Modern maintenance strategies increasingly focus on vast amounts of diverse data and multifaceted analytical approaches in order to make efficient use of given resources and unveil hidden potentials. While there is often no universal solution approach to a specific case at hand, it is still possible to observe recurring problem classes for which generic solution templates can be applied and thus the establishment of a reusable knowledge base appears beneficial. To this end, we apply a taxonomy development approach to identify and systematize dimensions and characteristics of recurring data analysis problems in data-driven maintenance scenarios. Our research method integrates findings from a systematic literature review and expert interviews with data scientists from industry. Thus, we add descriptive theory to the field of maintenance analytics and propose a taxonomy that distinguishes between analytical maintenance objectives, data characteristics and analytical techniques.