Internal · editorial operations

Content Quality Dashboard

The refresh-cadence view: what is overdue, what is missing trust primitives, and where the data is going stale. Generated Aug 1, 2026, 10:26 PM.

Status reviews — nothing overdue

Jurisdiction intel freshness — stalest first (0 outside SLA)

Showing the 20 stalest of 59 jurisdictions. Refresh via /update-states.

P2 structured state coverage — weakest first (45 complete)

This measures verified underlying rows across the 50 states and DC. A component shell does not earn credit. DC's RHTP lane is handled as explicitly ineligible; no award is implied.

Sources / jobs / pay shows current verified employer sources / live posting rows / rows with structured employer-posted compensation. Manual source reviews expire after 45 days and never re-stamp themselves. P2 requires all nine measured lanes; salary absence is reported, never imputed.

P3 jurisdiction living loops — weakest first (59 complete)

This measures all 59 state, DC, territory, and Compact nodes. Weekly delivery remains conditional on a new matching signal; a quiet week does not generate filler email.

The prebuild living-loop guard traverses a full 59-day spotlight cycle, resolves an exact state follow into a source-backed digest, and verifies the common plus dedicated CA/TX page controls.

Composite P4 readiness — every applicable gate joined

“Promotion-ready” means every applicable evidence component currently passes. It is not automatic P4 promotion: an editor must still make an explicit promotion decision. Full P4 applies to the 50 states, DC, and Puerto Rico. The seven TT jurisdictions use the honesty-first P4-analog contract, where WARN and the state-track anchor-profile gate are outside the contract rather than silently failed or invented.

P4 uses nine gates: P1, operating model, P3, current claim-level attribution, WARN source health, a current 12-value source-backed anchor profile, cadence in SLA, source diversity, and dated change-history monitoring. P4-analog uses the seven applicable gates and displays WARN/profile as N/A.

P4 refresh cadence — three clocks, no inferred sweeps

A source publication date, a hash-pinned claim review, and a completed jurisdiction source sweep are different facts. Only an explicit run with inspected-source evidence can set the sweep clock or derive the next deadline. Quiet weeks may pass as no-change; absent proof remains unknown.

P4 reviewed attribution gate — Tier 1 first

This is the audited denominator. Each decision is pinned to the claim id, source URL, and content hash. Organization and mixed claims count as eligible; genuinely systemwide claims do not. A jurisdiction must review every current claim and name organizations in at least 80% of eligible claims before this gate passes. Passing this gate is one P4 requirement, not full P4 promotion.

The queue lists the next five unreviewed claims per jurisdiction and prioritizes claims whose source/content review is already current. Eligible-but-unnamed claims remain visible as remediation failures; they are never silently reclassified as systemwide.

P4 claim-change accountability — 59 jurisdictions monitored

The committed manifest pins every current claim identity. An addition, wording or source change, move, or retirement fails closed until an explicit decision classifies it. The public log begins July 14, 2026; zero logged events never means the page had no earlier edits.

P4 anchor-employer evidence — one largest system per jurisdiction

Each jurisdiction has one deterministic anchor selected by reported patient volume, then site count. Actual values, source-backed values, reviewed absence, and unknown fields are counted separately. A reviewed-not-found field never earns profile-depth credit. The extended queue covers every anchor plus every organization reporting at least 50,000 patients.

P4 WARN source-health gate — silence is not zero

Source discovery, retrieval, and FQHC matching are separate. A manual, blocked, unavailable, research-needed, failed, or not-yet-run source always means the observation is unknown. Only checksum-pinned completed adapters can pass this component, and only human-confirmed directory matches can become public events or affect a score.

The official-row count describes the source artifact, not the number of FQHC layoffs. A zero in the confirmed-match column means no reviewed directory identity is published; it does not mean no event occurred.

Territory and Compact operating analog

These six lanes replace state-only placeholders with the governing financing, fiscal, licensure, referral/medevac, employer, and exclusion evidence for five territories and three freely associated states.