
Gauge. Learn. Assign. Rally. Ensure.
The GLARE Method™ is Red Glare’s methodology for turnout intelligence. It starts from one claim. Turnout is decided in the final weeks, by whoever can still reach a supporter and be trusted. That access is built or lost months before, one supporter at a time, and won back by the right messenger, on an open channel, asking for a commitment.
Turnout is decided in the final weeks, by whoever can still reach a supporter and be trusted, and that access is built or lost months before. The GLARE Method turns that claim into an operating discipline. It places every supporter on a two-fact Support grid, watches for movement down the Quiet Curve, sends the messenger each person will listen to, wins a commitment in the voter’s own words, and lets the county’s voter file decide what counted. It runs all term, not the last sixty days. It persuades voters without an R to say yes, and it never discourages anyone from voting. This is the public overview. The full working paper, with every decision rule, goes to the campaigns that run it.
Turnout is decided late. Access is lost early.
Turnout is decided in the final weeks, by whoever can still reach a supporter and be trusted. That access is built or lost months before, one supporter at a time, and it’s won back by the right messenger, on an open channel, asking for a commitment. If access is lost one person at a time, the unit of work is the person, not the list. If it’s lost months early, the work runs all term. If the messenger matters more than the medium, the hardest decision is who asks.
Figure 1. The Quiet Curve. The path a supporter takes from engaged to absent, in five stages, each one visible in the record. Illustrative shape, not data.
A yes on file, replies, opens the door
Replies slow, a poll skipped
Silent past the phase’s threshold
The request window passed, or the ballot sits out
Didn’t vote
Every stage down the curve costs a rung up the ladder. The cheapest place to turn the curve is the earliest, and the earliest is the term.
Quiet depends on the calendar, and the curve only applies to people who once gave a signal. Someone never engaged isn’t slipping. They’re not yet asked.
The thesis has limits. National mood, the top of the ticket and the candidate set how many people vote, and no method controls them. Early contact fades, so the chase always reinforces what the term built.
The curve isn’t a sentence. A neighbor’s ask can lift a supporter from Quiet back to Engaged, and Gauge sees the move the day it happens.
Trust the R. Blast. Hope.
Most turnout programs run on three moves. Trust the R means treating a registration as a vote, and never asking the voter without one who would say yes. Registration work itself stays, as term work. Blast means the same message to everyone on the cheapest channel in the last two weeks. Hope means counting activity. None of it says who voted.
The term is the campaign.
GLARE fails if it starts in September. By then the Support grid is empty, the channels are closed and there’s nobody to send. The term builds three stocks. The window, the chase and the cure spend them.
Read monthly. The share of supporters with a yes on file.
Built from vote-intent and issue polls, door conversations, pledge pages.
Read monthly. The share with at least one channel open.
Built from opt-ins at QR codes and pledge pages, messages worth reading.
Read monthly. The count of people in the Champions box.
Built from one supporter at a time, asked in a quiet week.
Figure 2. Two modes across one election cycle. The three stocks a chair reads every month, and what builds each one. Widths show order, not measured time.
Those three numbers in August set the ceiling on October.
Registration is term work too. New movers, unregistered housemates of supporters and voters turning 18 are signed up before the deadline, and each enters the file as someone to ask.
The on-ramp
A cold file has no yeses, no Champions and almost no consent. The on-ramp is how a county earns its first yeses without burning its number on a cold blast. Champions come first, face to face. Consent comes before text, because a phone number on the voter file is not consent.
The term never really ends. New Jersey votes every November and the primary comes in June. What Ensure learns from one election seeds the next term’s Gauge the day the county posts its vote history.
Two facts. Four boxes.
The Support grid places every supporter on two facts, whether they’ve told you yes and whether they always vote. The box decides what happens next. Silence alone means little. The grid weighs it against vote history, so a quiet voter who has voted in eight straight elections is a Regular, owed a reason to say yes, not a nudge to show up.
