Check Data Trust before you quote a number

Draft — 2026-09-26

Drafted from the screen's code on main (kareIntelligence 7455702). Not yet reviewed against a live session. Roles listed are the defaults the app's role matrix (@collabkare/ability 0.12.0) gives this screen's route; your organization's role overrides can change them.

Overview → Data Trust shows the records behind every number in kareIntelligence: when the pipeline last ran, whether it reconciled, what it loaded and which versions computed the results. Open it before quoting a figure outside the app.

Cards

• Data freshness — the period end of the last run that passed reconciliation. A newer failed run never moves this date.
• Reconciliation — a run is "ok" only when the mart's counts match counts derived independently from the raw export (claim lines against exported EOB lines, members against exported patients, member-months against patients × months). A clean run says "Zero mismatches."
• Pipeline heartbeat — the last 10 runs, newest first, with the rows each run loaded and its origin (a run started by the schedule says so).
• Versions — the mart schema, the metric seed and the CQL content that produced the numbers.
• CQL evaluation — the background measure evaluations: how many measures have results and whether any latest result is degraded (degraded measures show a badge on their screens, never 0).
• Services — what the engine could reach when it built the page. An unreachable peer is amber, never red.
• Schedule — when the pipeline starts itself. It is read-only here: "Managed in Cloud Manager → Analytics Operations". Without one it reads "No schedule set — runs are kicked manually".

Empty state

A new organization sees "No pipeline runs yet — This is an onboarding state, not an error."