Skip to content

ATMs in a tier-1 city

Transaction demand ÷ machine capacity, with a digital-payments twist.

moderate
5 min read
infrastructuredemand-capacity

Estimate the number of ATMs in an 8-million-person tier-1 city. Build cash-withdrawal demand, divide by per-machine capacity — and address the elephant: UPI has eaten most small transactions, but cash demand hasn't vanished.

Population → card-holding adults → two cash-behaviour segments → transactions → ÷ machine economics. UPI lives inside the frequency split.
1

Users

8M population → ~5.5M adults; ~80% have bank accounts with cards and occasionally withdraw → 4.4M users.

2

Frequency

Pre-UPI this was 4–6 withdrawals/month; today salaried users pull cash ~1–2×/month, cash-economy workers more → blended ~2/month → 8.8M/month ≈ 290K transactions/day.

3

Capacity

An ATM is economically viable at ~80–100 transactions/day (interchange fees vs rent + cash logistics + maintenance).

4

Count

290K ÷ 90 ≈ ~3,200 ATMs.

4.4M users × 2/mo ÷ 30 ÷ 90 txns/machine ≈ 3,250 ATMs

How to defend it

Acknowledge UPI explicitly and put it in the frequency assumption, not as a vague caveat. If pushed on "is the count rising or falling," reason from machine economics: falling transactions per machine push marginal ATMs below viability → consolidation.