How to Calculate Process Capability (Cp, Cpk) — With Real‑World Examples

Process capability indices Cp and Cpk are powerful, concise measures of how well a manufacturing or service process fits within specified limits. This guide explains what they mean, how to calculate them step by step, and shows two worked examples you can follow.

Figure: Measurement and inspection are central to estimating process capability.

Quick definitions

• Specification limits: Upper Specification Limit (USL) and Lower Specification Limit (LSL) — the customer or design boundaries.
• Process variation: The natural spread of your data, usually summarized by standard deviation (σ).
• Cp: A measure of potential capability — how wide the specification window is compared to process spread.
• Cpk: A measure of actual capability considering process centering — how close the process mean is to the specification midpoint.

Formulas

Standard formulas (for normally distributed data):

Cp = (USL - LSL) / (6 * σ)

Cpk = min( (USL - μ) / (3 * σ), (μ - LSL) / (3 * σ) )

Where μ is the process mean and σ is the process standard deviation. The divisor 6σ represents ±3σ, which covers ~99.73% of a normal distribution.

Step-by-step calculation

1. Collect a representative sample — use data gathered while the process is stable (statistical control).
2. Calculate mean (μ) — the average of your measurements.
3. Calculate standard deviation (σ) — use the sample standard deviation formula (usually s for sample data).
4. Note the specification limits — USL and LSL from design/customer.
5. Compute Cp and Cpk using the formulas above.
6. Interpret the results — higher values mean a more capable process. Typical rule of thumb:
   • Cpk < 1.0: process is not meeting specs — many defects expected.
   • 1.0 ≤ Cpk < 1.33: marginally capable.
   • Cpk ≥ 1.33: generally acceptable for many industries.
   • Cpk ≥ 1.67: often used for high-performance or critical processes.

Worked Example 1 — Manufacturing: Metal Shaft Diameter

Scenario: A factory produces metal shafts. The drawing specifies LSL = 19.90 mm and USL = 20.10 mm (target 20.00 mm). You collect 30 measurements from a stable run. The sample mean μ = 19.982 mm and sample standard deviation s = 0.028 mm.

Calculate Cp

First, compute the specification width:

USL - LSL = 20.10 - 19.90 = 0.20 mm

Now Cp:

Cp = 0.20 / (6 * 0.028) = 0.20 / 0.168 = 1.190

Calculate Cpk

(USL - μ) / (3σ) = (20.10 - 19.982) / (3 * 0.028) = 0.118 / 0.084 = 1.405
(μ - LSL) / (3σ) = (19.982 - 19.90) / 0.084 = 0.082 / 0.084 = 0.976
Cpk = min(1.405, 0.976) = 0.976

Interpretation: Cp = 1.19 suggests the process has the potential to meet specs (spread is relatively small compared to spec range). But Cpk = 0.976 (<1 .0="" 20.00="" and="" bias="" center="" closer="" fail="" improve="" is="" likely="" limit="" lower="" lsl="" mean="" meaning="" mm="" off-center="" on="" or="" p="" parts="" process="" reduce="" should="" shows="" side.="" team="" the="" to="" toward="" variation.="" will="">

Worked Example 2 — Service: Response Time SLA

Scenario: A support team promises a ticket triage response time between LSL = 0 min and USL = 30 min. Over 50 tickets, the mean response time μ = 22 min and sample standard deviation s = 6 min. The spec window is 0–30 but note LSL at 0 is one‑sided — Cpk still applies but interpret carefully.

Calculate Cp

USL - LSL = 30 - 0 = 30 min
Cp = 30 / (6 * 6) = 30 / 36 = 0.833

Calculate Cpk

(USL - μ) / (3σ) = (30 - 22) / (18) = 8 / 18 = 0.444
(μ - LSL) / (3σ) = (22 - 0) / 18 = 22 / 18 = 1.222
Cpk = min(0.444, 1.222) = 0.444

Interpretation: Cp < 1 indicates the process variation is larger than allowed by specifications — even if the mean were perfectly centered, the spread would cause many tickets to miss the 30‑minute SLA. Cpk is worse (0.444) because the process mean is closer to the USL (30 min) and many responses exceed the SLA. For service processes, reducing variation (e.g., standardized triage steps, automation) and shifting the mean down are remedies.

Practical tips & common pitfalls

• Use stable data: Cp/Cpk assume the process is in statistical control. If the process has special causes of variation (assignable causes), remove them first and re‑measure.
• Watch distribution shape: Cp/Cpk assume approximate normality. For strongly skewed data, consider transformations (log) or non‑parametric capability methods.
• Sample size matters: Small samples give unreliable σ estimates. Aim for 30+ points when possible — more is better.
• Understand Cp vs Cpk: Cp only measures spread; a high Cp with a low Cpk means the process is not centered.
• One‑sided specs: If you only have an upper or lower spec, the formulas still work but interpret carefully — sometimes Ppk or other metrics are used.
• Long‑term vs short‑term σ: Some organizations use short‑term (within subgroup) σ for Cp and a long‑term σ for Pp/Ppk. Be clear which σ you used and why.

Quick checklist to report capability

Item	Yes / No
Data collected from stable process
Sample size >= 30
Distribution roughly normal
Mean, standard deviation computed
Cp and Cpk computed and interpreted

Conclusion

Process capability indices Cp and Cpk are compact, actionable numbers that help you understand whether your process has the potential to meet specifications and whether it actually does so when centered. Remember: Cp shows potential (spread), while Cpk shows actual performance (spread + centering). Use representative, stable data, check distributional assumptions, and combine capability analysis with root cause work — reduce variation and center the process to improve capability.

If you want, I can generate a printable PDF of this article, or create a downloadable worksheet (Excel) that calculates Cp and Cpk when you put in your sample measurements. Tell me which format you prefer.