Process Capability Index Calculator — Cp, Cpk, Cpm & PPM

Process Capability Index Calculator

Check whether a process can stay within its specification limits. Get Cp, Cpk, Cpm, and Cpkm, the expected defects per million, and the sigma level—from summary statistics or straight from your measurements.

Assumes a stable, roughly normal process

Enter the mean and standard deviation, or paste raw measurements in production order. Leave one limit blank for a one-sided specification—for example, a maximum allowed impurity with no lower limit.

Process data

How to Use This Calculator

  1. Enter the lower and upper specification limits from the drawing or customer requirement. Leave one blank for a one-sided limit.
  2. Optionally enter the target (nominal) value. If you leave it blank and both limits are set, the midpoint is used.
  3. Choose Mean & SD and enter the process mean and standard deviation, or choose Raw Measurements and paste your readings in the order they were produced.
  4. Press Calculate Capability. You get Cpk with a rating, all related indices, the expected defects in parts per million, and a chart of the process against the limits.

What Is a Process Capability Index?

A process capability index calculator measures how well a process fits within specified tolerance limits. It calculates capability indices such as Cp and Cpk using the process mean, standard deviation, and specification limits. Higher Cp and Cpk values indicate a process has greater capability to produce results within its required specifications.

Each index compares how much room the specification allows with how much the process actually varies. If the tolerance band is wide and the process is tight and well centered, the index is high and almost every part meets specification. If the process spreads close to or past a limit, the index drops and defects follow.

The indices are dimensionless, so the same scale works for millimeters, grams, minutes, or ohms. That makes them a common language between manufacturers and customers: an automotive supplier, for example, is often asked to show Cpk of at least 1.33 on critical characteristics.

Voice of the customer

Specification width, USL − LSL.

Voice of the process

Natural spread, 6σ—about 99.73% of output.

Capability

How many times the process fits inside the specification.

Background reference: ASQ — What Is Process Capability?.

Process Capability Index Formulas

Cp — potential capability

Cp = (USL − LSL) ÷ 6σ

Cpu and Cpl — one side each

Cpu = (USL − μ) ÷ 3σ
Cpl = (μ − LSL) ÷ 3σ

Cpk — actual capability

Cpk = min(Cpu, Cpl)

Cpm — Taguchi index

Cpm = (USL − LSL) ÷ 6√(σ² + (μ − T)²)

Cpkm — third-generation index

Cpkm = Cpk ÷ √(1 + ((μ − T) ÷ σ)²)

Cp tells you what the process could achieve if it were perfectly centered. Cpk tells you what it achieves now, because it measures the distance from the mean to the nearer limit. When the mean sits exactly in the middle, Cp = Cpk. The gap between them is the capability lost to poor centering, and it is usually the cheapest gap to close—moving the mean is easier than reducing variation.

Formula reference: NIST/SEMATECH e-Handbook — Process Capability.

Worked Example: Machined Shaft Diameter

A shaft is specified at 10.00 ± 0.05 mm (LSL 9.95, USL 10.05). The process runs at a mean of 10.01 mm with σ = 0.012 mm.

  • Cp = 0.10 ÷ (6 × 0.012) = 1.389. The spread alone would pass a 1.33 requirement.
  • Cpu = (10.05 − 10.01) ÷ 0.036 = 1.111 and Cpl = (10.01 − 9.95) ÷ 0.036 = 1.667, so Cpk = 1.111. The upper side is the problem.
  • Cpm = 0.10 ÷ (6 × √(0.012² + 0.01²)) = 1.067, reflecting the 0.01 mm offset from target.
  • Defects: about 429 PPM above the USL and 0.3 PPM below the LSL.
  • Fix: moving the mean back to 10.00 mm raises Cpk to 1.389 and cuts defects to about 31 PPM—with no change to the machine's variation.

Cpk, Defects, and What the Numbers Mean

For a centered, normal process with two-sided limits. With one-sided limits, or a process shifted fully toward one side, divide the PPM by about two.

CpkSigma levelDefects (PPM)YieldTypical verdict
0.672σ≈ 45,50095.45%Not capable
1.003σ≈ 2,70099.73%Barely capable—any drift causes defects
1.334σ≈ 6399.994%Capable; common minimum for production
1.504.5σ≈ 6.899.9993%Good
1.675σ≈ 0.5799.99994%Common target for new or safety-critical processes
2.006σ≈ 0.00299.9999998%World class (Six Sigma, before the 1.5σ shift)

The often-quoted "3.4 defects per million" for Six Sigma assumes the mean drifts 1.5σ over the long term, leaving 4.5σ to the nearer limit. Without that shift, a true 6σ process produces about 0.002 PPM.

Cp/Cpk vs. Pp/Ppk: Which Standard Deviation?

