GOTV: get out the vote

Months of canvassing produce a list of people who said they will vote for you. On voting days that list is the whole campaign. Municipal turnout is often 30% or lower, so every supporter who forgets, gets busy, or assumes it is in the bag is a vote you already earned and then lost. Get out the vote (GOTV) is the work of making sure each identified supporter actually casts a ballot — a knock or a call to remind them, tell them where and when, and offer a ride if they need one.

It also solves a practical problem: on election day you will have more volunteer energy than you have ever had, and no doors left to persuade. GOTV gives every volunteer a list, a map, and a job that directly converts into votes.

Open GOTV in admin. DoorLine builds the target list from your canvass data: for every door (or unit), it takes the most recent answered response. If that support level is one you chose and the person can vote here, the door is a target. A later knock with no answer does not remove them. A later conversation where they are no longer in those levels takes them off, unless a volunteer has already worked that door. The name and phone from the response come with it.

Choose the levels with the three buttons: 5s only (strong supporters, smallest list), 4s and 5s (the usual GOTV universe, the default), or 3s, 4s and 5s (adds undecided-leaning doors — bigger list, more persuasion needed). Tick Signs, Donate, or Volunteer to add those doors as well. A door only needs to match one. Signs means a residential sign that is up, requested, or damaged. Public sign locations stay off, and a yard sign with no matched door cannot be knocked. Hit Refresh from canvass data after a canvass night to pull in new supporters.

Each poll in a ward votes at a different place, so a GOTV team needs to know which poll every door is in and where that poll votes. The clerk's voters' list has that: a poll number beside every address. Open GOTV → Polls, voting places and stickers and load the .xlsx or .csv the clerk sent you. The whole thing takes a minute. Here is what happens to the file, step by step.

1. The file is read in your browser, and the headers decide what is kept

Nothing is uploaded when you pick the file. Your browser reads it, strips any invisible watermark characters some clerks embed in every cell, and looks at the header row. It guesses the address column and the poll column from the headers and marks every other column as discarded, flagging the ones that look personal (name, elector ID, birth date, phone). You confirm or correct the two guesses on this screen. There is no way to keep a third column.

2. Only address and poll pairs cross to DoorLine

When you press Match, the browser collapses the rows to one line per civic address (units are folded away, because a poll is a property of the building) and sends just those pairs, about one line per address, with the poll beside it. Names and the rest were already gone. DoorLine then places each address on your map: exact matches first, then addresses written a little differently (22A where the map says 22, a half number), and a model decides the few genuinely ambiguous ones. Anything it cannot place is listed so you can pick the door or skip it; most of those are lots that are not on the map at all.

3. What is stored, how it is tagged, and how to purge it

Every poll DoorLine takes from the list is written as one row, address → poll, tagged with the import that produced it. Voting places you type in, and any poll you set by hand or accept from a suggestion, carry no tag: they are your data. Canvass data (knocks, supporters, GOTV status, signs) is never written to by any of this. A counts-only log records the date, who loaded it, the row and match counts, and the header names, and nothing from inside the rows.

When you are done with the list, or if you are asked to remove it, hit Purge voters' list data on the Polls page. It deletes exactly the tagged rows and stamps the log with the purge date, and leaves your voting places, hand-set polls, and every knock untouched. Remove all polls is the wider option that also clears voting places and hand-set polls. Replace list loads a newer export over the old one.

4. Voting places and the doors the list missed

Fill in each voting place (name, street address, hours) as the clerk publishes them. Blank is fine until then; splitting by poll does not need them, and sheets, stickers, and the app pick up whatever is filled in when you print or open them. A GOTV door the list did not cover (a new build, no registered voters, a sign at a business) appears under Doors to turn out without a poll with a suggestion from the nearest door that has one, same street first. Set it with one click. You can also change any door's poll from the Poll column on the GOTV target table.

When the clerk publishes who has already voted at advance polls, openAdvance poll results on the GOTV page. That card is separate from Polls and voting places, which only assigns each door a poll. The results file marks those doors voted, draws a check on the map, and leaves them off new walk sheets, new targeted lists, and new GOTV lists. Sticker counts drop them too.

The file is the same shape as the voters' list: a name, a property address, a poll, and a poll date. Names, the ward, and any extra code columns are dropped in your browser. DoorLine only receives the civic address, the unit when the address has one, the poll, and the date. Nothing unmatched is marked voted. Addresses it cannot place are listed so you can pick the door or leave them out.

A unit on the address marks that unit only. The building pin stays a normal door, the sheet names the unit, and only that unit drops out of GOTV. An address with no unit marks the whole door, even when more than one elector lives there. DoorLine does not match names, so it cannot tell those people apart.

On the canvasser map and the admin map, a door that has voted stays its support colour with a white check. Tap it and the sheet says already voted. A canvasser can still mark Already voted on a knock; that uses the same check and is not undone if you later remove the advance list.

Open turfs are not rewritten. The check appears the next time the canvass loads, including on a phone that was offline. Remove advance votes clears the list and puts those GOTV targets back, except doors a canvasser marked voted themselves.

Estimated Votes adds a card beside confirmed votes: how many matched doors you had already identified as supporters, how many as strong opponents, and how many you have not knocked. It is a count of doors, not a change to the model. Where the list had no unit, more than one elector can live at that door, so the number is a floor on ballots you can tie to an ID.

The top of the GOTV page shows targets, still to turn out, contacted, and voted, with a percentage. Each voting-day canvass shows its own voted count and a progress bar so you can move teams to where the list is longest.

Confirmed votes also appear on Estimated Votes in a separate “Confirmed votes” box. They are deliberately kept apart from the estimate: the model is a projection of the whole ward, while confirmed votes are real ballots you have watched go in. One is a forecast, the other is a floor. After you load an advance list, a second card counts the matched doors you had already identified. That count does not change the estimate either.

At the top of GOTV, add the dates you need a list for — advance polls, election day, a print deadline. DoorLine fills a table with first-time vote IDs at those dates. The levels and the Signs, Donate, and Volunteer columns are the same switches as Who is on the list. Turning one on changes who is on the target list, not only this forecast.

Weekly ID counts in real campaigns jump around (one Saturday can be three quiet weeks). The projection does not use an exponential. If new IDs have been rising, it keeps a gentle rise, pulled toward the last three complete weeks and capped at 1.5× your busiest week. If they have not been rising, it holds that recent weekly pace. It does not estimate how many volunteers you need — density and how people travel change that more than a formula would.

• Start with advance polls. Every supporter who votes early is one fewer door on election day, and the list you build for Tuesday is shorter for it.
• Refresh before each build. Supporters identified last night should be on today's list.
• Reminded is not voted. Keep knocking reminded doors on later passes until they confirm.
• Tell people where and when. Load the voters' list, split by poll, and sticker the voting place on the back of the lit. Volunteers then have it at every door without looking anything up.
• Rides win elections. Use the note field to flag anyone who needs one and have a driver work the exported list.