A structural model comparison for finding the best performing models in a collection

An improvement or redesign of a process often starts by modifying the model supporting the process. Analysis techniques, like simulation, can be used to evaluate alternatives. However, even a small number of design choices may lead to an explosion of models that need to be explored to find the optimal models for said process. If the exploration depends on simulation, it often becomes infeasible to simulate every model. Therefore, we define a notion of monotonicity to reduce the number of models required to be simulated whilst guaranteeing that the optimal models are found. We define and prove our monotonicity for throughput time. Furthermore, within our experimental evaluation we obtain very promising results in terms of running time because fewer models need to be simulated.