​
In the app

The analytics view shows:
  • Query volume — how many questions the twin is getting over time
  • Confidence distribution — the split of high / medium / low confidence
  • Escalation rate — what share of questions come back to you
  • Top sources — which material is doing the work
  • Channel breakdown — where questions arrive from: Slack, iMessage, WhatsApp, CLI, app

​
From the CLI

Who’s asking what, what escalated, where the documentation gaps are, and the pending escalation queue.

​
How to read it

A high escalation rate early is correct, not a failure. It means the twin is declining to guess. The number to watch is whether it falls as you answer things — that’s the loop working. A repeated escalation is a to-do item. The same question escalating three times is a gap you can close permanently by answering it once, properly. Confidence that’s uniformly high with thin sources is the warning sign worth chasing. Run bestmate trace on a few answers and check what actually grounded them.

trace & analytics

Digging into a single answer.