Status
What the pipeline collected, what is overdue, and what looks wrong.
Last run
The pipeline runs once a day at 04:15 UTC. The last scheduled run started 27 Sept 2026, 04:16 UTC (13 h ago). A source that fails is isolated: the others still run, and yesterday's value stays in the score. 1 source failed on the last run. 1 source has missed its own publication cycle.
Sources
Two different things fail independently, so we show both: collection (did we reach the source last night?) and data vintage (what period does the source actually publish?). An annual source collected successfully every night still serves a value from last year — that is its normal regime, not an outage.
Open data-quality issues
Automated checks flag values outside physically plausible bounds and jumps beyond four standard deviations. They never block the pipeline — a human arbitrates. A flag is not a mistake: a genuine shock is supposed to trip the four-sigma test, and a series that predates the modern era is supposed to fall outside bounds calibrated for it.
Scores computed under methodology v2.6.0, last recomputed 27 Sept 2026, 04:20 UTC — the same version the deployed backend is running. How each score is built is on the methodology page, and every revision of the method is listed, newest first, under revision history.