Ward-level election analytics, done properly

The question a campaign actually has

It is never "who are my voters". With thirty workers and two weeks, it is: which booths first, which message, and how many doors per day. Those are counting questions, and counting questions are answered by aggregates.

A ward of 1,400 voters across 240 households needs a different plan from a ward of 1,400 across 600. One booth with twice the first-time-voter share needs different material. None of that requires knowing a single person's name.

What aggregate analysis gives you

• Electorate size and change

  How many voters per ward or booth, and how that has moved since the last roll — additions, deletions and net change. This alone reshapes how a campaign allocates people.

• Age distribution

  Share of first-time voters, under-30s, and voters over 60, by booth. Determines both the message and the channel, since these groups are not reachable the same way.

• Gender ratio

  Male and female voter balance per booth, which is often strikingly uneven and matters for both canvassing and polling-day logistics.

• Household structure

  Average voters per household and the distribution of household sizes, derived from address grouping. Useful for planning door-to-door coverage — how many doors equal how many voters.

• Geographic concentration

  Which streets and localities hold the most voters, so canvassing routes are built on density rather than guesswork.

• Turnout history by booth

  Published booth-level turnout from previous elections joined to current rolls, to separate persuadable booths from low-turnout ones. These need completely different strategies.

• New and deleted voters

  Comparing consecutive published rolls to find additions and removals per booth — valuable for both campaigning and for spotting roll irregularities worth raising formally.

What we will not produce

We are explicit about this before a project starts, because it is usually the first thing we are asked for.

• Lists of named individuals with inferred community, religion or caste.
• Per-voter profiles combining identity, address and predicted political preference.
• Matching voter rolls against phone numbers, Aadhaar or any other database to identify individuals.
• Any output designed to target, pressure or exclude identifiable people.

This is not only an ethical position, it is a commercial one. Individual-level political profiling has drawn criminal complaints in this state, surname-based community inference has an error rate nobody can quantify, and a vendor who will build that for one client will build it about another. The aggregate work answers the strategic question without any of that.

How the work runs

1. 1
   Source the published roll

   From the official source, for the wards or booths in scope, with the publication date recorded.

2. 2
   Extract and structure

   Scanned rolls are processed through our Indian-language pipeline into structured records, with counts reconciled against the totals printed on the roll.

3. 3
   Aggregate immediately

   Individual rows are collapsed into counts by booth, age band, gender and household. The aggregate tables are the working dataset from this point on.

4. 4
   Analyse and report

   Tables, charts and a written summary of what stands out, in English or Telugu, built for the person deciding where to send workers tomorrow.

5. 5
   Delete the raw data

   Individual-level extraction is deleted on an agreed schedule once aggregates are produced.

Questions

Is analysing electoral rolls legal?

Electoral rolls are published by the Election Commission and are available to candidates and parties for electoral purposes. Aggregate statistical analysis of a published roll is ordinary political research. What is contested is bulk commercial profiling of identifiable individuals, which is a different activity with a different legal exposure, and is not something we do.

Why will you not do caste-wise voter lists?

Three reasons. It creates real legal exposure — in 2019 a firm in Andhra Pradesh faced police raids, an FIR and a special investigation team over allegations of exactly this kind of voter profiling. Surname is an unreliable proxy, so the output carries an error rate nobody can measure. And it is the use of this data most likely to cause harm to the people in it. Aggregate analysis answers the strategic questions anyway.

Can aggregate analysis really help a campaign?

Yes, and usually more than a name list does. Knowing that three booths hold 40% of a ward's voters, that one has twice the first-time-voter share, and that another had the lowest turnout last time tells you where to put your people. A list of names does not tell you where to go first.

What do we actually receive?

Ward and booth-level tables and charts, a written summary of what stands out, and the underlying aggregate dataset. No individual records.

How current is the data?

As current as the published roll you supply or we obtain from the official source. We date-stamp every dataset, because rolls revise and a stale analysis quietly misleads.

Can you do this outside Andhra Pradesh?

Yes. The pipeline handles rolls in Telugu, Hindi, Tamil, Kannada, Malayalam, Marathi, Bengali, Gujarati, Punjabi, Odia and Urdu, so the same analysis works for municipal, ZPTC, MPTC and assembly rolls in most states.

General information, not legal advice. Obligations around electoral roll use vary and are evolving under India's data protection framework — take your own legal advice before commissioning work of this kind.