Nowadays, mobile apps are used for nearly every situation: for planning the day, communicating with colleagues, ordering goods, or entertaining and socializing. To understand users expectations in each situation and to provide context-aware services, researchers and app vendors started to capture users' interaction with the smartphone and to model user's behavior. This paper reports on a behavioral study based on app usage data logged over one year and the corresponding apps descriptions from the app store. Using Topic Modeling and clustering techniques, we segmented the usage data into meaningful clusters that correspond to different ``states{''}, in which users normally use their smartphone, e.g. socializing or consuming media. Researchers and app-vendors can use the insights from our work to improve their contextual recommendation techniques and the overall usage experience.