Election day operations
One day, no second chances, and every decision has to be made in minutes. The work that makes election day calm happens in the fortnight before it.
Two weeks out: build the day on paper
Lock the list. Decide now who you are contacting on the day. It is almost always your identified supporters – the people already scored as support on the five-point scale – not a persuasion universe. Election day is turnout, not persuasion.
Build that audience and freeze it. A segment that keeps recalculating on the day is a segment you cannot reconcile against.
Map the booths, then the shifts. Every booth needs a named person for every hour it is open. Put the shifts in the calendar as shifts, not as a spreadsheet, so volunteers see them where they already look and you can see the holes.
Name your holes early. The 6am and the 5pm slots are the ones that go unfilled. Fill them first, from the people who have shown up most.
One week out: rehearse the comms
Agree the escalation path. Who does a booth captain contact when something goes wrong, and who does that person contact? Write it down. On the day nobody invents a process.
Set up the broadcast. One-tap broadcast push reaches every volunteer's phone at once. Draft the messages you can predict now – a weather call, a booth-short call, the final-hour push – so you are editing rather than composing at 3pm.
Test the reply path. If you are running an SMS program on the day, the shared inbox needs organisers rostered on it exactly like a booth. Inbound peaks in the first two hours and the last two.
The day before
- Confirm every shift by direct message, individually. Group messages do not confirm anything.
- Check that every volunteer has the app installed and can open it. The morning of is too late.
- Print the fallback. Paper walk lists for the booths where you know the signal is bad. The app queues offline, but a phone with a flat battery does not.
- Get the materials to the booths, not to your house.
On the day
Run the action room, not your phone. The room refreshes active volunteers roughly every ten seconds. Watch coverage, not individuals – you are looking for a booth that has gone quiet, not for a person who is slow.
Make decisions on a clock. Set three checkpoints – mid-morning, midday, mid-afternoon – and at each one ask the same two questions: which booth is under-covered, and are we on pace? Move people between checkpoints, not continuously.
Broadcast for real things only. Three or four messages across the day. A channel that pings constantly stops being read by the hour it matters.
Record as you go. Turnout data recorded at 7pm from memory is fiction. The point of recording through the day is that the last two hours are targeted at people who have not yet voted.
The last two hours
This is the only part of the day where a change of plan is worth the disruption. Pull volunteers from covered booths to under-covered ones, push the final broadcast, and work the list of identified supporters who have not been marked as voted.
After the close
Do two things before anyone goes home:
- Thank people, by name, in the room. The single biggest predictor of whether someone volunteers next time.
- Export the data. Results export to CSV. Get the day's record out and stored while it is complete.
Then debrief within 48 hours, while it is still accurate rather than remembered.