Skip to content
Intelligence · Service

Election Survey & Voter Analytics

Most campaigns in India are still run on the confidence of the loudest man in the room. We replace that with a stratified sample, a caste-and-community weighted model and a booth-wise scoreboard you can actually act on.

±3%
Typical margin of error
100%
GPS + audio audited responses
72 hrs
Baseline report turnaround
01

What a real constituency survey looks like

We start with a baseline survey 150–180 days out: 1,200 to 25,000 respondents depending on seat size, stratified by booth cluster, age band, gender and community so the sample mirrors the electoral roll rather than whoever picked up the phone. Field agents collect on a mobile app with GPS stamping and voice recording, so every response is auditable. Nothing is entered from memory at the end of the day.

02

The four questions every survey must answer

One: what is the current vote share split and how soft is it. Two: what is the candidate's personal favourability versus the party's, because those numbers diverge more often than campaigns expect. Three: which three issues actually move a vote in this seat — in most Indian constituencies it is water, roads and jobs, in that order, not the state-level narrative on TV. Four: which booths are winnable, which are lost, and which are the 15–20% swing booths where the entire budget should go.

03

Tracking, not a one-off report

A single survey is a photograph. We run a baseline, a mid-campaign tracker at T-60, and a rolling 3-day tracker in the final fortnight with a smaller rotating sample. That is what tells you whether a rally, a scheme announcement or an opposition attack actually shifted anything — and lets you kill spending on messages that are not landing.

04

Delivered as a decision tool

You get a live dashboard, not a 90-page PDF nobody reads. Booth-wise win probability, issue heat maps, community-wise swing, and a ranked action list for the candidate and each ward in-charge. Your war room, booth teams and messaging calendar all pull from the same numbers.

FAQ

Election questions, answered directly

How large a sample do we need for an assembly seat?

For a typical Indian assembly constituency of 2–2.5 lakh voters, 2,400–3,600 respondents gives a ±2–3% margin at booth-cluster level. For a Lok Sabha seat we run 8,000–25,000. Anything under 1,000 for an assembly seat is a vibe check, not a survey.

How do you stop field agents from faking responses?

Every response is GPS-stamped at the point of capture, 15% of interviews are audio-recorded and back-checked by a separate team, and we run statistical outlier detection on agent-level response patterns. Agents whose data fails back-check are removed and their entire batch is re-collected.

Can you survey a seat we have not decided to contest yet?

Yes, and it is the cheapest decision you will ever make. A pre-entry feasibility survey costs a fraction of a campaign and has talked more than one client out of a seat they would have lost.