Tracking in XOM means keeping a live, synchronised record of state as it changes — who is watching, what a user has earned, how a stream is performing, and what an operator did. Visual tracking of physical objects through a camera is a different capability and is not implemented in the reviewed codebase.
01Live session tracking
The broadcast and live stack carries the platform's most time-sensitive tracking. Dedicated endpoints exist for viewer tracking against a stream identifier and for stream status, and the viewer-side component set includes a live viewer count and a stream health readout. On the broadcaster's side, the console devotes a full analytics tab to live viewer counts, engagement metrics and stream health, with a real-time analytics dashboard defined in the subsystem specification for monitoring stream health during a broadcast.
Presence and other live state are held in Realtime Database rather than the document store, which is the split the architecture defines: Firestore for documents, Realtime Database for presence and live features.
02Progression and consumption tracking
The gamification layer tracks user activity and converts it into experience points and level progression, with an achievements module that separates in-progress tracking from completed unlocks, and leaderboards that rank users across experience, contributions, streaks and vertical-specific metrics. The quest engine adds waypoint tracking and a database sync module that persists quest state between sessions.
On the account side, the specification defines a consumption tracker behind a usage analytics view, alongside subscription tier control, add-on entitlement, and invoice history.
03Operational and safety tracking
Administrative tracking is specified separately from user-facing metrics. The admin control specification defines a system oversight dashboard with an AI health analyser, a security event monitor for reviewing security logs, and a quota management system for configuring limits. The admin panel adds system logs with an audit trail, and moderator actions are recorded with automatic audit stamps.
Behavioural tracking also feeds safety: the AI layer includes a fraud detection subsystem doing behavioural analysis for fake accounts and manipulation, and session tokens carry an anti-abuse risk score used for adaptive moderation.
04What is not tracked
There is no computer-vision pipeline, no camera capture path, and no pose or marker tracking in the code reviewed. Content that carries three-dimensional assets does so by embedding an external viewer from a URL stored on the post record, which means any tracking of that scene belongs to the embedded viewer, not to XOM.