We carried out an empirical investigation among producers of packaging machines collecting information about their rivals, i.e., those few competitors which they closely monitor. We found interesting regularities that take geographically clustered firms apart from isolated firms, that existing theories left unexplained. By constructing an agent-based model we were able to formulate a simple and plausible heuristic for rival selection which is able to generate the empirically observed facts. We submit that this case is exemplary in showing what agent-based models can do, namely, providing sufficiency proofs that help theory-building.