Process Capability Cp, Cpk, Pp, Ppk & Specification Fit

The within-subgroup sigma is estimated by the range method, Rbar / d2, which is the AIAG default and is accurate for subgroup sizes up to about ten; beyond that the pooled standard deviation or sbar / c4 is the better estimator and this tool will read slightly low. The overall sigma is taken as the sample standard deviation of the individual readings without a c4 bias correction, matching AIAG practice; the correction is under half a percent for a hundred readings and can be ignored at these sample sizes. Everything here assumes the characteristic is measured on a continuous scale and is at least roughly normal - capability indices computed on a skewed characteristic such as neps per gram or a bounded one such as shade difference are not comparable with those computed on a symmetric one, and ISO 22514-2 describes the non-normal distribution families to use instead. The indices say nothing about measurement error: if the gauge repeatability and reproducibility consumes a meaningful share of the tolerance, part of the spread being penalised here belongs to the test method rather than the process, and a measurement systems analysis should be run before any capability improvement project is launched. Finally, capability is only interpretable on a process in statistical control; a large capability gap is direct evidence that this precondition has failed, and the correct response is to find and remove the special cause rather than to report the index.