UPI worked because the state built a shared rail and left the products on top of it to private firms. That division of labour is the most reusable idea in India's recent economic policy, and it is also easy to copy badly. A public rail succeeds when it is open, cheap to join and narrow in what it does. It fails when it tries to become the product.
What the rails actually do
The Unified Payments Interface, run by the National Payments Corporation of India, is a common protocol for moving money between bank accounts in real time. Any licensed app can plug into it, and any bank can be reached through it. The consequence is that a payments company does not have to negotiate bilateral connections with every bank or build a closed network of merchants. It competes on design, reliability and the services it layers on, while the settlement underneath is shared.
Other rails follow the same pattern at different layers. Aadhaar-based authentication and electronic KYC let a regulated firm verify who someone is, with consent, at a small fraction of the cost of paper onboarding. The Account Aggregator framework, regulated by the Reserve Bank of India, lets a person share financial data from one institution with another through a licensed consent intermediary, so that a lender can see bank statements without asking for a photocopy. The Open Network for Digital Commerce applies the idea to buying and selling: a buyer application and a seller application can transact without sharing a single platform.
In each case the rail takes a cost that every participant used to pay separately, such as identification, connection or data collection, and makes it a shared utility. The private reward comes after: lower cost to acquire a customer, new products that were uneconomic before, and a bigger addressable market because small users are suddenly cheap to serve.
What the model needs to repeat
Not every sector has the same shape as payments, and the attempts to extend the model have had uneven results. ONDC, to be candid, has grown more slowly than UPI did. That is unsurprising: payments has one transaction type and a clear regulator, while commerce has a thousand categories, logistics dependencies and entrenched platforms with strong reasons to stay closed.
Three conditions seem to separate rails that take hold from those that do not. The first is a narrow, standard core: UPI standardises a payment instruction and little else. The second is a participant that must join. Banks were required to be on UPI, and that gave every app a reason to build for it. The third is a use case that is frequent and cheap enough that adoption is habit and not an event.
The state's job is to build the road and publish the rules of the road, not to run the buses.
Health is the obvious next candidate. The Ayushman Bharat Digital Mission is building health identifiers and consent-based sharing of records, which could let a diagnostic lab, a hospital and an insurer work from the same patient history only when the patient agrees. Logistics has a start in the Unified Logistics Interface Platform, which opens government data on cargo, vehicles and documents to private logistics operators through standard interfaces. Skills and credentials are a plausible third, where a verified record of training and employment would cut the cost of hiring and lending to workers outside the formal system. None of these is proven. All of them have the same structure.
Where the model can go wrong
The first risk is who pays. UPI has been free for most merchants, since the government set the merchant discount rate at zero for UPI and RuPay debit transactions from 2020 and funded part of the cost through incentives to banks and payment firms. That has been brilliant for adoption. It is not obviously sustainable for the operators who keep the rail reliable at enormous volume. Our view is that a rail with no price is a rail with a quality problem in waiting, and that the sector should settle a clear, transparent way of funding it before the next generation of rails adopts the same habit.
The second is concentration. A handful of apps handle most UPI volume, and the regulator's cap on any one app's share of volume has been postponed more than once. An open rail can still produce a closed market on top of it.
The third is consent. Systems built on data sharing depend on people understanding what they agree to. The Digital Personal Data Protection Act of 2023 sets the legal frame, including a role for consent managers, but consent fatigue is a design problem as much as a legal one. A rail that people click through without reading has not earned its trust.
The fourth is operational. Shared infrastructure concentrates failure as well as efficiency, and an outage on a rail that everyone uses is an economic event. We are watching how new rails build in resilience and dispute handling at the start, since retrofitting them after the volume arrives has been the more expensive path.