पाठशाला Pathshala · विचार Vichār, The idea · Lesson 07 · Start

India Stack as a market map: UPI, Aadhaar, ONDC, Account Aggregator

Public digital rails make a category of business nearly free and a different category newly possible. Read the four big ones as a map: build where the rail stops, not where it runs.

Pathshala, The Founder Library · 11 October 2026 · 10 min read

In most countries the plumbing of a digital economy is owned by companies, and a founder pays rent to it: a few per cent of every payment, a fee for every identity check, a cut of every order found through a platform. India built a good part of that plumbing as public infrastructure and priced it at or near zero. That decision has made some businesses impossible to charge for and others possible for the first time. A founder who reads the rails as a map knows which is which before writing a line of code.

This lesson covers the four rails a founder will meet first: Aadhaar, UPI, the Account Aggregator framework and ONDC. For each it gives the scale as of the latest published figures, the rule for what the rail commoditises and what it creates, and the business types on each side of the line. The figure in the middle lets you walk the map rail by rail. The closing section is the trap list, because rails have owners and owners change the rules.

A public rail takes a function that used to be scarce and expensive, does it for everyone at a price close to zero, and does it the same way for everyone. Two things follow. First, nobody can build a business on charging for that function; the margin is gone and will not return. Second, everything that was impractical because the function was expensive becomes practical at once, and that is where the new companies live. The first Aadhaar number was issued in September 2010 and one crore had been issued within nine months; the architecture has been argued for in public ever since by Nandan Nilekani, who led the project, and the [thinkers page](/library/thinkers) has more on him.

The map is therefore drawn with one question per rail: what did it make free, and what does that make newly possible? The answer to the first question is the list of businesses not to start. The answer to the second is the list to study.

UPI: payment is free, so payment data is the product

The scale first, because investors will ask. The Press Information Bureau’s August 2026 factsheet puts UPI at 2,365.8 crore transactions worth ₹29.87 lakh crore in July 2026, with 741 banks live, growth from about 2 crore transactions in FY 2016–17 to over 24,162 crore in FY 2025–26, and roughly 70 per cent of all digital payment transactions in the country. There is no other payments rail on earth at that volume with no charge to the merchant.

The zero is a policy, not an accident. The government made the merchant discount rate on UPI and RuPay debit transactions zero from January 2020 and has since paid banks an incentive from the budget to carry small-merchant transactions, ₹1,500 crore for FY 2024–25 at 0.15 per cent on person-to-merchant payments up to ₹2,000. The consequence for founders is blunt: payment acceptance is not a business in India. Payment gateways survive on cards, on enterprise integration and on value added around the transaction, not on UPI. Wallets lost their reason to exist. And the two largest UPI apps together hold most of the volume, with NPCI’s rule capping any one third-party app at 30 per cent of volume deferred to the end of 2026 and still unenforced; nobody should plan a company around being the next one.

What UPI makes possible is everything that assumes money can move for free at any ticket size and leave a record. A ₹20 payment is now worth collecting, so pay-per-use pricing works for the first time in a market that could never afford card economics on small tickets. Every kirana, auto driver and tuition teacher now has a verifiable sales record in a bank account rather than a notebook, and that record is lending data, insurance data and working-capital data. The cash economy that was unbankable because its transactions were invisible is visible. Those are the businesses to study: not moving the money, but what the movement reveals.

Aadhaar: identity is free, so the first session is where the product begins

UIDAI reported 231 crore authentication transactions in November 2025 alone, 47.19 crore of them e-KYC, and 28.29 crore face authentications, more than double the year before. Each of those replaced a photocopy, a signature, a courier and a person checking the two against each other.

Identity verification was a business before Aadhaar and is a feature after it. A KYC startup in 2026 is a module inside someone else’s onboarding flow; the verification teams that banks and telecom companies once ran are gone. That is the commoditised side. The possible side is the compression of time: a customer can be verified, onboarded and transacting in a single session on a phone, which means the first session is the product rather than a form, and the companies that win design for that. It also means a person with no paper trail can open an account, get a SIM or receive a benefit on the strength of being who they are, which opened the bottom of the market to everyone willing to serve it. The rules on how private companies may use Aadhaar authentication have changed more than once and will again; a founder building on this rail reads the current UIDAI circulars before the business plan, not after.

Account Aggregator: data moves with consent, so underwriting costs rupees

The Account Aggregator framework rests on RBI directions issued in September 2016 and went live on 2 September 2021 with eight banks. An aggregator is a licensed non-bank through which a customer consents to their bank, deposit, insurance or securities data being shared with a lender or adviser for a stated purpose and period. The directions are explicit that the data shall not reside with the aggregator and shall not be its property: the aggregator is blind by design, which is what makes the framework trustworthy and also what makes the aggregator itself a low-margin utility. By 31 August 2026 Sahamati’s dashboard showed 33.8 crore accounts linked and 56.6 crore consents fulfilled.

