Published 8 August 2026 · v1.0.0 · SBI Research, “New Insights from UPI Data” (18 August 2025) · PSS Act Section 10A amendment of 4 August 2026

The shield that came off, briefly

One law change, every merchant category affected differently.

Since 1 January 2020, Section 10A of the Payment and Settlement Systems Act prohibited any merchant discount rate on UPI transactions. On 4 August 2026, the Taxation and Other Laws (Amendment) Bill replaced that blanket ban with a notification power: the government will now specify which payment modes stay fee-exempt, and modes left off the list may lawfully attract a fee. No fee has been notified yet. This piece’s companion article covers the law change and the ₹437 crore incentive scheme in full, and works the fuel-retail case in detail — the most leveraged single category, at 3.87% dealer margin against a full-ticket fee. What follows here is the same arithmetic run across everywhere else UPI is actually spent.

Is UPI heading toward card-network pricing?

Four rows, one question: does a future UPI MDR converge on what Visa, Mastercard and RuPay already charge?

India’s card networks already answer the leverage question this article keeps asking — they have been pricing merchant acceptance for decades, at rates that are either RBI-capped (debit) or market-set (credit). The comparison matters because one payment mode already sits on both sides of the line: a RuPay credit card routed through the UPI interface is not the zero-MDR bank-to-bank rail this blog has otherwise been describing — it already carries a card-style fee, today, no new notification required.

The leverage curve, once more

Fee rate ÷ gross margin is the whole model.

An MDR is a percentage of the ticket; the merchant absorbs it out of a gross margin; so the share of margin an MDR consumes is simply the fee rate divided by the margin. That holds for every counter that accepts UPI, regardless of category:

Where familiar UPI expense categories conventionally sit on that curve — stated as characterisations, not measured margins:

• Thin-margin, regulated tickets sit at the painful end, alongside fuel. PDS kerosene’s dealer commission is 4.5% of its ticket (₹2.65 on ₹59.06 at Mumbai, PPAC build-up) — verified elsewhere on this blog. LPG refills share the structure — a fixed regulated distributor commission on an ₹800-plus cylinder — but the current commission value is not in this author’s PPAC documents, so it is named and not numbered.
• Jewellery is the unregulated twin of the petrol pump: tickets in the tens of thousands of rupees on making-charge margins conventionally in single digits. At a 2–5% margin, a 0.30% fee reads 15% to 6% of margin gone — and a ₹1 lakh purchase generates ₹300 of fee on one swipe.
• Kirana and pharmacy sit mid-curve — conventionally low-double-digit margins put a 0.30% fee at roughly 2–3% of margin: real, recurring, and the reason small-merchant exemptions exist in every MDR debate.
• Services with high gross margins — restaurants, salons, education fees — barely feel the curve: at 25–50% gross margin, even 1% is 2–4% of margin.

Where NPCI says the volume actually is

29 published categories, 82% of P2M volume, and groceries lead by a wide margin.

NPCI has begun publishing merchant-category classification data, decoded by SBI Research in an August 2025 note. It does not contain margins — NPCI does not publish those — but it says where UPI’s person-to-merchant volume and value actually sit, for July 2025:

Food and grocery: the counters most Indians actually stand at

Three categories, 43.7% of UPI’s P2M transaction count.

Groceries, full-service restaurants and fast food are three separate NPCI categories, but they describe one behaviour — buying something to eat, at a counter, at a small ticket. Summed, they dwarf every other category in this table by volume, and their combined fee exposure is where an MDR would actually be felt by the largest number of people, even though no single one of them tops the value ranking:

Two things follow. First, at a ₹179 average ticket, the absolute rupee fee on any single grocery or food purchase is tiny — 45 paise at 0.25% — which is exactly why a flat MDR on these categories would be politically invisible transaction-by-transaction while generating a multi-thousand-crore annual pool in aggregate. Second, groceries, restaurants and fast food are structurally different merchants even though NPCI's data treats them alike: a kirana store's margin is set by wholesale procurement and thin retail markup, while a restaurant's is set by menu pricing against food cost, and the two do not sit at the same point on the leverage curve.