The formulas are identical; only the σ changes. Mixing the two is the most common reason capability figures disagree between suppliers and customers.

FeatureCp / Cpk (capability)Pp / Ppk (performance)
σ usedWithin-subgroup (short-term)Overall sample SD (long-term)
Estimate for individual valuesAverage moving range ÷ 1.128Ordinary SD with n − 1
Includes drift and shiftsNoYes
AnswersWhat the process can do when stableWhat the customer actually received

In Raw Measurements mode the calculator reports both. If Ppk is much lower than Cpk, the process is moving between periods—shift changes, tool wear, new material lots—and the fix is to find that cause, not to tighten the machine.

Within-capability reference: Minitab Support — Potential (within) capability.

Before You Trust the Index

Check first

  • The process is stable—a control chart shows no trends or out-of-control points.
  • The data are roughly normal; check a histogram or normal probability plot.
  • The gauge is good enough—measurement error is small next to the tolerance.
  • There are enough data points; 25 subgroups or 100 values is a common minimum.

Common mistakes

  • Reporting Cpk from a handful of parts—with 30 values, a true Cpk of 1.33 can easily show up anywhere from about 1.0 to 1.7.
  • Using Cpk on skewed data such as flatness, runout, or impurity, which are bounded at zero.
  • Comparing Cpk from one supplier with Ppk from another.
  • Treating a high Cpk as proof the process will stay that way without monitoring.

Frequently Asked Questions

What is a good Cpk value?

1.33 is the most common minimum for an established manufacturing process, which means at least 4σ between the mean and the nearer limit. Many customers ask for 1.67 on new or safety-related characteristics, and quality management systems in industries such as automotive often write these thresholds into supplier requirements. Below 1.0, the process is producing defects even when everything runs normally.

Why is my Cpk lower than my Cp?

Because the process average is not in the middle of the tolerance. Cp only looks at the process spread; Cpk looks at the distance from the mean to the nearer limit, so it also reflects process centering. Moving the mean toward the target value raises Cpk toward Cp. Cpk can never be higher than Cp.

Can Cpk be negative?

Yes. A negative Cpk means the process mean lies outside the specification—above the upper specification limit or below the lower specification limit—so more than half of the output fails conformance and the yield is below 50%. It calls for immediate adjustment of the process setting before production continues.

How do I calculate Cpk with only one specification limit?

Use only the side that exists: Cpk = Cpu for a maximum limit, Cpk = Cpl for a minimum limit. Cp and Cpm cannot be calculated, because there is no tolerance width. The same rule applies to Pp and Ppk. Leave the missing limit blank in the calculator.

When should I use Cpm instead of Cpk?

When being on target matters, not just being inside the limits. Cpm follows Taguchi's idea that any deviation from nominal carries a cost to product quality, so it drops as soon as the mean drifts from the target, even if Cpk still looks fine. Cpkm combines both penalties into a single process capability index.

How do I improve process capability?

First center the process, which raises Cpk up to Cp. Then reduce process variation by addressing its sources—material, tooling, setup, temperature, operator method—which raises both. Statistical process control, a core quality control method, helps here: a control chart shows when the process shifts so you can react before parts go out of spec. Widening the tolerance also raises the index, but only if the customer's design allows it.

What if my data are not normally distributed?

The PPM estimates can then be badly off, especially in the tails, because they assume a normal distribution. Plot your measurements first; a skewed dataset is common for characteristics bounded at zero. Common options are to transform the data (for example with a Box–Cox transformation), to fit a better-matching distribution such as lognormal or Weibull, or to use percentile-based indices that replace 6σ with the distance between the 0.135th and 99.865th percentiles.

How is the sigma level related to Cpk?

The short-term sigma level is 3 × Cpk—the number of standard deviations between the mean and the nearer limit. Cpk 1.33 is a 4σ process and Cpk 2.0 is a 6σ process. Some Six Sigma tables add 1.5 to account for long-term drift, which is why they report process performance (Ppk) as well as capability—check which convention a source uses. Note also that sigma level is measured against the specification limits set by the customer, not against control limits. Control limits are process limits at ±3σ, calculated from a sample of production data and used to monitor process stability.

Other Useful Calculators

Disclaimer: This calculator provides process capability estimates for educational, informational, and preliminary analytical use. Results assume a stable process and approximately normal data, and depend entirely on the specification limits, statistics, or measurements entered. Defect rates are model-based estimates; capability indices calculated from small samples carry substantial uncertainty and may differ from those produced by software that uses other estimation methods or subgroup structures. For product qualification, customer submissions (such as PPAP), regulatory, or safety-related decisions, follow the applicable standards and customer requirements and consult a qualified quality engineer or statistician.

Last updated: October 5, 2026