Calculate Cpk from process mean, standard deviation, and specification limits. Evaluate whether your process meets customer requirements reliably.
The Process Capability Index (Cpk) measures how well a process fits within its specification limits, accounting for both the spread and the centering of the process. Unlike the simpler Cp index that only considers spread, Cpk penalizes processes that are off-center by comparing the process mean to the nearest specification limit.
Cpk is calculated as the minimum of two ratios: (USL − μ) / (3σ) and (μ − LSL) / (3σ). A Cpk of 1.0 means the process just barely fits within specs. A Cpk of 1.33 is typically the minimum acceptable value for existing processes, while 1.67 is required for new processes in many industries. A Cpk of 2.0 represents Six Sigma capability.
This calculator takes your process mean (μ), standard deviation (σ), upper specification limit (USL), and lower specification limit (LSL) to compute Cpk, Cpu, Cpl, and the estimated defect rate. Use it to validate process capability before production release and during ongoing quality monitoring.
Cpk answers the fundamental question: can this process consistently produce parts within specification? A high Cpk means fewer defects, less scrap, reduced inspection cost, and confident customer delivery. It is the standard capability metric required by automotive (PPAP), aerospace, and medical device customers. Consistent measurement creates a reliable baseline for tracking improvements over time and demonstrating return on investment for process optimization initiatives.
Cpu = (USL − μ) / (3σ) Cpl = (μ − LSL) / (3σ) Cpk = min(Cpu, Cpl) Estimated PPM = based on the limiting tail of the normal distribution
Result: Cpk = 1.25
Cpu = (10.5 − 10.05) / (3 × 0.12) = 0.45 / 0.36 = 1.25. Cpl = (10.05 − 9.5) / (3 × 0.12) = 0.55 / 0.36 = 1.53. Cpk = min(1.25, 1.53) = 1.25. The process is closer to the upper limit. At Cpk 1.25, approximately 88 PPM defects are expected.
Production Part Approval Process (PPAP) submissions in the automotive industry require capability studies demonstrating Cpk ≥ 1.33 for all critical and significant characteristics. During initial production, Ppk (preliminary capability) may be reported until enough data accumulates. Cpk studies use only within-subgroup variation, while Ppk uses overall variation.
Cpk can be improved by reducing variation (increasing σ), centering the process (moving μ toward the midpoint of specs), or widening specifications (when functionally possible). Process centering is often the quickest win because it requires only adjusting target settings, not reducing variability.
A product may have dozens of measured characteristics, each with its own Cpk. Report the worst-case Cpk to represent overall process capability. Use capability histograms showing the distribution of Cpk values across all characteristics to identify where improvement efforts should focus.
Cp measures only process spread relative to specification width, ignoring centering. Cpk also accounts for centering by looking at the distance from the mean to the nearest spec limit. Cpk ≤ Cp always. If Cpk = Cp, the process is perfectly centered.
A Cpk of 2.0 corresponds to 6σ capability (6 standard deviations between the mean and the nearest spec limit), yielding 3.4 DPMO with the 1.5σ shift. In practice, a Cpk of 1.5 is already considered very capable.
Yes. A negative Cpk means the process mean is outside the specification limits. This indicates a severely incapable process where the majority of output is out of spec.
At minimum 30 individual measurements. For reliable confidence intervals, collect 100+ data points. Use subgroup data from control charts for within-subgroup standard deviation to distinguish short-term capability from long-term performance.
Use only the applicable index: Cpu for upper-only specs, Cpl for lower-only specs. Cpk requires both limits. Many characteristics in manufacturing have bilateral specs, but one-sided specs are common for surface finish, concentricity, and hardness.
Yes. Standard Cpk assumes normally distributed process data. For non-normal data, use alternative methods like the Pearson or Johnson family of distributions, or transform the data before calculating Cpk.