Confidence

P5-attribution Bayesian confidence dashboard. Per-tool Discovery (probability of being found) and Use (probability of being reused within 7 days) are estimated with a Beta(1,1) prior over a 30-day trailing window, and reported alongside 95% credible intervals.

Headline — overall Discovery / Use

Trial-weighted aggregate. Targets: Discovery 90% / Use 80% by T+90d, 95% / 92% by Y1.

Discovery (weighted)

…

Target T+90d 90% / Y1 95%

Use (weighted)

…

Target T+90d 80% / Y1 92%

Trials (30 days)

…

Posterior credible width depends on trial count

Per-tool — Discovery / Use by tool

Posterior mean and 95% credible interval for each tool. Tools with few trials show wider CIs (greater uncertainty).

Tool	Discovery	95% CI	D hits / trials	Use	95% CI	U hits / trials
Loading…

Trend — daily confidence (from T+0)

Past snapshots (analytics/confidence_<DATE>.json) joined as a time series. The faint band is the 95% credible interval. The public dashboard reads from a static asset in the AutonoMath public repository.

Discovery Use Discovery 95% CI Use 95% CI

Cadence: one cron snapshot per day (scripts/cron/confidence_update.py) + 5-minute in-memory cache. Source: GET /v1/stats/confidence (live) · /analytics/confidence_index.json (history). Methodology: confidence_methodology.md. Last generated: —