Strategies in Modeling: A Personnel Scheduling Example

The process of selecting the best alternative mathematical representation when several competing models are available has received scant attention in Management Science. Instead of systematically evaluating alternatives as a well-posed decision problem, the model builder usually selects the appropriate tool, i.e., model, by judgmental assessment. Yet the lack of successful modeling implementations illustrates how little we know about this selection process. This paper is an initial attempt at developing a workable procedure for comparing models. There are five crucial ingredients: 1 performance, 2 realism/complexity, 3 information requirements, 4 user friendliness, and 5 computational costs. A successful application of this procedure in the context of a large faculty scheduling problem is described.