In this paper, we outline a framework for appropriate and proper usage of simulated data for biometric authentication research. Currently, there are no formal guide-lines concerning the use of simulated data in the biometric authentication literature. Some have suggested the usage of simulated or synthetic data while others have advised against it. Our position is that there is a place for simulation data in biometrics research but that such implementations need to meet certain requirements. To that end we describe conditions under which it is reasonable to use such data, as well as criteria for evaluating the appropriateness of a data generation methodology. This criteria is that models for generation of artificial data should be flexible, consistent and parsimonious. Along with justifying these criteria, we illustrate how simulated data might be used to evaluate a classifier.