Analytics & Utilization (PL-T199)

Behind the dashboard is an AnalyticsRollup model that pre-aggregates bookings, revenue, no-shows, customer cohorts, and funnel events into day / week / month buckets with idempotent upserts so re-runs are safe. Three cron jobs maintain it: analytics_rollup_hourly recomputes the current hour's bucket, _daily closes the previous day at midnight, and _monthly_close finalizes the previous month on the 1st. On read, the dashboard's compute_rollup function adds an in-memory live-delta on top of the latest hourly rollup so today's KPIs include the last few minutes' activity without waiting for the next cron run.

The endpoints — GET /chain/analytics/dashboard / utilization / revenue / customers / conversion / heatmap and POST /chain/analytics/export — are gated by chain.analytics.read and chain.analytics.export capabilities so a viewer role can browse but not export. The dashboard surfaces KPIs (revenue, AOV, no_show_rate, NPS, utilization) plus drill-down views: AnalyticsUtilization with per-resource occupancy, AnalyticsRevenue with group_by=resource_type or color_type, AnalyticsHeatmap for the 7×24 ISO-week peak-time grid, ConversionFunnel reading BookingFunnelEvent rows (unique-session-deduped) for visit → wizard → submit → paid drop-off, and CustomerCohortChart for repeat-vs-first-time analysis. Exports under 92 days run synchronously; longer ranges queue to a background job and email the operator when the file is ready.

The PDF export uses a templated monthly-report layout that managers can attach straight to a board pack.