The election night monitoring checklist

Why election night breaks monitoring

Most breaking news is a surprise attended by a few outlets. Election night is the opposite: a scheduled story that every outlet in the country pushes at once, all night. Volume goes up tenfold, every push looks urgent, and the real signal (a race actually called, a concession actually given) is buried in an avalanche of live-blog updates saying nothing. The other special hazard is speed itself: on election night, being first is how newsrooms compete, and being first with someone else's wrong projection is how they end up in journalism reviews.

So the checklist has two goals: make the important push visible inside the noise, and make it structurally hard for your desk to mistake one outlet's speed for a settled fact. The examples below use the US midterms on 3 November 2026; the structure works for any country's election.

The week before: build the wall

1. The referee column. Electoral authorities, secretaries of state, the public broadcaster, the wire that makes the official-adjacent calls (AP in the US). Small column, few pushes, every one matters. Sound alert on.
2. The callers column. The five or six organisations whose projections move the night: networks, the wire, the big papers. Unfiltered, so you see every call the moment it goes to their readers.
3. The competitors column. The outlets you measure yourself against, unfiltered. On election night this answers the only workflow question that matters at 23:40: "have they pushed it yet?"
4. Race alerts. One word filter per decisive race or candidate, for instance ("Smith" OR "Ohio Senate") AND !poll, applied across all sources. Route these into a dedicated Slack channel (#election-calls) so the whole desk sees each match with source, time and permalink.
5. The consensus column. All sources, consensus set to 3+. This is the de-noised version of the night: only stories at least three independent newsrooms have pushed. When something appears here, it is real enough to act on.
6. The outside view. One column of foreign outlets covering your election, translated. How the night reads from abroad is a story your morning edition can own, and almost nobody watches it live. (The reverse workflow, for covering someone else's election, is its own guide.)
7. Dry run. Two days before: sound alerts audible in the room, Slack routing tested, and one person who knows how to edit a word filter at midnight without asking.

On the night: rules of engagement

• Watch the consensus column, work from the race channels. The unfiltered columns are for answering questions, not for staring at. If your eyes live anywhere, make it the 3+ column, where the night's actual events surface one by one.
• Every claim gets a timestamp. When you cite another outlet's call, the notification's permalink records who pushed what, when, to the second. Screenshot nothing; link the record.
• Distinguish the three kinds of push. Live-blog pulses ("polls close in ten minutes") are rhythm. Results pushes ("Smith leads with 40% counted") are data. Calls ("networks project Smith wins") are news. Only the third kind changes what your desk does.
• Let the quiet columns speak. If the referee column is silent while the callers column is loud, that gap is itself information: projections are running ahead of the count.

Handling calls and projections

The oldest election-night failure is treating one fast outlet as the result. The discipline worth writing on the wall:

• One call is that outlet's claim. Report it as theirs, attributed, if you report it at all.
• Independent agreement is news. When several organisations that do their own decision-desk work have called the same race, the ground has actually moved. This is exactly what a consensus threshold automates: an alert that fires only when three or more newsrooms have pushed the same story cannot, by construction, be one desk's itchy trigger finger.
• Concessions and official results outrank everything. The referee column ends arguments.
• Write your fallback language in advance. "AP has called X; other networks have not yet followed" is a sentence to draft on Tuesday afternoon, not at 02:00.

The morning after: the retro

Election night is also the best measurement night of the cycle, because everyone pushes the same events. The permalinked record answers, precisely: which outlet called each race first, by how many minutes; how long your own desk took from first call to your push; which of your alerts fired late or not at all. Across a year of data the median gap between first and second outlet on a story is 16 minutes; on election night the gaps compress to seconds, which makes it the sharpest benchmark you will get. Run the retro while the timestamps are fresh, and fix the source list the same week, while everyone still remembers which column was useless.

The one-page checklist