Political survey company in India — GPS-verified surveys and campaign field operations

Election survey agency India playbook: an assembly survey in four stages

An assembly constituency typically has 250–450 booths and 2–3 lakh voters. Surveying every single house is neither necessary nor practical. The sound method is a fixed number of households from each panchayat or ward (say 30–50), spread across neighbourhoods and times of day. In Bharat Survey that target is set with the assignment, and the dashboard shows which panchayats have met it and which are slipping.

Stage 1: area and team

Choose the constituency — panchayats, villages, wards and (if you uploaded a booth master) booths arrive preloaded. Every volunteer gets one seat and one authorised device; each supervisor looks after 8–12 volunteers. Assign by panchayat or ward, not by block, so the out-of-area check means something.

Stage 2: questionnaire

Keep the first (baseline) survey short: 8–12 minutes. Neutral wording, with “don’t know / prefer not to say” on every question. Ask sensitive questions (caste, religion, past vote) only when truly needed, and only after our form review. A published form locks; any change creates a new version, so wave-on-wave comparison does not break.

Stage 3: field and review

A 45-minute morning briefing, 8–14 surveys per volunteer per day (depending on settlement pattern), and an evening pass through the QA queue by the supervisor, with back-checks on high-risk records. Within the first three days the team learns that the system notices, and quality improves on its own.

Stage 4: report and the next wave

Panchayat-wise issue lists, facility conditions and satisfaction scales — each with its time window and response count. That decides the next issue poll, and the survey’s public opinion feeds into booth ratings.

What a voter survey questionnaire should contain

A good questionnaire is not judged by how many questions it has, but by whether each answer can change a decision. A working structure:

• Introduction and consent — who is asking, the purpose, voluntary, no benefit or harm. (A mandatory screen in the app)
• Area — panchayat/ward and booth, filled from location; the surveyor only confirms.
• Respondent basics — age group, gender, number of voters in the family; all optional.
• Biggest local issue — single-choice list + “other” (text). This is the most useful question on the form.
• Facilities — water, power, roads, health, school: a 1–5 rating, with a photo if wanted.
• Schemes — benefits such as ration, housing, pension, Ujjwala: yes / no / application pending / don’t know.
• Satisfaction — with the sitting representative or panchayat, on a 5-point scale; it can be asked without naming a party.
• Help request — with separate consent, in a separate ticket.
• Visit outcome — complete/partial/refused/locked: essential for the report’s denominator.

How to catch fake surveys — six signals

No single signal is proof; three or four together make a pattern. Bharat Survey combines them into a risk score automatically, but the method works on paper too:

• Location versus area — the submission’s location falls outside the assigned panchayat’s approximate boundary.
• Time profile — a 10-minute form done in 90 seconds, or one volunteer’s 40 surveys spaced exactly 3 minutes apart.
• Mock location — Android’s signal that an app generated the location.
• Duplicates — the same household label plus nearly the same location, or the same answer pattern again and again.
• Impossible sequence — village A at 09:10, village B 12 km away at 09:25.
• Back-check — the supervisor revisits 5–10% of households and asks 3 questions; if disagreement is high, hold the whole batch.

Keep in mind: a weak signal is not fraud. That is why “weak/stale signal” is its own state — it goes to review, not to rejection.

A 90-day campaign field plan

This is a sample framework that we adapt to your constituency in the demo. Field activity during the model code of conduct, the pre-poll silence period and the exit-poll embargo must follow local rules — that responsibility sits with the campaign.

• Day 1–7 — setup and baseline: project, seats, assignments, 45-minute training; a 3-day pilot in 2–3 panchayats, then the whole constituency.
• Day 8–30 — baseline done + booth master: every panchayat meets its target; booth master imported, booth in-charges and panna pramukhs assigned, first A/B/C/D ratings.
• Day 31–60 — issue polls and facilities surveys: 2–3 short waves; a photo list of public facilities; help tickets open and close; daily booth reports begin.
• Day 61–85 — weak booths and revisits: survey priority on C/D booths, a list of “revisit” households, rating updates with reasons.
• Day 86–90 — polling-day preparation: polling-day mode rehearsal, Offline Permit refreshed on booth in-charges’ phones, dry run on the war room board.

