What Is Process Capability Index?
Process capability index (Cpk) measures how effectively your process produces output within customer-defined specification limits. It combines two critical factors: the spread of your process (variation) and how well it's centered between the upper and lower limits.
Consider a food service operation that must maintain soup temperatures between 64–80 °C. If the average temperature drifts toward one boundary while variation increases, the process produces more units outside acceptable ranges. Cpk captures this dual problem in a single metric.
A Cpk value of 1.0 means your process barely fits within specification limits. A value of 1.33 or higher is considered capable—providing a safety margin that accounts for natural process shifts. Without adequate Cpk, even small changes in mean or standard deviation can push output beyond customer tolerance.
Process Capability Index Formulas
Four variations of process capability indices address different scenarios. Use Cp when your process is perfectly centered; use Cpk when centering is uncertain. The Taguchi index (Cpm) and combined index (Cpkm) penalize deviation from a target value rather than just specification limits.
Cp = (USL − LSL) ÷ (6 × σ)
Cpk = min[(USL − μ) ÷ (3 × σ), (μ − LSL) ÷ (3 × σ)]
Cpm = Cp ÷ √[1 + ((μ − T) ÷ σ)²]
Cpkm = Cpk ÷ √[1 + ((μ − T) ÷ σ)²]
USL— Upper specification limit (maximum acceptable value)LSL— Lower specification limit (minimum acceptable value)σ— Standard deviation of the processμ— Process mean (average)T— Target value (ideal output)
Interpreting Cpk Values
Cpk interpretation follows industry-accepted benchmarks:
- Cpk < 1.0: Process is not capable. Specification limits sit within ±3 sigma, meaning defects occur regularly.
- Cpk = 1.0–1.33: Process is marginally capable but vulnerable to centering shifts. Minor process deterioration causes failures.
- Cpk ≥ 1.33: Process is considered capable. The specification band extends beyond ±4 sigma, allowing for natural variation without defects.
- Cpk ≥ 1.67: Process demonstrates strong capability. Even with a 1.5-sigma mean shift, defect rates remain negligible.
The benchmark of 1.33 exists because manufacturing processes naturally drift approximately 1.5 standard deviations from center over time. A Cpk of 1.33 ensures that even with this drift, output remains within limits.
Common Pitfalls in Capability Analysis
Several mistakes can lead to misleading Cpk values and poor quality decisions.
- Ignoring Non-Normality — Cpk calculations assume normal distribution. If your process data is skewed (e.g., tool wear causing increasing defects), standard deviation misrepresents true variation. Collect larger samples and test for normality using statistical software before relying on Cpk alone.
- Using Outdated or Inconsistent Data — Cpk reflects only the data period analyzed. If measurement practices, equipment, or operators change mid-analysis, your index becomes unreliable. Ensure data spans a representative production window with consistent procedures.
- Confusing Specification Limits With Control Limits — Specification limits (±3-sigma range) define customer requirements, while control limits (±2-sigma range) detect process shifts. Cpk uses specification limits; mixing them up yields incorrect conclusions about true capability.
- Overlooking Centering When Cp Looks Good — A high Cp with low Cpk indicates poor centering. Your process variation fits the band, but it sits off-center. Adjust the mean before investing in variation reduction.
Practical Example
A beverage bottler fills containers to a target of 500 mL with specification limits of 495–505 mL. Historical data from 100 fills shows a mean of 500.3 mL and standard deviation of 1.2 mL.
Cpk calculation:
Cpk = min[(505 − 500.3) ÷ (3 × 1.2), (500.3 − 495) ÷ (3 × 1.2)]
Cpk = min[(4.7 ÷ 3.6), (5.3 ÷ 3.6)]
Cpk = min[1.31, 1.47] = 1.31
A Cpk of 1.31 sits just below the ideal 1.33 threshold. The process is marginally capable, with the upper limit (4.7 mL above mean) representing the constraint. Small increases in standard deviation or negative drift risk exceeding 505 mL and triggering customer complaints.