Abstract: Formal models play a crucial role in advancing knowledge about brain function. However, there are many incoherencies and questions typically asked in the field: what are good model goals? Which modeling tools are appropriate for a given question? When can we compare models? How can we know if model choices are appropriate and justified? This highlights some of the complexity of the modeling enterprise and suggests that modeling should be regarded as a high-dimensional decision process. Unfortunately, there is currently a lack of structure, principles and guidelines regarding this decision process. Here, we propose an organization of the problem, model and outcome spaces by describing modeling as a decision process that links model space to model utility via a crucial action space, i.e. model selection. We then provide decision heuristics and guidelines for best practices to help the modeler in the decision process. We hope this will help both modelers and reader better appreciate different models as an exercise of fruitful complementarity rather than a battle over what’s the “right” model. Finally, we discuss a series of open problems in the field that we think need addressing in order to make progress in neuroscience research.