Process Capability Analysis: Cp, Cpk, Pp, and Ppk Demystified

Leave a comment

Process capability indices are the language Lean Six Sigma uses to translate raw statistical data into a clear answer to one question: can this process consistently meet customer requirements?

Cp, Cpk, Pp, and Ppk are the four most common indices, and the choice between them, along with the difference between them, is a frequent source of confusion in practice.

This guide explains what each index measures, when to use which, and how to interpret the numbers in the context of a real Six Sigma project.

Capability vs Performance: The Underlying Distinction

The four indices split into two pairs. Cp and Cpk are capability indices: they describe what the process is capable of doing under normal stable operation, using short-term variation. Pp and Ppk are performance indices: they describe what the process actually delivers over the long run, including all sources of variation.

A stable, well-controlled process will have Cp and Pp values that are close to each other. A process whose long-term behaviour drifts from its short-term snapshot will show Pp meaningfully lower than Cp, which is itself a useful diagnostic about how well the process is controlled.

The ASQ Exam Prep resource on process capability gives a worked numerical example with the formulas; this article focuses on the practical use of the indices once you have them.

The Four Indices

Cp: Short-term potential capability

Cp = (USL – LSL) / (6 sigma), where sigma is the short-term standard deviation, typically estimated from the average range of rational subgroups. Cp answers the question: if the process were perfectly centred between the specification limits, how would its spread compare to the tolerance?

A Cp of 1.0 means the natural process spread exactly fills the tolerance, leaving no margin. A Cp of 1.33 is the traditional minimum for capable processes; Cp of 2.0 corresponds to a Six Sigma process.

The critical limitation of Cp is that it ignores centring. A perfectly centred process with Cp = 1.5 and a wildly off-centre process with Cp = 1.5 produce the same index but very different defect rates.

Cpk: Short-term centred capability

Cpk = min((USL – mean) / (3 sigma), (mean – LSL) / (3 sigma)). Cpk accounts for centring by measuring the distance from the mean to the nearer specification limit. When the process is perfectly centred, Cpk = Cp. When the process is off-centre, Cpk is smaller than Cp, and the gap tells you how much capability is being lost to mis-centring.

In practice, Cpk is the more useful of the two short-term indices for most projects. It is the number the customer cares about, because it reflects what the process will actually deliver.

Pp: Long-term overall performance

Pp uses the same formula as Cp but with the long-term overall standard deviation calculated directly from the full data set. It captures the variation that includes all the things that happen over time: shift changes, operator changes, supplier batch differences, gradual drift. Pp is typically lower than Cp.

Ppk: Long-term centred performance

Ppk relates to Pp the way Cpk relates to Cp: it accounts for centring using the overall standard deviation. Ppk is the most realistic single-number summary of what the customer will experience over time.

Interpreting the Numbers

The conventional benchmarks have been stable for decades. Cpk or Ppk less than 1.0 means the process is incapable: defects will be common. A value of 1.0 to 1.33 means the process is marginally capable: some defects will occur.

A value of 1.33 to 1.67 means the process is capable: defects will be rare. A value above 1.67 is excellent, and the traditional Six Sigma target of Cpk = 2.0 corresponds to roughly 3.4 defects per million opportunities once the conventional 1.5 sigma long-term shift is accounted for.

A frequent question is what target to set for a new project. The answer depends on the customer’s tolerance for defects and the cost of variation. In high-volume automotive manufacturing, 1.67 is often the floor; in laboratory analytics, 1.33 is more common. The right answer is set by the business and the customer, not by a default in the textbook.

Prerequisites for Valid Capability Analysis

Capability analysis only produces meaningful indices if three conditions are met.

  • The process must be stable (in statistical control). A capability index calculated on an unstable process is misleading because the underlying distribution is not even a single thing to characterise.
  • The data should be approximately normal, or the analysis must be adapted (a transformation, or a non-normal capability procedure).
  • The measurement system must be capable. If Gage R&R consumes a large fraction of the tolerance, the apparent process variation is partly measurement noise, and the capability index is artificially low.

Skipping any of these prerequisites is one of the most common errors in capability work. Many published reports of low Cpk turn out, on investigation, to reflect measurement problems or special cause variation rather than the underlying process.

Capability Analysis in DMAIC

Capability analysis appears in two phases. In Measure, the baseline Cpk or Ppk quantifies how far the current process is from meeting requirements, which feeds the project charter and the financial case. In Control, the post-improvement Cpk or Ppk confirms the gain and becomes the headline metric the sponsor signs off on.

Between those two phases, the team works on stabilising the process (SPC), eliminating special causes, validating the measurement system (Gage R&R), and reducing common cause variation through redesign. The capability index moves as those activities succeed.

The Sigma Level Equivalent

Six Sigma practitioners often express capability as a sigma level rather than a Cpk. The translation is mechanical: a sigma level of Z = 3 sigma short-term corresponds to a Cpk of approximately 1.0, a Z of 4.5 sigma short-term corresponds to Cpk of approximately 1.5, and Z of 6 sigma short-term corresponds to Cpk of approximately 2.0. The convention adds a 1.5 sigma long-term drift, which is why ‘Six Sigma’ in industry shorthand means Cpk of 2 short-term and 3.4 defects per million long-term.

For deeper coverage of capability analysis within accredited certification, the ILSSI Black Belt syllabus includes capability for non-normal data, multivariate capability, and process capability for short-run processes.

A Practical Example

A pharmaceutical packaging line had a fill-weight specification of 100 grams plus or minus 2 grams. Twenty-five subgroups of five units showed a stable I-MR pattern, a mean of 100.4 grams, and a short-term sigma estimate of 0.45 grams. Cp = 4 / (6 x 0.45) = 1.48. Cpk = min((102 – 100.4) / (3 x 0.45), (100.4 – 98) / (3 x 0.45)) = min(1.19, 1.78) = 1.19.

The process is marginally capable, and the constraint is the upper specification limit because the mean is off-centre. Recentring the process by 0.4 grams would lift Cpk close to Cp and dramatically reduce the expected defect rate at the upper limit. This is the kind of focused, evidence-based recommendation that capability analysis enables.

Common Mistakes

  • Calculating capability on an unstable process and reporting the result as if it were meaningful.
  • Reporting Cpk without also reporting Cp, which hides the contribution of centring.
  • Assuming normality without checking it.
  • Ignoring measurement system error when the index is low.
  • Comparing Cpk values across processes with different tolerances, which is not a like-for-like comparison.

Final Thoughts

Process capability indices have outlived many fashions in quality management because they answer a question every customer asks: how often will this process let me down? When calculated correctly on a stable process with a validated measurement system, Cpk and Ppk are powerful summary statistics. When calculated carelessly, they mislead. The discipline of getting them right is the practical heart of statistical thinking in Lean Six Sigma.

To formalise these skills, explore the ILSSI certification programmes or browse the research archive for industry case studies.