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

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

August 15, 2026

A control chart tells you whether a process is stable. It does not tell you whether the process is good enough for the customer. A machine can run steadily all day and still make parts that are slightly too big. A team can resolve tickets at a consistent pace and still miss the SLA.

Process capability analysis answers that second question. It compares how much your process varies with how much the customer allows, and turns the answer into a single number. This guide explains Cp, Cpk, Pp and Ppk in plain language, with formulas, a fully worked example and clear benchmarks.

Quick Answer

Process capability measures whether a process can consistently meet its specification limits. Cp compares the specification width with the process spread; Cpk also accounts for how well the process is centred. Pp and Ppk use the same formulas but with overall (long-term) variation instead of short-term variation. A Cpk of at least 1.33 is a common minimum target, and 1.67 or higher is often required for critical characteristics.

In this guide

What Process Capability Means

Think of parking a car in a garage. The garage width is the specification: what the customer allows. The car width plus the natural wobble in how you park is the process spread. If the car is much narrower than the garage and you park in the middle, you will never scrape the walls. If the car barely fits, or you always park close to one side, you will.

Capability indices put numbers on that picture:

  • USL and LSL: upper and lower specification limits, set by the customer or design
  • μ (mean): where the process is centred
  • σ (standard deviation): how much the process spreads. Six standard deviations (±3σ) covers about 99.73% of output for normally distributed data.

Capability analysis is a core Measure phase tool in DMAIC, used to set the baseline. It is repeated in the Control phase to prove the improvement worked.

Three Checks Before You Calculate

Capability numbers are only meaningful if three conditions are met. Skipping them is the most common reason capability studies mislead.

  1. The measurement system is reliable. If your gauge or data source adds its own variation, the process will look worse than it is. Run a measurement system analysis first.
  2. The process is stable. Plot the data on a control chart. If there are special causes, capability numbers describe a moving target and cannot predict anything.
  3. The data is roughly normal. The standard formulas assume a bell-shaped distribution. Check with a histogram or a normality test. Skewed data, such as waiting times, needs a transformation or a non-normal capability method.

Use enough data, too: a common guideline is at least 25 subgroups, or around 100 individual readings, collected over a representative period.

Cp and Cpk Explained

Cp: could the process fit?

Cp = (USL − LSL) ÷ 6σwithin

Cp compares the width of the specification with the width of the process. It ignores where the process is centred, so it shows the potential capability: how good the process could be if it were perfectly centred. Cp = 1 means the process spread exactly fills the specification. Cp = 2 means the specification is twice as wide as the process spread.

Cpk: does the process actually fit?

Cpk = minimum of [(USL − μ) ÷ 3σwithin, (μ − LSL) ÷ 3σwithin]

Cpk measures the distance from the process mean to the nearest specification limit. If the process drifts towards one limit, Cpk falls even if the spread stays the same. Cpk can never be higher than Cp. If they are equal, the process is perfectly centred.

σwithin is short-term variation, estimated from within subgroups (for example, the average range of samples taken close together). It reflects what the process can do in a short window without drift.

Pp and Ppk Explained

Pp and Ppk use exactly the same formulas, but replace σwithin with σoverall: the standard deviation of all the data together. That includes shifts between shifts, batches, operators and days.

Index Variation used Considers centring? Answers
Cp Short-term (within) No Could the process fit if centred?
Cpk Short-term (within) Yes Does it fit right now, in the short term?
Pp Long-term (overall) No Could it fit over time if centred?
Ppk Long-term (overall) Yes What is the customer actually receiving over time?

The gap is the useful part. If Ppk is much lower than Cpk, the process is good in the short term but drifts or shifts over time. That points you towards causes like tool wear, shift changes, different batches of material, or settings that change between days.

Worked Example: All Four Indices

This is an illustrative example to show the method. A machining line produces shafts with a diameter specification of 25.00 ± 0.05 mm. The process is stable on its control chart and the data is roughly normal. A capability study gives:

Lower specification limit (LSL) 24.95 mm
Upper specification limit (USL) 25.05 mm
Process mean (μ) 25.01 mm
Short-term standard deviation (σwithin) 0.010 mm
Overall standard deviation (σoverall) 0.013 mm
LSL 24.95USL 25.05Target 25.00Mean 25.010.04 mm to USL (nearest limit)0.06 mm to LSL24.9524.97525.0025.02525.05
Illustrative: the process spread fits well inside the specification, but it is shifted 0.01 mm towards the upper limit.

