Pehchaan turns the UPI payments every Paytm merchant already receives into Blinkit-grade customer intelligence — who's slipping away, who to win back, and what just opened down the road. Spoken in Kannada on the Soundbox.
The moment a UPI payment lands, Pehchaan knows who it is, how often they come, and whether they were slipping away — and tells the shopkeeper out loud.
The kirana's only edge is the relationship. But a shopkeeper with 400 customers can't notice who stopped coming — until they're gone.
Nobody cancels a kirana. They just order from an app one week, and don't come back. There's no signal — unless you're counting.
Every UPI payment carries the customer's name and ID. Paytm sees it for ~1.5 crore subscription merchants — and today it's only used to say “payment received”.
A dip at one kirana is a bad week. The same dip at seven shops around one point is a new dark store — and only the network can see it.
Sources: Grant Thornton Bharat kirana survey via NEXT IAS (Jul 2026); Paytm subscription merchants via Inc42 (FY26)
Not “hasn't come in 30 days”. A daily regular gone for a week matters more than a monthly one gone for a month — the model knows the difference.
The Soundbox tells the owner who's slipping and who to expect — no app to open.
Trained to June, tested on 75 days it never saw — and on real CDNOW purchase data it recovers the published model parameters within 3%.
Chooses who's worth an offer, writes it, sends it through the Paytm app, and confirms the return at the counter.
Regulars leaving many shops in the same week reveal a new competitor. Paytm can warn merchants, pre-fund offers and brief field teams before the damage spreads.
per-customer correlation on months the model never saw
error locating a new dark store from aggregate drops alone
a month in spend slipping away at one kirana — now visible
Paytm subscription merchants it can run on without new hardware
No. Every number comes from statistical models fitted to payment history — MBG/NBD for churn and Gamma-Gamma for spend — chosen by testing four approaches out of sample, on our shop and on real CDNOW purchase data. The LLM only writes the Kannada message; it never decides who or how much.
From payments the merchant already receives: the payer name and UPI ID that appear on every Paytm for Business statement. No new app, no data entry, no loyalty card.
Merchants never see phone numbers. Offers are delivered as Paytm app notifications under Paytm's consent settings, and customers can opt out. Network insights are aggregated across shops.
It is designed to: payment data is processed only in India (RBI payment-data storage rules), payer data is never shared outside Paytm (NPCI UPI norms), customers opt in and can withdraw or be erased (DPDP Act 2023), messages go only to opted-in customers with a monthly cap (TRAI), and scores are never used for credit. Every agent action is in an exportable audit log.
Pehchaan only sees UPI regulars — today the majority of a city kirana's repeat customers, and growing every month. Cash customers simply aren't scored.
Three things in one place: the Soundbox at the counter to recognise people, the consumer app to deliver offers, and payment data across ~1.5 crore subscription merchants to see neighbourhood-level shifts.
Open the live dashboard for Sri Manjunatha Stores, HSR Layout.