ONDC, three years in: the number that matters isn't the headline

In the fiscal year ending March 2026, the ONDC network processed 218 million transactions. That is a genuine milestone. The network has crossed 10 million monthly transactions, onboarded over 700,000 sellers across 1,200 cities, and expanded from 31 to 72 seller apps in roughly 18 months. Read those numbers in a press release and you would conclude ONDC is working.

It is. But for a different product than the one originally announced.

Open the data, look at the sectors

ONDC publishes transaction volume by sector at opendata.ondc.org. It is public, updated monthly, and broken down by category. The 218 million headline is accurate. The sector split is what most coverage leaves out.

In April 2025, mobility (ride-hailing and auto-booking) accounted for 58% of all ONDC transactions. Six months earlier, its share was 40%. Monthly mobility orders grew from 3.6 million in April 2024 to 9.4 million by April 2025, a 2.6x increase in a year.

Retail moved the other way. Monthly retail orders peaked at 6.5 million in October 2024 and fell to 4.6 million by February 2025. Retail's share of total ONDC transactions dropped from 47% to 29% over those same months.

What drove ONDC's retail decline

Two things happened around the same time: ONDC reduced the financial incentives that had been partly subsidising the network, and the incumbent experience gap became harder to overlook.

On incentives: ONDC's early retail growth was partly subsidy-funded. Sellers discounted more than the unit economics justified because the network covered part of the gap. When ONDC pulled back those incentives in late 2024, order volumes dropped — not because demand disappeared, but because the marginal case for a buyer to choose an ONDC app over Amazon or Flipkart had depended partly on price, not just experience.

A survey of ONDC users in early 2025 found that 54% described the experience as cumbersome. For a network competing against companies that have spent years and billions on UX, that is a difficult number to recover from on organic growth alone. Amazon's checkout, Zomato's real-time tracking, Flipkart's returns flow — these are capabilities that live outside the ONDC protocol. Every buyer app has to build them independently.

Why the UPI comparison doesn't fully hold

The narrative that ONDC will do for ecommerce what UPI did for payments is everywhere in Indian tech circles. The logic is appealing: UPI democratised payments through an open protocol that anyone could route through. ONDC should work the same way for commerce.

The analogy holds at the protocol level. It breaks down structurally.

UPI works because a payment is a cognitively simple job. Enter amount, confirm, done. The protocol carries the entire user task. There is no ambient concern about delivery time, returns policy, or whether the product matches the photo. UPI's "experience" is the payment flow, and that is owned entirely by the buyer app.

Ecommerce is a chain of jobs: discovery, selection, checkout, delivery, post-sale support, returns. ONDC handles checkout, one link in that chain. Everything the customer actually worries about happens elsewhere. Zomato built food delivery by owning logistics end to end. Amazon built ecommerce by owning warehousing, seller support, and search. ONDC specifies the message format for the checkout handoff. It cannot specify what happens around it.

This is not a criticism of the protocol's design. It is a structural fact about what open protocols can and cannot carry. UPI works because the job it solves is narrow. The job ecommerce solves is not.

Why mobility succeeded, and what that tells you

Mobility's trajectory on ONDC is worth understanding on its own terms, because it illustrates the conditions under which the protocol actually creates value.

Apps like Namma Yatri built a real alternative to Ola and Uber by eliminating the aggregator take-rate that incumbents charge drivers. On ONDC, the driver keeps a larger share of the fare. That is a genuine value proposition — not a subsidy, not a discount. It is a structural cost advantage.

Critically, the mobility job on ONDC is simple: match a rider to a driver, confirm a fare, track the journey. There are no SKUs, no delivery windows, no warehouse. The protocol is sufficient for the entire transaction. The buyer app does not need to build inventory management, returns handling, or a fulfilment network. The driver is the inventory.

This suggests where ONDC's strongest long-term opportunities lie. The sectors most likely to scale are those where incumbent take-rates are high, transaction complexity is low, and the economic benefit of removing the aggregator is clear to both buyer and seller. Mobility fits all three. Parts of B2B commodity trade may also. Rural and artisanal commerce, where the barrier to entry on established platforms is high and the seller economics are poor, is another.

The sectors where ONDC is struggling (general retail, grocery, food delivery) are precisely those where incumbents have the strongest moats and the experience gap is hardest to close through protocol design.

What to do with this if you are building

A few honest observations for anyone evaluating an ONDC-based strategy:

The sector composition matters more than the headline volume. If your product touches retail, food, or grocery, you are building in a segment where ONDC organic traction is currently declining. That does not make it a dead end — but it means pricing in the UX investment required to compete without subsidy support.

The open data is underused. Most commentary on ONDC is based on press releases and quarterly statements. The actual transaction data is public, granular, and updated monthly. If you are assessing whether a specific category is growing or contracting, look at it directly before drawing conclusions from narrative.

The buyer app is the product. ONDC's buyer-seller protocol separation means the buyer app experience is the entire product from the customer's perspective. The mobility teams that have succeeded understood this and built around it. The retail buyer apps that defaulted to ONDC as a distribution mechanism without solving experience are the ones seeing volume drop.

B2B and financial services: the less-discussed verticals

The conversation around ONDC focuses almost entirely on retail and food. Two other sectors are worth tracking.

B2B commerce maps well to the mobility pattern: procurement for commodity goods tends to be a higher take-rate, relatively simple transaction. A manufacturer sourcing raw materials or a small retailer restocking inventory does not need the same UX sophistication as a consumer app. The underlying job — find supplier, confirm price, place order — is narrow enough that the protocol is sufficient. Early pilots in agricultural commodities and industrial supplies have shown traction, though volumes are still small.

Financial services is newer. ONDC has been developing a credit flow protocol that connects lenders, loan service providers, and borrowers on a single network. The logic is the same as retail: break the bundling that lets incumbents extract margin. In credit, that margin is substantial. If the credit protocol scales, it may produce more durable volume than the retail story ever did, precisely because the incumbent take-rate in lending is high and the transaction complexity is low enough that protocol-level standards can carry it.

Neither vertical has hit scale yet. But both fit the pattern of sectors where ONDC creates structural value rather than just substituting for an existing UX that incumbents have already optimised.

What comes next is less about the protocol

ONDC has built real infrastructure in three years: genuine transaction volume, a working mobility protocol that created actual competition for Ola and Uber, and a seller onboarding machine that has reached 700,000 sellers in markets incumbents were not serving well.

Whether retail recovers depends less on changes to the protocol and more on whether any buyer app builds an experience that earns loyalty on its own terms — without the incentive subsidies that drove early adoption. The protocol is infrastructure. Infrastructure has been enough for mobility because the job was simple and the incumbent was extractive. In retail, neither condition is as clean.

The builders worth watching are not the ones asking whether ONDC will become the UPI of ecommerce. They are the ones asking: in which specific verticals is the incumbent take-rate high enough, and the transaction simple enough, that the protocol advantage outweighs the experience gap? In those verticals, ONDC may build something durable.