An assembly seat in a large Indian state is about 2.5 lakh voters, 300 booths, one winner and a Rs 40 lakh legal ceiling on what the candidate may spend (Election Commission limit for larger states, revised January 2022). Most candidates meet the seat in its last 40 days, through the people who want something from them. The ones who win narrowly usually met it a year earlier, through the data. This is the year, month by month.
Months 12 to 10. The map
Before anything is said, the seat is read. Form 20 booth-wise results for the last four or five elections, assembly and parliamentary, from the Chief Electoral Officer's website. The current roll by part, with growth since the last poll. Census ward data. Scheme beneficiary counts by panchayat or ward. On one screen, three colours: booths that never move, booths that always move, and the 50 to 70 that decide the seat. The campaign's money will go there. Most candidates have never seen this map. Every serious consultancy starts with it. So do we.
Months 10 to 8. The survey
Trained enumerators, in homes, several rounds. What changed in your lane. What you are angry about. Who you trust. What would make you switch. Half the enumerators women, or the campaign hears half the roll. The findings go into a single page the candidate is made to read twice: the three issues that actually move this seat, in the voter's own words. The slogan is written from that page, not from the manifesto.
Months 8 to 6. The machine
A named person for every booth. A booth committee for the decisive 60. A grievance desk on a helpline, a missed-call number, SMS and a web form, feeding one CRM, so that a complaint received in January is closed in February with a photograph, and logged. Not on WhatsApp: Meta's WhatsApp Business policy bars political campaigns from the Business Platform, and bulk tools on ordinary numbers get banned. A content engine in the local language: short video of work done, not speeches. Google presence so that when the constituency searches the candidate's name, the candidate's own account comes first. All of this costs least now and impossible in the last month.
Months 6 to 4. The message test
Three messages, sometimes five. Small ad budgets, real audiences inside the constituency, measured by saves, shares and replies. The one that moved people is the one that goes on 40,000 posters and 300 wall paintings. The one the committee liked goes nowhere. One week, one bruised ego, several thousand votes.
Months 4 to 2. The ground log
Door to door begins in the decisive booths, then widens. Every visit logged: house, booth, response, follow-up. A volunteer who reports 90 homes is asked for 90 names. The war room, a single screen, reads the log nightly: homes covered by booth, complaints opened and closed, ad reach by ward, opponent moves by day. A problem in booth 141 is known on Wednesday. A WhatsApp group with 400 members is not a war room.
Months 2 to 0. The run
The nomination is filed and, with it, the declared social media handles. Every paid advertisement is pre-certified by the Media Certification and Monitoring Committee. Ad accounts on Google and Meta are verified in the right name and the spend reconciles to the ledger. The expenditure register is kept daily against the Rs 40 lakh ceiling, because an observer will read it. Rallies happen, because they must, but they are the visible ten per cent. The counted ninety is the ground log.
The last 48 hours
Section 126 of the Representation of the People Act silences campaigning through television and similar apparatus for 48 hours before the close of poll, and the Commission's instructions with the platforms' 2019 Voluntary Code extend that to paid digital. The switch-off is executed to the minute, in writing, by someone whose job it is. Then the booth committees do the only thing left: turnout. The 60 booths. The households on the list. A vehicle for the elderly. A phone call to the family that moved to the city.
The five mistakes the log catches
One. The campaign spends the first six weeks in the candidate's own strongest booths, because that is where the welcome is warmest. The map says those votes were already won. Two. The message changes every week, because a new adviser arrived. The test said one message. Three. The volunteer count is real but the visit count is not, and nobody checks the names until it is too late. Four. The outreach runs on a bulk WhatsApp tool, the number is banned in week three, and the grievance list goes with it, because the platform does not allow campaigns on it in the first place. Five. The expenditure register is reconstructed from memory the night before the observer visits. Every one of these is visible on a single screen, weekly, if the screen exists.
The booth committee, the unit that wins
Everything above reduces to one unit: a booth committee of five to ten people who live in the booth, know the 800 to 1,200 voters on its roll by face, and own three jobs. Before polling: the list of households, visited and logged. On polling day: turnout, household by household, with a vehicle for anyone who needs one. After: the Form 20 number for their booth against the target, so that the next campaign's map starts the morning after this one ends. A campaign with strong booth committees in the decisive 60 can lose the air war and still win. The reverse is rarely true.
What the candidate does, personally
Reads the map once and the war-room screen nightly. Sits in on one survey debrief. Chooses the message the test chose, not the one the family liked. Visits the decisive booths twice and the safe booths once. Signs the expenditure register weekly. Everything else, the machinery, the content, the compliance file, the dashboard, is delegated to people whose job it is, and checked. A candidate who does the machinery personally has no time to be the candidate.
After the count
Win or lose, the machine is the asset. The grievance desk, the content engine, the map and the database do not switch off. For a sitting MLA they become the four-year campaign that makes the fifth year easy. That is the part of political consulting that is simply good operations, and the part Apex builds, directly for a candidate or under a consultancy that owns the strategy. The scope, the fee logic and the refusals are on the political consulting page.
Frequently asked questions
How many booths decide an assembly seat?
In a typical seat of about 300 booths, the swing seats are the 50 to 70 where past margins were narrowest, the roll grew fastest or intermediaries changed. Four or five elections of Form 20 data make them visible.
What is the spending limit for an MLA candidate?
Rs 40 lakh in large states (Rs 28 lakh in smaller ones), set under Section 77 of the Representation of the People Act, counted from nomination to the declaration of results, and audited by an expenditure observer.
When should an MLA campaign start?
Twelve months out, with the booth map and the survey. The grievance desk, content engine and search presence cost least to build six to eight months out and nearly impossible in the last month.
What is Form 20?
The booth-wise result sheet the Returning Officer publishes after every count, available on each state Chief Electoral Officer's website. It is the foundation of any honest constituency analysis.