Product Analytics

Product Analytics is the surface where the platform's growth is measured rather than its operations. The product-usage dashboard plots feature-adoption rates across tenants, per plan, and per country, sourced from `ApiUsageEvent` rolled up nightly. Funnels track the canonical signup → first match → active path both per tenant and aggregated, exposing the leak points that warrant feature work.

Retention cohorts produce weekly and monthly heatmaps sliced by signup source, plan, and country, so stickiness can be diagnosed by acquisition channel. Feature-flag experiment readouts reach into GrowthBook for variant assignments and combine them with conversion events to compute lift, confidence interval, and sample size per variant; the readout is the gate before flipping a flag from experiment to default. The custom-query playground is intentionally read-only and `sys_engineer`-gated: queries cap at 10 000 rows, include a syntax-validated MongoDB driver, and persist as `sys_saved_queries` so a useful one-off becomes a shared diagnostic.

The product-metric catalog is the most important piece of the surface even though it has no UI of its own: it stores executable definitions for DAU, WAU, MAU, stickiness, retention M1, active tenants, and signups, and every dashboard pulls from the catalog. Two dashboards quoting the same metric will always see the same number, because there is exactly one definition. New metrics land here first, then in dashboards.