Figure 3. The Support grid. The real signal is movement between boxes. People move up on a yes and right on vote history, and they can slip back. A no sits outside the grid, counted on its own. A persuadable voter steps in the day they say yes.
Beyond the R
The R on the file is a starting point, not a boundary. The top row of the grid reads said yes, not registered Republican. An unaffiliated voter who says yes is a supporter, sorted and turned out like any other.
Persuasion moves fewer people than turnout work, and it moves them most with a messenger they know, on an issue they named. That’s why it lives in the term. Supporters are found by testimony and the household, never by guessing from demographics.
One grid, five ways to see it
The Support grid opens as Glare. Everyone you haven’t asked yet sits at the core, the inventory every ask draws from. The four boxes sit around it, one mark per person. Switch to Cross, Meter, Table or Rows and it’s the same people, drawn another way. Every count opens the people behind it.

Figure 3a. Glare, the default view. Its redesign, broken down by precinct, is shipping next. Sample numbers for a made-up county.
Five verbs. One owner each.
Every step has an owner and a measure. A method nobody owns is a poster. The working paper sets out each step’s deliverable and decision rule.
See who’s slipping while there’s time.
Supporters don’t vanish in November. They go quiet in June. Gauge reads every reply, knock, silence, poll answer and ballot state as it lands, keeps every supporter in the right box on the Support grid, and raises the people moving the wrong way.
It prevents trusting the registration. An R on the file is not a vote.
Name the reason, against the calendar.
Quiet isn’t one problem, so it can’t have one fix. Learn gives every name on the Slip List exactly one reason, drawn from the record and judged against what the election calendar still allows. A reason the calendar has closed is no reason to act. Ballot and channel states are flags the system tracks on its own. A reason is what the voter needs from a person.
It prevents one risk score that chases your most reliable voters hardest.
Send who they’ll listen to.
The messenger is the message. Assign matches every reason to the messenger most likely to be heard, on a channel that’s still open, and puts a name on every claim so nobody knocks the same door twice.
It prevents blasting the whole list on the cheapest channel and calling it outreach.
Win the commitment.
A contact isn’t the goal. A commitment is. Every conversation ends in one concrete thing the voter agrees to do, written to the record in their own words, and the follow-up quotes it back from the person who asked.
It prevents a friendly “we’ll see” counted as a win.
Let the voter file decide.
A click isn’t a vote, and neither is a promise. Ensure checks every commitment against the county’s files, keeps one ledger of who actually voted, and feeds what it learns back into Gauge.
It prevents counting activity, then learning the wrong lesson for three cycles.
Where does your operation actually sit?
Every operation sits on one of five levels, and any chair can place their own in thirty seconds. The test is what they can show, not what they say.
We have the file.
Can show. A count of registered Republicans.
We text everyone.
Can show. Messages sent, doors knocked.
We see who’s quiet.
Can show. A Slip List, with a reason on each name.
We get plans.
Can show. Every box on the Support grid counted, and the share of yeses with a plan.
The voter file confirms it.
Can show. The three stocks by month, who moved boxes this week, and a holdout read against vote history.
Figure 6. Five levels of turnout operations. A county that sends everyone the same text in October is at level 2. The jump from 2 to 3 is the term’s work. The jump from 4 to 5 is a holdout and the patience to wait for the county’s files.
GLARE for a small county
Most county committees run on a handful of volunteers and go quiet between November and summer. GLARE is built to work at that size. In the term, a small county runs it once a month. One poll or one message that gives, one look at who moved boxes, one round of asks.
The software does the sorting, so people only show up where the voter needs a person. The daily rhythm starts in the window and the chase. Templates are approved once, so a volunteer never waits on a captain in a cure window. That county is at level 2 or 3, and that’s a real program.
Seven rules that keep it honest.
The method only works if it stays a loop. These rules stop it turning back into a list and a guess, and keep it on the right side of the line.
The last sixty days spend what the eighteen months before them built.
What a supporter told you outranks what a model guessed about them.
Who asks matters more than how it’s sent. Send the neighbor.
