Currently we store the derived distributions and time series (install, uninstall, and sponsors error codes, countries, OSs, etc.) in files and in relational tables. It is much more convenient and efficient to manage that kind of data when they are stored in a JSON-based DB (a document versus a table or file). The goal is to create corresponding schemas, migrate the data, and start loading new data into the new store. On the reporting side, start using the data. This work could be done in incremental steps without disruption of the reporting services.