Control plane
Analytics — actionable hotspots
Pick a store, choose a window, then act on what the data shows. The policy-gap summary leads with the top three rules driving escalations and refunds; the policy-gaps table below carries the full per-rule breakdown. Top-products flags the topics that drive the most refund + return pressure. All panels are admin-only — non-admin signers see a friendly empty state instead of an error so the rest of the page stays usable.
All panels below require admin access.
Ticket volume · admin

Daily ticket volume — channel × status

Conversations bucketed by their openedAt day over the selected window — broken down by channel (Shopify · Amazon · eBay) and resolution status (open · resolved · escalated).

Pick a store to inspect ticket volume trends.

Policy gap summary · admin

Top 3 rules driving escalations

One row per rule, scored by escalations × 3 + denies across the selected window. Pick a row to jump into the policy editor pre-focused on that rule — the drill-down table below keeps the full per-rule breakdown.

Pick a store to see which rules drive the most escalations.

Policy gaps · admin

Which rules are tripping the most

One row per rule. Ranked by rejection rate × total evaluations across the selected window. The customer messages sitting behind every deny are right below the row — read the exact phrasing that tripped the rule.

Pick a store to see which rules deny the most.

Top refund/return hotspots · admin

Where refund + return pressure stacks up

Resolved tickets grouped by topic, ranked by refund + return intensity over the selected window. Topic is the SKU stand-in — the data model has no product table today.

Pick a store to compare refund/return hotspots.

About these aggregates
Policy-gap summary ranks each rule by `escalations × 3 + denies`. Policy gaps rank each rule by `rejection_rate × total_evaluations`. Top products rank by `(refunds + returns) × 5 + tickets`. All three endpoints share the same window picker and can be reset with the Refresh button after picking a new store.
Need the broader digest trend (week-over-week, defect-signal pairs, cron-pipeline history)? The dashboard analytics view covers that side of the operator story.