What Is the Difference Between Cp, Cpk, Pp, and Ppk?

By Hélène · May 11, 2026

Cp, Cpk, Pp, and Ppk are the four indices most quality and process engineers reach for when a customer, an internal scorecard, or a launch readiness review asks for a capability number. They look similar, they use the same specification limits, and they are often reported side by side. They do not, however, mean the same thing, and confusing them is one of the most common reasons capability conversations on the floor go sideways.

This guide explains what each index measures, when to use it, how to interpret the gap between Cpk and Ppk, and what to check before trusting any of these numbers in a decision.

What each index actually measures

All four indices compare the voice of the process (how much the data varies) to the voice of the customer (the specification limits). The difference is in how they estimate process variation and whether they account for the process being off-center.

• Cp - Potential short-term capability. Compares the specification width to short-term variation, estimated from within-subgroup variation (typically using R-bar or S-bar). Ignores the process mean. Answers: if my process were perfectly centered, could it fit inside the specs?
• Cpk - Actual short-term capability. Same short-term variation as Cp, but uses the distance from the process mean to the nearer specification limit. Penalizes off-center processes.
• Pp - Potential overall performance. Compares the specification width to the overall standard deviation of all observed data, with no subgrouping assumption. Ignores the process mean.
• Ppk - Actual overall performance. Uses the overall standard deviation and the distance from the mean to the nearer specification limit. The most realistic single-number summary of how the process actually performed during the study window.

The shortcut most engineers remember: Cp and Cpk describe what the process is capable of when it is behaving; Pp and Ppk describe what it actually delivered, drift and all.

Short-term vs long-term variation

The deeper distinction is in how the standard deviation is estimated.

Cp and Cpk use a within-subgroup estimate of sigma. If you collect five parts every hour for a week, the within-subgroup variation is the variation inside each hourly subgroup. It captures common-cause variation while the process is in a stable state, and it deliberately filters out hour-to-hour or shift-to-shift shifts.

Pp and Ppk use the overall sample standard deviation across every measurement in the study. That picks up everything: common cause, shift changes, tool wear, lot-to-lot material differences, operator changeovers, and any drift over the study window.

That is why Cpk is almost always equal to or higher than Ppk for the same data. When Cpk is much higher than Ppk, the process has meaningful between-subgroup variation that the within-subgroup estimate is hiding. When Cpk and Ppk are close, the process is behaving consistently across subgroups.

In quality reviews, this gap is one of the first things experienced engineers look at. A wide Cpk-to-Ppk gap is rarely a measurement problem; it is usually a signal that the process is not as stable as the control chart summary suggests.

A manufacturing scenario that makes the difference visible

Consider a CNC machined shaft with a diameter specification of 10.00 mm plus or minus 0.05 mm. The team collects five parts per hour for one week and runs a capability study.

• The within-subgroup standard deviation is small because parts machined back-to-back are very similar.
• The overall standard deviation is larger because the tool wears between regrinds, ambient temperature drifts across shifts, and a new bar stock lot was introduced midweek.

In that study, Cp and Cpk look strong because they only see the tight short-term variation. Pp and Ppk are noticeably lower because they include the drift and the lot change. If the team only reports Cpk to the customer, they are reporting the best case. If they only report Ppk, they are not separating out which part of the variation comes from a stable process and which part comes from special causes that could be reduced.

Both numbers are useful. Reporting only one usually hides the real story.

When to use Cp and Cpk vs Pp and Ppk

The choice depends on what decision the index is supporting.

Use Cp and Cpk when:

• the process has been demonstrated stable on a control chart for the study period
• you are characterizing inherent process capability for design tolerancing or DOE follow-up
• you are answering "is this process technically capable if we can keep it centered and stable"
• subgrouping is rational and the within-subgroup estimate genuinely reflects common-cause variation

Use Pp and Ppk when:

• the process is new, recently changed, or not yet demonstrated stable
• you are running a PPAP or a production part approval where overall performance is the contract
• you are reporting actual delivered performance to a customer or internal scorecard
• you want a single index that does not assume stability

In practice, most automotive, aerospace, and medical device customers expect Ppk during initial qualification (the process has not earned a stability claim yet) and Cpk during ongoing production (with a control chart maintaining the stability assumption).

If you are switching between Minitab, Excel, and other tools to run these studies, the index definitions are the same, but the subgrouping defaults differ. Choosing the right tool for stratified capability work is its own decision and worth thinking through before a launch.

