Guides

Name matching across Indian identity documents: a field guide

APIZONE · ·updated 8 Sep 2026 · 1 min read

The same person's name is spelled three ways across their PAN, Aadhaar and bank account. Here is how to match them without false rejects.

On this page
  1. Sources of variation
  2. A matching pipeline
  3. Tune with your own data

Sources of variation

Initials vs expansions (R Kumar vs Ramesh Kumar), order (given-name-first vs surname-first), honorifics, spelling of the same sound (Mohammed, Mohammad, Muhammad), and joined vs separated words. All of these are the same person.

A matching pipeline

Normalise: lowercase, strip punctuation and honorifics, collapse whitespace. Tokenise. Compare token sets with allowance for one being an initial of the other. Score with a string-similarity metric on the remainder. Accept above a threshold, review the band below it, reject only clear non-matches.

Tune with your own data

Pull a sample of known-good pairs from your existing verified customers and set the threshold where it clears them without letting through obvious mismatches. Revisit quarterly.

Share X LinkedIn WhatsApp