Analytics Infrastructure in Two Weeks

What the talk is about

Oleksandr Chernin explains how to build an in-house analytics infrastructure in about two weeks on top of third-party analytics platforms, mainly for casual and hyper-casual games. Instead of manual exports from several services, scripts in Python or R pull data through APIs into a single database, clean it and feed visualization tools. He recommends keeping five basic fields in every event, including event date, install date and user ID, which simplifies cohort analysis and joining events. He also covers data processing: dropping faulty records while logging errors, filling missing fields from other tables, and keeping a reference calendar of events. By his estimate it costs around 1,000 dollars a month, mostly for servers. Useful for companies starting with analytics and for analysts.