Occupant behavior is nowadays acknowledged as a main source of discrepancy between predicted and actual building performance; therefore, researchers attempt to model occupants' presence and adaptive actions more realistically. Literature shows a proliferation of increasingly complex, data-based models that well fit the cases analyzed. However, the actual use of these models by practitioners is very limited. Moreover, simpler models might be preferable, depending on the aim of investigation. The present study proposes shifting the focus to fit-for-purpose modeling, in which the most appropriate model for a specific case is characterized by the lowest complexity, while preserving its validity with respect to the aim of the simulation. A number of steps are taken to achieve this shift in focus. The existing models are presented according to complexity. The available inter-comparison studies are critically reviewed. Subsequently, a list of parameters that affect the choice of an appropriate modeling strategy is presented as a first attempt to derive guidelines and generate a framework for investigation. To support such claims the effect of some of the listed parameters is evaluated in a case study. The main conclusion to be drawn is that determining the best complexity for occupant behavior modeling is strongly case specific. (C) 2016 Elsevier B.V. All rights reserved.