Interpreting the numbers

Common reference points used across manufacturing:

• 1.00 - process variation just barely fits inside the specs if perfectly centered. Almost no margin.
• 1.33 - historically considered minimum acceptable for many production processes.
• 1.67 - typical target for safety-related, automotive, and aerospace characteristics.
• 2.00 - sometimes required for critical-to-safety features or where process capability needs significant headroom.

These thresholds apply to Cp, Cpk, Pp, and Ppk in the same way. The number itself is comparable; what differs is the variation estimate behind it.

A few interpretation rules worth applying every time:

• If Cp is high but Cpk is low, the process is capable but off-center. The fix is mean adjustment, not variation reduction.
• If Cpk and Cp are both low, variation is the issue and centering will not save you.
• If Cpk is high but Ppk is much lower, the process drifts between subgroups. Investigate special-cause sources before trusting the short-term number.
• If Ppk is acceptable but the control chart shows out-of-control points, the index is misleading. Capability indices assume the process is stable in the first place. Reading control chart rules carefully before trusting any capability number is part of basic decision hygiene.

Practical action block: what to check before trusting any capability number

Before sending a Cp, Cpk, Pp, or Ppk number into a customer report, launch decision, or capability scorecard, verify:

1. Measurement system is acceptable. Run or confirm a recent gage R&R. A capability study on a poor measurement system tells you almost nothing about the process.
2. Control chart shows stability. Cp and Cpk assume stability. If the chart has out-of-control signals during the study window, the within-subgroup sigma is not a fair estimate.
3. Subgrouping is rational. Subgroups should capture short-term, common-cause variation. Five consecutive parts is rational; five parts pulled across an entire shift usually is not.
4. Sample size is adequate. Aim for at least 25 to 30 subgroups for short-term studies and at least 100 individual measurements for overall performance studies when practical.
5. Distribution is reasonable. Cp, Cpk, Pp, and Ppk assume approximate normality. Heavily skewed or bounded data may need a transformation or a non-normal capability method.
6. The right index is being reported. Match the index to the decision: Ppk for "what did we deliver," Cpk for "what is this process capable of when stable."
7. Cpk-to-Ppk gap is interpreted. A wide gap is a signal, not a scorekeeping convenience. Investigate before reporting.

Common mistakes

In quality reviews, the same handful of mistakes show up repeatedly:

• Reporting Cpk when the process has never been demonstrated stable. The number looks fine; the assumption underneath does not hold.
• Reporting Ppk only and using it to argue that variation is inherent, when most of the variation is special-cause and reducible.
• Comparing capability indices across two products with different measurement systems or different subgrouping strategies. The numbers are not comparable.
• Using Cpk to claim "we will hit this tolerance going forward" without a control chart in place to keep the stability assumption alive.
• Treating any single capability number as the final answer instead of one input into a process decision.

How leaders should read these numbers

For plant leaders and quality directors, the practical questions are:

• Is the team reporting Cpk or Ppk, and does that match the decision being made?
• Has the process been demonstrated stable, or is the team assuming it?
• What is the gap between Cpk and Ppk, and what is being done about it?
• Are these numbers based on a measurement system that has been validated recently?
• What changes when the index moves from 1.33 to 1.67? If nothing operationally changes, the number is being collected as a metric rather than used as a decision tool.

A capability index is only as trustworthy as the stability, measurement, and sampling behind it. Asking these questions consistently is how leaders build the statistical judgment in their teams that turns capability reporting from a paperwork exercise into a real decision input.

When a Cpk drops and the team cannot explain why, the next step is usually structured investigation, not a recalculation. Treating a capability decline as a symptom and walking it from symptoms to root cause is faster than re-running the study with different assumptions.

Reporting capability cleanly

When the audience is a plant manager, a customer, or a launch review board, the most useful summary usually includes:

• The index used (Cp, Cpk, Pp, or Ppk) and why
• The specification limits and the sample size
• A short note on stability (control chart confirmed or not)
• A short note on the measurement system status
• The actual number, with context for what it means for the decision at hand

A clean format like this turns capability into a defensible answer instead of a number on a slide. Reporting process capability to plant managers in a way that supports decisions is its own skill, separate from running the study itself.

Key Takeaways

• Cp and Cpk use within-subgroup, short-term variation and assume the process is stable. Pp and Ppk use overall variation and make no stability assumption.
• Cp and Pp ignore centering. Cpk and Ppk penalize the process for being off-center between the spec limits.
• A wide Cpk-to-Ppk gap is a signal that the process is not as stable as the short-term number suggests.
• Use Cpk for ongoing production with a maintained control chart. Use Ppk for new processes, recent changes, and PPAP-style overall performance reporting.
• A capability index is only as trustworthy as the measurement system, the stability evidence, and the rational subgrouping behind it.