The comparison of performance between training organisations often stumbles over the fact that the different groups of beneficiaries do not have the same difficulties. The question is how to ensure one is comparing like with like. A few years ago an evaluation team had to compare 133 school districts in different states in the U.S. The team decided to compare each school district to a "benchmark group" of 14 other school districts with similar pupil characteristics (poverty; home language; parents' level of education; belonging to a minority group; etc.). To form the standard groups, the evaluation combined a factor analysis and expert panels. At the end of the evaluation, the validity of the performance comparisons were accepted without question.

Henry et al, (1992)