The TUM Campus App helps students manage their everyday lives by providing valuable information about their grades, canteen menus, course-related data, and more. To further increase the usability and the usefulness of the app, the next milestone is to predict what kinds of data a given user might be interested in at any given point in time. Given their adoption inside many web applications nowadays, recommender systems are gaining popularity, as they can help guide users in their choices. We want to build a system that provides context-sensitive recommendations inside the TUM Campus App. A smart widget should display the data that are likely relevant to the user. This feature requires us to get an understanding of their usage pattern. We can achieve this by collecting usage data while the user is using the app, including time and location data, and the type of content they are viewing inside the app. We can then analyse the gathered data by implementing an algorithm that forms the basis for the suggested data inside the smart stack widget. We use hypothesis tests as well as user feedback to further assess the quality of the recommendations from a theoretical, as well as from a subjective standpoint.