By Sivaram Kowta, President – Digital Banking, Zeta
A shopper stands at the checkout with ₹12,000 of appliances in her trolley and ₹8,000 in her account. Inside UPI today, she has no option. She will put something back, or leave and return another day. Her bank holds the account, knows her salary and carries the risk on every payment she makes. But the app on her phone is the one that sees this moment, and it is the app she will open next time.
That scene is the real story of Indian digital payments, and it is bigger than any single fee. NPCI has just deferred the MDR on larger UPI payments to immediately post Diwali, on a date still to be notified. Policy of this kind shifts who funds the system, which costs an estimated ₹20,000 crore a year to run. The more consequential question is what banks do with the advantages they already have.
The bigger cost never showed up on the P&L
For six years, UPI has been free to use, and the subsidy has drawn most of the attention. It was never the most expensive cost. The real price was that customers stopped coming to the bank at all. Banks built the rails, and three apps built the experience on top of them and built it well. Those three now handle about 85% of UPI volume. The accounts and the risk still sit with banks, but the daily relationship sits elsewhere.
The bank sees a settlement instruction. The app sees what the customer buys, where they buy it, how often, and whether the money runs out before the month does. One of those is a record. The other is a relationship.
That relationship is valuable, and others have learned to monetise it. CRED turned credit card bill payments into a habit for customers with scores above 750 and now earns about ₹2,000 a year from each of its 1.26 crore monthly users on a lending book of ₹22,000 crore, much of it the upsell those issuers would once have made themselves. The pattern is broader than one company: NBFCs write 91 of every 100 personal loans in this country, on CRIF High Mark’s numbers, while banks hold 56% of the value and 7.3% of the volume.
Free UPI gave the apps engagement, and engagement let them build businesses in credit, insurance and investments. The largest more than doubled its lending revenue in FY26 without earning a rupee from payments. Once the merchant fee takes effect, they will also earn at least 8 basis points on every qualifying transaction, and will rightly reinvest it in growth. If apps reinvest while banks merely recover past costs, the distance between the account and the customer will only widen.
To be sure, banks are not without advantages. They own the account, they alone hold the licence to lend at the moment of payment, and they carry decades of trust that no funding round can buy. Each of those points to something worth building, and I would take them in that order.
Make the account a product worth staying with
Customers choose a UPI app out of habit, and habit forms in sequence: reliability creates it, convenience consolidates it, rewards reinforce it. Banks have largely sorted the first, but only on one side of the equation. Technical declines are now under 1%, almost all of that from issuer-side investment. The half left loose is the PSP end, where the customer actually is, and it is precisely the half three apps have taken over.
Convenience is where the bank’s position is strongest and least exploited. It holds the account, so balances, standing instructions, deposits and loans already sit in one place. Add-on UPI IDs against that account, with limits by person and category, would let a driver or a college-going child pay without anyone sharing a PIN. A bank’s UPI ID can be saved with the merchants a customer already uses, much as cards are. And when a payment goes wrong, the bank can resolve the dispute end to end.
Rewards come next, and in that order. Merchant fees, supplemented by merchant funding, can finally pay for UPI rewards that are earned, visible and redeemable instantly. They are best aimed above ₹2,000, where both the revenue and the behaviour worth shifting sit. For small everyday payments, speed and certainty will always matter more than points. Each of these makes the bank’s UPI ID harder to abandon.
Use the one thing only a bank can do
A pre-sanctioned credit line, which only a bank can issue, gives the shopper the option they lack. Suryoday Small Finance Bank has sanctioned such lines to over five lakh customers, 97% of them new to the bank, through a partner app rather than its own. Roughly 150 million Indians have a credit record and no credit card at all. A bank does not need to own the front end to own the funding source.
Deepen trust, which has to be earned every day
EY India found this year that 82% of customers trust their bank with their data, but only 73% believe fraud gets resolved effectively. That gap is the opportunity. Most UPI fraud is not a broken system but a stranger persuading someone to enter a PIN, and it ran to ₹805 crore in the first eight months of FY26. Device-bound authentication, and the trusted-person approval the RBI proposed in April, would close that door. Banks would do better to build it now than adopt it later as a mandate.
Getting past the plumbing
None of these ideas is new inside banks. The constraint is that savings, cards and lending sit in separate systems built in different decades, and most of the budget goes into keeping them running. The answer is not to replace the core in a hurry, but to build a modern layer alongside it to carry the new capabilities.
One trap to avoid: do not build only for the transactions that now earn MDR. Credit on UPI averages ₹1,125 a transaction and is used around 40 times a month, against eight for a card. The habit is built below the threshold while the revenue sits above it. A product that serves only the second will never earn the first.
MDR settles an old question about who pays for UPI. It leaves a bigger one open. For six years, banks funded the rails while apps built the everyday experience, and UPI is better for both. Banks now have the money, the licence and a finite window in which to make their own app worth opening too.