The calculations

Index Calculation Result
Cp 0.10 ÷ (6 × 0.010) = 0.10 ÷ 0.060 1.67
Cpk min(0.04 ÷ 0.030, 0.06 ÷ 0.030) = min(1.33, 2.00) 1.33
Pp 0.10 ÷ (6 × 0.013) = 0.10 ÷ 0.078 1.28
Ppk min(0.04 ÷ 0.039, 0.06 ÷ 0.039) = min(1.03, 1.54) 1.03

What the numbers tell the team

  • Cp 1.67 vs Cpk 1.33: the process spread is good, but it is off-centre by 0.01 mm. Simply re-centring it at 25.00 mm would lift Cpk to 1.67 without any other change.
  • Cpk 1.33 vs Ppk 1.03: short-term performance is acceptable, but over weeks the process varies more. Something changes between subgroups, perhaps tool wear between changes or differences between shifts.
  • Ppk 1.03: this is what the customer actually experiences. It is too low for most automotive customers.

Action plan: adjust the machine offset to centre on 25.00 mm, then investigate the long-term drift. Comparing data by shift and by tool life on a control chart is the natural next step, and a fishbone diagram helps list possible sources of drift.

What Is a Good Cpk Value?

Cpk Approx. sigma level Approx. out-of-spec (centred, two-sided) Typical verdict
Below 1.00 Below 3σ More than 0.27% Not capable
1.00 3σ About 2,700 ppm Barely capable
1.33 4σ About 63 ppm Common minimum target
1.67 5σ Under 1 ppm Often required for critical characteristics
2.00 6σ Practically zero World class

Out-of-spec figures assume a perfectly centred, normally distributed process with no long-term shift. Real processes drift, which is why the familiar Six Sigma figure of 3.4 defects per million assumes a 1.5σ long-term shift.

Your customer's requirement overrides general rules. Automotive and aerospace customers usually specify their own minimum Cpk or Ppk for each characteristic. To convert defect rates into DPMO and sigma levels, see our guide on calculating DPMO.

How to Improve a Low Cpk

Compare Cp with Cpk first. The gap tells you which problem to fix.

Pattern Diagnosis What to do
Cp high, Cpk low Off-centre Adjust the process mean towards the target. Often quick and cheap.
Cp and Cpk both low Too much variation Reduce variation: root cause analysis, better methods, maintenance, mistake-proofing. See how Six Sigma reduces variation.
Cpk good, Ppk much lower Drift or shifts over time Find what changes between shifts, batches or days. Use control charts to catch drift early.
All indices low despite efforts Process or design limit Consider new equipment, a design change, or reviewing whether the specification is realistic with the customer.

Use FMEA to protect the characteristics with the lowest capability, and keep monitoring them with statistical process control.

Process Capability Outside Manufacturing

Capability works for any measurable output with a requirement. Some examples:

  • IT support: resolution time against an SLA of 8 hours (an upper limit only)
  • Banking: loan processing time against a promised turnaround
  • Healthcare: lab report turnaround against a clinical target
  • Food and pharma: fill weight against declared quantity

When there is only one limit, as with an SLA, use the one-sided index: Cpu = (USL − μ) ÷ 3σ, or Cpl = (μ − LSL) ÷ 3σ. Time-based data is often skewed, so check normality carefully before trusting the number.

Common Process Capability Mistakes

  • Calculating capability on an unstable process. Fix special causes first.
  • Ignoring non-normal data. Standard formulas can overstate or understate capability badly.
  • Using too little data. A few readings from one good day will flatter the process.
  • Reporting Cp alone. Cp ignores centring, so a process can have a high Cp and still make defects.
  • Mixing up Cpk and Ppk. Customers often ask for Ppk at launch and Cpk for ongoing control. Confirm which one they need.
  • Skipping MSA. A poor gauge makes a good process look incapable.

Master capability analysis with hands-on practice

Our Lean Six Sigma Green Belt programme teaches Cp, Cpk, Pp and Ppk with real datasets in Excel and Minitab. Black Belt covers non-normal capability, transformations and advanced statistical analysis.

Green Belt Course Black Belt Course

New to Six Sigma? Start with what Six Sigma is, or read the Green Belt certification guide for India.

Frequently Asked Questions

What is the difference between Cp and Cpk?

Cp compares the specification width with the process spread and ignores centring. Cpk measures the distance from the process mean to the nearest specification limit, so it falls when the process drifts off-centre. Cpk is never higher than Cp.

What is the difference between Cpk and Ppk?

They use the same formula, but Cpk uses short-term variation within subgroups while Ppk uses the overall variation of all data. Ppk shows what the customer actually receives over time.

What is a good Cpk value?

A Cpk of 1.33 is a common minimum target, 1.67 is often required for critical characteristics, and 2.0 corresponds to Six Sigma performance. Below 1.0, the process is not capable.

Can Cpk be negative?

Yes. A negative Cpk means the process mean is outside the specification limits, so most of the output is out of spec.

Does the process need to be stable before calculating Cpk?

Yes. Capability indices predict future performance only for a stable process. Check stability on a control chart first and remove any special causes.

Can process capability be used in service industries?

Yes, for any measurable output with a requirement, such as resolution time against an SLA. Service data is often skewed, so check normality and use one-sided indices when there is only one limit.

Leave a Comment

No comments yet. Be the first to comment.