The fields of statistics and econometrics have developed powerful methods for testing the validity (specification) of a model based on its fit to underlying data. Unlike statisticians, operations researchers and managers are typically more interested in the quality of decisions derived from a given model, rather than its statistical validity per se.

We propose a framework and a statistical test that captures this perspective. The details of the approach will be fleshed out in the context of a typical revenue management problem, where a key objective is to infer a demand model/relationship. An illustrative empirical example will be given, focusing on a data set used to optimize consumer loans. This work and its conclusions may provide further reasoning why simple, parsimonious, and potentially incorrect models could still be deemed ``good enough'' for operations purposes.

(*) joint work with Omar Besbes (Wharton) and Robert Philips (NomisSolutions)