1.33 as a working minimum for many processes, but the team checks what their customer or standard actually requires.

• Indices are tracked over time, so a slow drift in Cpk is caught before it produces defects.

What bad interpretation looks like

• Quoting a glowing Cp while ignoring a poor Cpk, and calling the process fine when it is quietly hugging a limit.

• Computing capability on a process that is not in statistical control, then trusting the number.

• Chasing more decimal places of Cpk instead of asking whether the spec itself reflects real customer needs.

• Treating 1.33 as a sacred law rather than a guideline, and over-engineering a process whose customer is satisfied at less.

Used together and honestly, Cp and Cpk are a fast read on whether a process can be trusted and whether it needs centring or tightening. Used carelessly, they become a number people quote to feel safe. Confirm control, read potential against reality, fix centring before you fight variation, and hold the index to what your customer actually needs.

If you want to build capability analysis into how your organization runs its key processes, XNM's strategic advisory can help you put it to work.