What the rail commoditised: the upload-your-statement step, PDF parsing, screen scraping, and the quiet advantage a large bank had from being the only one that could see a customer’s cash flows. What it created: lending underwritten on real cash flows for a few rupees of API calls, which makes a ₹50,000 working-capital loan to a shop viable where a few thousand rupees of manual underwriting once killed it; personal finance, tax and insurance products built on what the customer actually has rather than what they remember; and credit for people with thin bureau files and thick bank statements, which in India is most people. The [why-now lesson](/library/why-now-timing-argument) works this example in detail. The point here is where the line falls: do not build the aggregator, build the lender, the adviser or the underwriting engine that uses one.

ONDC: the attempt to do to discovery what UPI did to payment

ONDC is an open protocol rather than a marketplace. A buyer on one participating app can find and order from a seller on another, with logistics, payment and cataloguing supplied by other participants; the network does not own inventory or operate a store. The PIB’s backgrounder describes thirteen domains, from grocery and food to mobility, credit and skilling, and sellers in more than 616 cities as of January 2024. By July 2024 the network was processing about 1.2 crore orders a month, a third of them mobility, with around 6.3 lakh sellers and service providers on board. Set against a single large marketplace that is small. Set against ONDC’s own figure of two years earlier it is a steep curve, and the direction is the thing to watch.

If the network reaches scale it commoditises the one asset a marketplace defends hardest: the customer’s attention at the moment of search. A seller app for a single trade or city would no longer need to build demand, because demand would arrive from every buyer app on the network. Buyer apps for niches the large platforms ignore would be possible with the catalogue supplied. Logistics, credit and cataloguing could be sold as network services to thousands of small sellers at once. All of that is conditional, and the honest version of a pitch built on ONDC says so: here is the monthly order number, here is the number at which our model works, here is what we do until it gets there. A founder who writes that paragraph is taking a position on a rail. One who writes “ONDC will unbundle commerce” is writing a forecast.

A public rail is not a market. It is the reason a market next to it now exists.

A worked example: mapping one idea across the rails

A team in Coimbatore wants to serve the small textile units around the city, which sell on credit to traders and wait sixty to ninety days to be paid. The idea is invoice financing. Map it. UPI: the unit’s receipts from traders now land in a bank account rather than as cheques, so the repayment history exists; the rail also means a ₹8,000 collection costs nothing to make. Aadhaar: the proprietor is onboarded on the phone in the first session, with no branch and no courier. Account Aggregator: with consent, the lender reads eighteen months of the unit’s actual cash flows in minutes and underwrites on them, which turns a product that needed a field officer into one that needs an API and a credit model. ONDC: not relevant today, and the map says so; if the units begin selling through the network, their order data becomes a fourth source.

Now the commoditised side of the same map, which is the list of things not to build. Not a payments layer for the units: UPI is free and already there. Not a KYC product for lenders: it is a feature. Not a statement-aggregation tool: the framework exists and is regulated. The company is the lender and the underwriting model, and its edge is the relationship with the textile cluster and the data that accumulates from lending to it, neither of which any rail provides. The rails lowered the cost of everything around the business to nearly nothing; what is left is the business.

The traps of building on a rail

Building the rail’s next feature. If the thing you are building is an obvious extension of what the rail does, assume the rail will do it. NPCI, UIDAI and the aggregator ecosystem all add capabilities every year, and a company whose entire product is one of those capabilities is a feature with a cap table.

Depending on a price that is a policy. Zero MDR is a government decision renewed annually with a budget line. The 30 per cent cap is an NPCI rule whose deadline has already been extended by two years. Aadhaar usage rules have been rewritten by the courts and the regulator. A business whose margin exists only because a policy is set a particular way has a why-now with an expiry date; know the date, and know what the business looks like if the policy flips.

Mistaking the rail’s scale for your market. Two thousand crore UPI transactions a month are not your addressable market; they are the reason your market can exist. The [market sizing lesson](/library/market-sizing-an-investor-will-believe) applies in full. Count your customers and your price, and use the rail’s number only to show the customers can now be reached.

Betting on a rail before its number arrives. ONDC in 2026 is the live case. The architecture is sound, the backers are serious, and the monthly order figure is still small against the incumbents. A company that needs the network to be large today is early, and early is a polite word for a cash problem. Build the version that works at today’s volume and is positioned for tomorrow’s.

A quarterly ritual: the rails sheet

Keep one sheet with a row per rail your business touches: the metric that matters to you, its source, this quarter’s value and last quarter’s, and the policy that could change it with the date it is next up for review. UPI monthly volume from NPCI or PIB; e-KYC volume from UIDAI; accounts linked and consents from Sahamati; ONDC monthly orders from DPIIT or the trade releases. Update it on the first working day of each quarter. Then write two lines under it: what the rails made cheaper for us this quarter, and what they made cheaper for our competitors. If the second line is longer than the first, the rail is doing more for the market than for you, and the strategy needs a sentence about what you own that the rail does not.

Nothing here is legal, regulatory or investment advice. The rails are governed by NPCI, UIDAI, RBI and DPIIT rules that change; the figures were checked in October 2026 against the sources below, and the lists of what is possible and what is commoditised are this library’s reading, not anyone’s official position.