Start with specifications and variation
A manufactured feature rarely measures exactly the same way on every part. A filling machine may dispense slightly different amounts into successive bottles; a cutting process may produce pieces with slightly different lengths. Specifications define the acceptable range for a feature, usually with a lower specification limit (LSL) and an upper specification limit (USL). Measurements outside that range do not meet the stated requirement.
Cp and Cpk summarize how a process’s measurements compare with those limits. They are useful when a feature is measured on a continuous scale, such as length, weight or diameter. Neither number replaces looking at the measurements themselves. A capability calculation is most meaningful when the measurement method is dependable and the process is stable enough that its recent variation is a reasonable guide to its behavior.
Cp asks whether the process spread fits
Cp compares the width of the specification range with the process’s estimated short-term spread. Its formula is Cp = (USL − LSL) ÷ (6 × σ), where σ represents an estimate of within-process standard deviation. The six-standard-deviation span is a conventional way to describe process spread, particularly when measurements are approximately normally distributed.
A higher Cp means the estimated spread is narrower relative to the allowed range. But Cp does not consider where that spread sits between the limits. Imagine a machine that makes very consistent parts, all clustered close to the upper specification limit. Its Cp may look promising even though some parts exceed that limit. Cp answers a question about potential fit if the process is centered; it does not, by itself, show whether the process is centered.
Cpk accounts for the process location
Cpk checks the distance from the process average to each specification limit, then uses the smaller result. Its formula is Cpk = the smaller of (USL − μ) ÷ (3 × σ) and (μ − LSL) ÷ (3 × σ), where μ is the process average. The smaller distance matters because the nearer limit is generally the one the process is more likely to cross.
Suppose an acceptable length runs from 9.5 to 10.5 units. If the process average is 10.2 units and the estimated within-process standard deviation is 0.1 unit, Cp is 1.67: the specification width of 1.0 divided by a spread of 0.6. Cpk is 1.0 because the average is only 0.3 unit below the upper limit, or three estimated standard deviations away. The process is relatively consistent, but it is shifted toward one edge of the acceptable range.
Cpk cannot exceed Cp when both are calculated from the same limits, average and variation estimate. If they are close, the process is roughly centered between the limits. If Cp is much higher than Cpk, improving the process location may matter more than reducing its spread. If both are low, reducing variation may be necessary as well.
Check the process before trusting the index
Capability numbers can conceal important patterns. A process that drifts during a shift, changes after a tool replacement or alternates between two settings may produce a single average and standard deviation that poorly describe either condition. Review measurements over time and investigate unusual changes before treating one capability value as a dependable summary.
The way variation is estimated also matters. Cp and Cpk typically use a within-process estimate intended to reflect short-term variation. Pp and Ppk use an estimate based on overall variation in the collected data, which may include longer-term shifts. The two pairs can therefore tell different stories. Always check which measure a report presents, how the samples were collected and whether the measurement system can distinguish meaningful differences between parts.
These indices are easiest to interpret under assumptions such as an approximately normal distribution and a reasonably stable process. Skewed data, mixed batches or sparse samples call for closer examination rather than a confident conclusion from a single number. A reported Cpk should not be treated as a guaranteed reject rate.
Use the numbers to guide the next question
There is no universal Cp or Cpk value that makes every process acceptable. The appropriate target depends on the feature, its consequences, customer requirements and how the process is monitored. A useful review starts by confirming the specification limits, plotting the measurements and checking whether the process is stable. Then compare Cp with Cpk: a large gap points toward centering, while two similarly low values point toward spread.
Finally, connect the calculation to a practical action. That might mean adjusting a setting, addressing a source of variation, improving the measurement method or collecting data across more operating conditions. Cp and Cpk are most valuable as compact descriptions that prompt investigation—not as substitutes for understanding how the product is made.
