where USL and LSL are the upper and lower specification limits, respectively. Note that Cpp is a normalized version of
the mean squared error; that is, Cpp is equal to the biassquared plus variance and then divided by a normalizing constant D2
. If m and s are unknown, then Cpp may be estimated from a sample of size n taken out of the process. The unbiased and consistent point estimator for Cpp, proposed by Greenwich and Jahr-Schaffrath [5], is