What an election survey costs depends on these factors

We do not print a price on this page, because no single number is right for every constituency. What shapes the quotation:

• Number of seats — volunteers active at the same time. This is the main factor; Viewers and field roles (booth in-charge, panna pramukh) need no seat.
• Duration — a 30, 60 or 90-day campaign window; longer plans lower the monthly cost.
• Projects / constituencies — one panel, many projects, for agencies and parties.
• Setup — parsing the booth master PDF, custom forms, XLSForm import, training sessions.
• Modules and services — booth management and the war room switch on only for political tenants; field teams from us are quoted separately.

The bigger cost than software is volunteer honorarium and travel — and that is the money wasted most when surveys are fake. Verification protects exactly that spend. The same platform serves our non-political work too, which is why teams looking for a market research survey agency in India sometimes land here: the fieldwork discipline is identical.

Seven common election survey mistakes

These mistakes come from planning, not software — but the right tool surfaces them early.

• Assigning by block — tell someone “the whole block is yours” and nobody knows which panchayat was missed. Assign by panchayat/ward; only then does the out-of-area check mean anything.
• Only main-road houses — volunteers pick easy houses. Give instructions with the target: different neighbourhoods, different times. If the map shows pins bunched in one place, the supervisor steps in.
• A 20-minute questionnaire — after the tenth minute, respondent and surveyor are both in a hurry. Keep it to 8–12 minutes; move the rest to the next wave.
• Changing questions between waves — change a word and the comparison breaks. That is why a published form locks and edits create a new version, named in the report.
• Not recording “refused” and “locked” — counting only completed surveys produces false percentages. Record every visit outcome, and the denominator stays honest.
• Calling a convenience sample “public opinion” — people found at home are not a representative sample of the constituency. Our reports do not claim that; every number carries its response count and time window.
• Treating back-checks as distrust — revisiting 5–10% of households is not suspicion of the team; it is insurance for the data. Say in week one that it applies to everyone.

What the evening report contains — a sample structure

There are four report categories, and each exports to PDF, Excel or CSV. No number appears on any page without its time window, form version and response count.

1. Work progress

Target vs visits vs completed vs accepted, by panchayat or ward; each volunteer’s work for the day; which panchayats are done and which are behind. Example (sample): “Barauli — target 40, visits 46, completed 39, accepted 37”.

2. Data quality

The split between accepted / in review / out of area / unclear signal; average interview length; back-check match rate; how many of each volunteer’s records came back for correction, and why.

3. Local issues

The biggest issue per panchayat, 1–5 facility ratings, scheme benefit status and a located photo list of public problems — so the leader knows before a visit what each village will ask about.

4. Help requests

Tickets taken with separate consent: how many are open, in progress and resolved per panchayat. Contact details are visible only to the person the ticket is assigned to.

Glossary — campaign and survey terms

• Assembly constituency (AC) / Lok Sabha constituency (PC) — electoral constituencies; the PC → AC → polling station structure is already built in.
• LGD code — the code from the government’s Local Government Directory that gives every district, block, panchayat and village a fixed identity.
• Booth / part number — a polling station, usually ≈1,000–1,400 voters. Mandal / shakti kendra — a group of 5–7 booths.
• Booth in-charge, BLA, panna pramukh — the booth team; a panna pramukh looks after one page of the voter list (≈30–60 voters). More on the booth management page.
• Wave — a short survey repeating the same question set, so change over time becomes visible.
• Denominator — the base of a percentage. “62% satisfied” means something only when you know 62% of what — 37 answers or 370.
• Arrival check — location, time and accuracy recorded on reaching the assigned area.
• Risk score — a score built from signals such as area, timing, duplicates and mock GPS; the QA queue is sorted by it.
• Back-check — a supervisor’s independent revisit, linked to the original record.
• Offline Work Permit — a 24-hour permission to work without a network; a lost phone cannot keep data indefinitely.
• Tenant — each candidate’s or project’s separate, closed data space. Viewer — a view-only role that needs no seat.
• XLSForm — the standard format for questionnaires written in Excel, used by Kobo/ODK/SurveyCTO and imported directly here.
• EVM demo link — people often search for an “evm demo link”; official EVM and VVPAT demonstrations come from the Election Commission of India, not from us. We do not handle voting machines.