Cp vs Cpk explained: process capability made simple.

SPC / Capability June 13, 2026 7 min read 1,300 words

Cp and Cpk both tell you whether a process can hold a tolerance — but only one of them notices when the process drifts off-center. Get the difference, the formulas, and the numbers your customer expects, in plain language.

What process capability actually measures

Capability compares the spread of your process to the width of the tolerance. If your parts vary far less than the tolerance allows, you are capable; if the variation eats most of the tolerance, you are not. Cp and Cpk put a number on that.

The Cp formula

Cp is the simple ratio of tolerance width to process spread:

Cp = (USL − LSL) / (6σ)

USL and LSL are the upper and lower specification limits; 6σ is six times the process standard deviation (roughly the full natural spread). A Cp of 1.0 means the process spread exactly fills the tolerance — cutting it fine. The catch: Cp assumes the process is perfectly centered, which it rarely is.

The Cpk formula

Cpk fixes that blind spot by measuring the distance from the process mean to the nearest spec limit:

Cpk = min[ (USL − mean) / 3σ , (mean − LSL) / 3σ ]

Because it takes the worse of the two sides, Cpk drops as the process drifts off-center, even if the spread is unchanged. That is why customers ask for Cpk, not Cp.

The one-line difference Cp asks "is the process narrow enough?" Cpk asks "is it narrow enough and centered?" Cpk is always less than or equal to Cp.

Cp vs Cpk: a quick example

Tolerance 10.00 ±0.10 mm (USL 10.10, LSL 9.90), σ = 0.025 mm. Cp = 0.20 / 0.15 = 1.33 either way. But if the mean sits at 10.05 instead of 10.00, Cpk = (10.10 − 10.05) / (3 × 0.025) = 0.67 — capable on paper (Cp 1.33), failing in reality (Cpk 0.67). The part is drifting toward the upper limit.

Target values your customer expects

CpkMeaningTypical use
< 1.00Not capable — expect rejectsAction required
1.33Capable — the common minimumMost automotive / general
1.67High capabilitySafety / critical characteristics
2.00Six Sigma levelWorld-class processes

Run your own numbers in the free Cp/Cpk calculator — paste your measurements and it returns Cp, Cpk, and a capability chart.

Common mistakes

Four traps Computing capability on an unstable process (fix stability first) · too small a sample (use 30+ readings) · reporting Cp when the customer wants Cpk · ignoring measurement error — a weak gauge inflates σ. Check your gauge with a Gauge R&R study first.

FAQ

What is a good Cpk value?

1.33 is the common minimum for capability; 1.67 or higher is expected for safety-critical characteristics. Below 1.00 the process will produce out-of-spec parts.

Why is Cpk lower than Cp?

Cp ignores centering; Cpk penalizes any shift of the mean away from the target. If the process is perfectly centered, Cp and Cpk are equal.

How many samples do I need?

Use at least 30 measurements from a stable process; more is better. Small samples give unreliable estimates of σ.

Does Cpk relate to PPAP?

Yes. PPAP's initial process studies element typically requires capability evidence, often Cpk ≥ 1.33 on significant characteristics.

Where do the measurements come from?

From inspecting each characteristic on the drawing. Auto-balloon the drawing first with CadNexa's Smart Detect, then measure against each ballooned characteristic.

RR
Rajadurai R
Founder, MetricMech & CadNexa · Mechanical engineer, 14 years plant-head experience