When looking for the best model, simulation is often used for "what-if" analysis. The properties of a modelled process are explored by repeatedly executing the model. Based on the outcomes, parts of the model may be manually modified to improve the process. This is iteratively done to obtain the model best suited to a user's requirements. In most cases, this is a labour-intensive and time-consuming task. To improve on the state of the art, we propose a framework where the user defines a space of possible process models with the use of a configurable process model. A configurable process model is a compact representation of a family of process models, i.e., a set of process models related to each other. Within the framework, different tools can be used to automatically compute characteristics of a model. We show that, when used on data from multiple real-life models, our framework can find better models in an automated way.