Abstract: A method for identifying and characterizing shape variations in parts. Measurements are taken of a population of parts and the shapes of the parts are expressed as a function having a plurality of coefficients. Discrete or function error maps are developed from the coefficients and a principal components analysis is performed on the error maps to identify the principal components of variation of the parts. The parts may be grouped into sub-populations representing ranges of variation along each of the principal components of variation, and downstream processes may be controlled differently for each sub-population. In one embodiment, a typical (re-generated) part shape is identified along multiple principal components of variation, and a tool path is controlled to be responsive to the typical part shape. Information regarding the principal components of variation may further be used to revise upstream manufacturing processes to advantageously affect the distribution of error in subsequently manufactured parts.