Explore your own questions
Explore groups events, players, sessions, events per player or session, and numeric property sums or averages by UTC day or hour, event name, platform, build, level, country, device model, OS version, GPU or a property. Grouping by country or a device field keeps the 50 biggest values for the whole window and says how many there were. Add a second grouping and up to 20 event names. Hourly queries cover at most 14 days. A second grouping keeps its eight biggest values for the whole window, and the answer says how many there were. Results are capped at 1,000 rows; the page and CSV explicitly mark truncation. Missing numeric values remain empty. Narrow your query before interpreting a partial result.
Enter event names to also see a feature engagement matrix. Breadth is the fraction of all active players in the selected window who used that event. Frequency is the mean event count among those players; per-active-player frequency includes everyone in the window. This is descriptive usage data, not evidence of causation.
Dashboard reads are included. CSV download uses the applied query and is capped at 256 KB, with spreadsheet formula escaping and a truncation note. The game-scoped API endpoints are /analytics/explore, /analytics/explore.csv and /analytics/engagement. HTTP queries also accept up to five prop.key=value filters and exact platform, gameVersion, country, deviceModel, osVersion and gpu filters.
Overview and Monetization show players, sessions, paying players and payer conversion by country, device model, OS version or GPU, from /analytics/dimensions/{dimension} (20 values by default, 50 at most, the rest as other). Player segments can name a country code and any of the three device fields, and every read that takes a segment honours them. Experiments and config rules still target platform and build only, because the rest is not known when an SDK fetches its config.
MCP explore and get_engagement_matrix, and REST /api/v1/data/games/{gameId}/explore and /engagement, use the standard programmatic rate and existing token scope. Group a property with groupBy=prop:difficulty; property sum and mean require propKey. These tools, along with realtime and anomaly reads, are available through MCP and REST rather than the hosted analyst toolset to preserve its request budget. AI playtest events remain excluded unless explicitly included.