What is the difference between Cpk and Ppk?

Cpk measures short-term process capability using within-subgroup sigma, which captures only common-cause variation. Ppk measures long-term process performance using the overall standard deviation of all the data, which includes both common-cause and special-cause variation. Cpk is what the process could do if stable; Ppk is what it actually did over the whole run.

Is Cpk always higher than Ppk?

For a stable, in-control process Cpk and Ppk are nearly equal. When the process drifts or shifts over time, the overall sigma grows larger than the within-subgroup sigma, so Ppk falls below Cpk. A large gap between Cpk and Ppk is a signal that the process is not in statistical control.

Does PPAP require Cpk or Ppk?

AIAG PPAP normally asks for Ppk on the initial submission because the run is short and not yet proven stable, with a typical requirement of Ppk greater than or equal to 1.67 for special characteristics. Once the process is established and in control, ongoing capability is reported as Cpk, often with a target of 1.33 or higher.

What is a good Cpk or Ppk value?

A common minimum for ongoing capability is Cpk 1.33, which is roughly 4 sigma and about 63 defects per million for a centred process. Safety or special characteristics often require 1.67 or higher. A value of 1.0 means the process spread just fills the tolerance with no margin, so it is not acceptable for most customers.

Cpk vs Ppk: capability vs performance.

Quality / SPC June 30, 2026 9 min read 1,650 words

Cpk and Ppk look almost identical and use almost the same formula, yet customers ask for different ones at different times. The difference is the sigma in the denominator: Cpk uses within-subgroup variation (short-term capability), Ppk uses the overall variation of all the data (long-term performance). This guide explains both, with a worked example and clear guidance on which to report.

Capability versus performance

Capability is what your process could do if it stayed perfectly stable. Performance is what it actually did across the whole production run, drift and all. Cpk is the capability index; Ppk is the performance index. Both compare the spread of your process against the tolerance band, and both account for centring — the difference is entirely in how the variation is estimated.

If you already know the basics of process capability, this article is the natural next step after our Cp vs Cpk guide, which covers spread versus centring. Here the question is short-term versus long-term.

What Cpk is

Cpk is the process capability index. It uses the within-subgroup standard deviation, estimated from the average range or average sigma of your rational subgroups. Because subgroups are small samples taken close together in time, this sigma captures only common-cause (random) variation — the inherent noise of the process when nothing special is happening.

The formula is the smaller of the two single-sided indices:

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

Because it uses the tight within-subgroup sigma, Cpk answers: "if this process were perfectly stable, how capable would it be?"

What Ppk is

Ppk is the process performance index. It uses the overall standard deviation calculated from every data point in the study, ignoring subgroup structure. That overall sigma absorbs everything — common-cause noise plus any shifts, drifts, tool wear, and between-shift differences that happened during the run.

The formula mirrors Cpk:

Ppk = min[ (USL − mean) / 3σ_overall , (mean − LSL) / 3σ_overall ]

Ppk answers the blunter question: "regardless of why, how well did the process actually meet the tolerance over this whole run?"

The one real difference: the sigma

Everything turns on which standard deviation goes in the denominator.

Term	Sigma used	Captures
Cpk (capability)	Within-subgroup (R-bar/d2 or S-bar/c4)	Common-cause only
Ppk (performance)	Overall (all data, n−1)	Common + special cause

For a process that is genuinely in statistical control, the two sigmas are close, so Cpk and Ppk land near each other. When the process drifts, the overall sigma swells while the within-subgroup sigma stays tight, and Ppk drops below Cpk. That gap is diagnostic gold — it tells you instability, not just spread, is eating your margin.

Cpk vs Ppk at a glance

Aspect	Cpk	Ppk
Name	Capability	Performance
Time frame	Short-term, potential	Long-term, actual
Sigma	Within-subgroup	Overall
Needs control?	Assumes stability	No assumption
Typical use	Ongoing monitoring	Initial PPAP run

Worked example

Take a shaft diameter with USL 25.10 mm, LSL 24.90 mm, so the tolerance is 0.20 mm and the target is 25.00 mm. You run 25 subgroups of 5 parts. Suppose:

Process mean = 25.02 mm
Within-subgroup sigma = 0.025 mm
Overall sigma = 0.035 mm (larger, because the mean drifted across shifts)

Cpk: the nearer limit is USL. (25.10 − 25.02) / (3 × 0.025) = 0.08 / 0.075 = 1.07.

Ppk: (25.10 − 25.02) / (3 × 0.035) = 0.08 / 0.105 = 0.76.

Cpk says the process is borderline capable; Ppk says actual performance is poor. The gap (1.07 vs 0.76) is the drift. Chasing a tighter machine here would be wrong — the fix is stabilising the mean across shifts. Drop your own numbers into the Cp Cpk calculator to see both indices side by side.

A high Cpk with a low Ppk is a warning, not a pass It means the machine is precise but the process is not controlled. Reporting only the flattering Cpk hides real risk that will surface as field rejections. Always look at both before you sign off.

Which index should you report?

It depends on where you are in the part's life:

Initial PPAP / new launch: report Ppk. The run is short and stability is unproven, so the conservative overall sigma is honest. AIAG PPAP commonly requires Ppk ≥ 1.67 for special characteristics.
Ongoing production: report Cpk once the process is demonstrably in control on a control chart. Typical target is Cpk ≥ 1.33.

For the full automotive submission context, see our PPAP guide for Indian OEMs, and remember capability numbers are only trustworthy if your measurement system is sound — validate it first with a Gauge R&R study.

Run the numbers Use the Cp Cpk calculator to compute capability from your data, then grab a ready inspection sheet from the free templates. To turn a customer drawing into a numbered characteristic list for the capability study, CadNexa's auto-ballooning tool reads dimensions and tolerances straight off the PDF.

Common Cpk and Ppk mistakes

Quoting Cpk on an unstable process. Cpk assumes statistical control. If the control chart shows out-of-control points, the within-subgroup sigma is meaningless.
Mixing the sigmas. Using overall sigma but calling it Cpk, or vice versa, makes the number uncomparable. Be explicit about which sigma you used.
Ignoring the Cpk–Ppk gap. A wide gap is free diagnostic information about drift. Do not average it away.
Reporting capability with a bad gauge. If Gauge R&R is poor, the measurement noise inflates sigma and your indices are fiction.
Treating 1.0 as acceptable. Cpk 1.0 leaves zero margin. Most customers want 1.33 minimum, 1.67 for special characteristics.

Rajadurai R

Founder — 14 years plant-head experience