The listed accounts agree with the recommendation: distribution earns, transactions do not

The answer

The report has argued that the investable business here is distribution rather than transactions. Four listed companies file accounts that test it, and they agree.

In FY2026 the transaction business, One97 (Paytm), turned its first profit, at 6.8% net margin. It did so on revenue 17.0% below its own FY2024 peak, which is the shape of a business that got better by getting smaller. In the same year Angel One earned 21.2% and PB Fintech 9.9% selling distribution, and CDSL, which operates the piece of market infrastructure that is permitted to charge a toll, earned 39.8%. Every distribution business in the panel earns a wider margin than the transaction business.

The ordering is the finding: the further a business sits from the transaction and the closer to the relationship or the toll, the wider the margin.

Three supporting arguments

1. The rail is the thinnest margin in the panel, at its best year yet. One97 (Paytm) carries the largest revenue of the four at Rs 8,192 crore and converts the least of it, 6.8%, against 21.2% at the best distribution business: a gap of 14.5 percentage points. Scale is not the constraint. Price is, which is what sub-module A said from the volume side and what sub-module H priced at 63.9% of transactions earning nothing.

2. Infrastructure that may charge a toll earns like infrastructure that may charge a toll. CDSL runs on the smallest revenue in the panel, Rs 1,145 crore, and returns 39.8%. It is the control case for the whole report: identical in kind to a payment rail, a piece of plumbing everyone must cross, and the difference in outcome is that its crossing has a price. UPI’s does not.

3. Distribution scales without the rail’s cost base. Both distribution businesses in the panel earn more per rupee of revenue than the transaction business while running a fraction of its revenue. That is the merchant-side flywheel sub-module A recommended, visible in filed accounts rather than argued from first principles.

So what

• For an operator: the pathway ranking in the diligence memo survives contact with audited accounts. Build toward the products the payment relationship carries, and treat the transaction as the acquisition cost it is.
• For an investor: the transaction business is the one that took the longest to earn anything and earns least now. Underwrite the distribution attached to it, not the volume running through it.
• For a policymaker: a rail whose only listed operator reached breakeven this recently is not a sector that can absorb a further pricing cut. Sub-module H already shows the state’s own subsidy thinning by arithmetic.

Method and its limits

Four filed annual income statements, retrieved through the same library the banking module already uses. Margins are computed from revenue and reported net income; nothing is estimated and no figure is typed in.

What this cannot support. These are not like-for-like companies: different revenue recognition, different perimeters, different regulators, and one of them is a regulated monopoly. Only the shape of the margin each model produces is compared, never a ranking of management. It is a panel of named companies, not cohort averages, because two of the three groups hold a single company and an average of one is not an average. Four companies also cannot stand in for a sector, and a bad year at any of them would move the comparison, which is why the verdict above is computed rather than asserted. Finally, this module says nothing about consumer credit, which remains the largest gap on the coverage map: distribution here means insurance and broking, the two that happen to be listed as near-pure plays.