Wire sources

The outlet is the newsroom; the owner is the balance sheet. Consolidation, layoffs, syndicated national copy under a local masthead, and whether a statehouse bureau exists at all are decided one level above the outlet — so this is the chart the ownership question actually lives on. Sourced in editorial/wire-ownership.json, resolved per item date so a station changing hands does not retroactively rewrite the record.

Bucketed on each item's isoDate — the time the outlet itself stamped on the story. This is the newsroom's day: when assignment desks file, when the evening push lands, who is still publishing at midnight.

The same headlines bucketed on firstSeen: when the cron first saw them. Deliberately kept as a separate chart rather than merged with the one above — the difference between the two is our instrument, not Kansas City. Where this chart spikes and the publisher chart does not, that is our cadence or a feed catching up. This page used to draw only this one, captioned as if it were the outlets' behaviour.

Per outlet: the gap between an outlet's own timestamp and our first sight of the item. A high median usually means a shallow or slow feed rather than a slow newsroom; a high p90 against a low median means the feed occasionally strands items until they nearly fall out of the window. Both are measurements of our reach into that outlet, and both bound how fast anything downstream can be.

The ledger records what got published. This records how loudly each outlet was talking, including the polls that found a feed silent — which the ledger cannot represent at all, because an absence leaves no row. Feed depth is the average number of items visible in that outlet's window per poll; it is not a story count. New is the story count. Saturated counts polls where every slot we took was new, meaning our per-source cap may have cost us items behind it — a nonzero number on a busy outlet is the signal to raise its maxPerFetch.