Case-note domain (Debt, Housing, Benefits, Employment, Family, Other) by day.

Not to be confused with "Service Demand", which counts anonymous triage-bot conversations, not advisor-handled cases.

AI-assisted completion time (recording finished → case note finalized) compared against NCAB's own pre-AI baseline. Nothing here is estimated — the baseline is whatever figure NCAB enters below, and the AI-assisted average only counts cases actually completed through this system.

Each finalized case note is automatically graded (0-3) against its category's real Casebook checklist — one point per section (Background/Notes, Info Sources, Next Steps) that's adequately covered. Only case notes saved since this feature shipped have a score.

% of queries where the knowledge base returned a confidently-relevant result (vs. a low-scoring or zero-result "coverage gap") — the same threshold the Retrieval Quality admin view uses.

A candidate signal for an emerging local issue. Method and thresholds are configured server-side (env config, no code change needed) and shown per card. Bar = this week's volume; the tick marks the trailing baseline.

Query volume heatmap — darker/brighter cells are busier.

Classified topic of incoming chat/triage queries by day (broader than case-note domain).

By outward postcode district only (e.g. "NR1") — combines the public triage flow's optional postcode question and the Case Record intake's optional postcode field. No full postcode or address is ever captured or shown here — district-level shading only, never a precise location.

Areas below the small-cell threshold, or with no cases on file, are shown with a muted fill and no category detail on hover — never as a specific count below the threshold. See the module notes in api/routes/leadership_analytics.py for why.