Engagement ideas from data

The result is an updated ideas file that keeps three things apart: what the data shows, what might cause it, and what to change. Use docs/ideas.md by default. If the project already has a canonical list (docs/IDEAS.md, a root ideas.md), update that one rather than starting a competing list, and name its path in your answer.

1. Rebuild the context

• Read AGENTS.md or CLAUDE.md, the existing ideas and decisions, the game's core loop in its code, its event catalog (what each event means and when it fires), and the telemetry: a saved snapshot's README.md and manifest.json in analytics-data/, or the answers playgama-metrics-snapshot read in this conversation. No telemetry yet, or only an old snapshot: read it first with playgama-metrics-snapshot (it asks whether to save it).
• Match the answers to the build, its commit and when it went live. Flag a dirty or unknown build, mixed traffic, an incomplete last day and the hours not yet processed (verify lists them).
• The code and a local build help explain an event, but they don't prove that players saw that exact interface in the measured build.

2. Measure the player journey