Every ask is voluntary, budgeted and logged in their words. Nobody’s record is shown to anyone.
Once, at the template and its audience, so nobody waits in a queue. The floor can stop a message. Only a person starts one.
The voter file decides, and a holdout says whether the work caused it.
GLARE brings your own people out and persuades new ones in. It is never used to keep anyone home.
What’s proven, what isn’t, and how we’ll find out.
GLARE is built from published research. It has not yet been shown to raise turnout as a whole method, and nothing Red Glare publishes says otherwise.
Face-to-face canvassing raising turnout is one of the best-replicated findings in field experiments.1 Social pressure from neighbors has strong evidence behind it.2 Research on vote plans is mixed, and later studies found smaller effects than the first.
The large social-pressure effect came from telling voters their record would be shown to neighbors. GLARE refuses to do that, so its neighbor-as-messenger version should be expected to move less. Every threshold in the method is a design choice, not a measured result. Early contact also fades, which is why the chase reinforces what the term built.
The plan, the counties and the comparison, filed before Election Day
Whole precincts or street segments, at random, before anyone is contacted
No contact, a standard October program, and full GLARE
Vote history compares the two. Inside the noise is chance.
Figure 7. The test built into the method. A congressional district or statewide race, precincts held back so neighbors can’t leak the treatment, about 20,000 people or more, three arms, and the answer published whichever way it falls.
Until that test is read, GLARE is a well-built bet. After it, GLARE is either proven or corrected.
- Gerber, A. S., and Green, D. P. (2000). The Effects of Canvassing, Telephone Calls, and Direct Mail on Voter Turnout. A Field Experiment. American Political Science Review, 94(3), 653-663. Cambridge
- Gerber, A. S., Green, D. P., and Larimer, C. W. (2008). Social Pressure and Voter Turnout. Evidence from a Large-Scale Field Experiment. American Political Science Review, 102(1), 33-48. Yale ISPS
The rules that run it stay with the campaigns that run it.
This page is the why. The working paper is the how, every rule a county needs to run GLARE on Monday morning. It goes to campaigns that run Red Glare, walked through on your own voter file.
The deliverable, the decision rule, the measure and the owner for all five verbs
The reasons, in priority order, and who to send for each
Who asks, on which channel, and what counts as a vote
How often anyone is asked, phase by phase
Who goes first on a cold file, and how a county earns its first Champions
The whole operation, week by week, from the term to the cure
Asked about the method.
Has the GLARE Method been proven to raise turnout?
Not as a whole method, yet. Its parts rest on published field experiments, and the method is built to be tested with a randomized holdout in a real election. Red Glare publishes no lift numbers.
What is the Quiet Curve?
The Quiet Curve is the path a supporter takes from engaged to absent, in five stages. Engaged, Fading, Quiet, Missed and Home. It falls during the term, months before most campaigns start, and every stage leaves a mark in the record.
Does GLARE only work with registered Republicans?
No. The Support grid sorts people by whether they said yes, not by party. An unaffiliated voter who says yes is a supporter, sorted and turned out like any other.
How does a county with a cold file get its first yeses?
Through the on-ramp, a consent-first way to earn a county’s first yeses face to face before anyone gets a text. Champions come first. The order, the sources and the measure are in the working paper.
Can I read the full method?
The full working paper goes to campaigns that run Red Glare. We walk through it on your own voter file, so every rule lands on real names. Book a walkthrough and bring your file.
Can a small county run GLARE?
Yes. In the term it runs once a month, with one poll or message, one look at who moved, and one round of asks. The software does the sorting. The daily rhythm starts near the deadlines, when more hands show up.
How is GLARE different from a turnout score?
A turnout score guesses once from history and demographics. GLARE reads what each supporter actually does as it happens, gives each one a reason and a messenger, and lets the voter file grade the result.
Run GLARE on your county.
Thirty minutes. Bring your voter file and we’ll show you where your supporters sit on the Support grid, who’s sliding down the Quiet Curve, and who to send.