• Groceries run thin. Organised and unorganised grocery retail in India conventionally operates on low-double-digit gross margins — procurement-driven, price-competitive, and the reason wholesale-format retailers exist. On the leverage curve (Exhibit 2), a 10–15% margin puts a 0.30% MDR at roughly 2–3% of margin: the same order of magnitude as kirana in the companion piece, because in practice it largely is kirana. At 3,033 million transactions in one month, groceries is UPI's single highest-volume merchant category of any kind — ahead of fast food, restaurants, and every other row in Exhibit 3.
• Restaurants and fast food are not the same margin story, even though NPCI groups them by volume similarly. Menu pricing conventionally runs food cost at roughly a quarter to a third of the ticket, which reads as a wide gross margin on the ingredient line — but rent, staff, delivery-platform commissions and (for fast food specifically) franchise royalties consume most of that before it becomes operating profit. This article’s framework prices an MDR against gross margin, where restaurants and fast food sit comfortably (Exhibit 2's 25–50% rows); a full-service operator worried about net margin would experience the same fee very differently, and that distinction — gross versus net, and how much of gross a payment fee actually reaches before opex — is not resolved by any data source used in this article.
• Delivery-platform transactions complicate this further and are outside NPCI's category as published. A UPI payment routed through a food-delivery app is likely coded to the platform's own merchant category, not to the restaurant, meaning the restaurant/fast-food rows in Exhibit 3 probably undercount the true food-service transaction volume that ultimately reaches a kitchen. This article does not attempt to size that gap.

The rest of NPCI's top 15

Apparel, electronics, liquor, gaming, marketplaces, and bill payments — six more shapes on the same curve.

• Apparel — clothing shops (₹10,874cr, ₹1,175 ticket) and uniforms/commercial clothing (₹5,969cr, ₹909 ticket) — sit at the comfortable end. Retail apparel conventionally runs 30–50% gross margin, so even a 1% MDR is 2–3% of margin (Exhibit 2's bottom rows). High ticket size plus wide margin is the mirror image of fuel: an MDR here is a rounding error.
• Electronics shops (₹10,068cr, ₹1,093 ticket) sit closer to fuel than to apparel. Consumer-electronics retail is conventionally thin-margin, single digits to low teens, competing on price against e-commerce; at a 5–10% margin a 1% MDR is 10–20% of it. High-ticket, thin-margin, competing on visible price — the same structural profile that makes jewellery and fuel MDR-sensitive.
• Liquor (₹5,920cr, ₹407 ticket) is a regulated-margin category, like fuel and kerosene, but the number is state-set and not in this article’s sources. State excise departments fix retail margins on liquor in most states (structurally comparable to the OMC dealer commission), but the specific commission percentage is state-by-state and not published in a form this article can cite — named, not numbered.
• Digital Goods: Games (₹10,077cr, ₹287 ticket) does not fit the merchant-margin frame at all. In-app and platform game purchases run through app-store or publisher billing with take-rates near 15–30% of the ticket, not a retail cost-of-goods margin — an MDR here would be a fee on a fee, a different question this article does not attempt to answer.
• Online Marketplaces (₹8,054cr, ₹660 ticket) is a similar structural mismatch. The MCC captures payment to the marketplace platform, not to the underlying seller, whose margin (and whose exposure to a fee passed through by the platform) is invisible in this data.
• Utilities (₹23,498cr, ₹1,345 ticket) and government services (₹11,298cr, ₹1,027 ticket) sit outside the margin frame on the other side — these are bill payments and fee collections, not sales with a cost of goods. An MDR on a power bill or a passport fee is not a merchant-margin question; it is a pass-through question, and given the payer is typically a captive household or citizen with no substitute vendor, this is arguably the category most exposed to a flat MDR precisely because there is no market discipline to resist one.

Sources. NPCI merchant-category classification data for July 2025, decoded and published by SBI Research, “New Insights from UPI Data” (Issue 20, FY26, 18 August 2025) — the top-15 merchant category table (Exhibit 3 here) and the 29-of-~300-MCC coverage figures are taken directly from that note; NPCI's own merchant-category pages returned an access error to this author on 8 August 2026 and could not be queried directly. Section 10A amendment of 4 August 2026 and the fuel-retail leverage case — this blog's companion article, which also cites the Taxation and Other Laws (Amendment) Bill reporting, the ₹2,000 crore → ₹437 crore incentive cut, and the January 2017 fuel-MDR standoff in full. PDS kerosene dealer commission (4.5% of ticket) — PPAC Ready Reckoner FY2025-26 H1, as verified in this blog's fuel-pricing piece. The combined groceries/restaurants/fast-food totals, fee-pool arithmetic and weighted average ticket in Exhibit 4 are this author's computation on the SBI Research figures above; all other category margin characterisations (apparel, electronics, liquor, gaming, marketplaces, utilities) are stated as retail-trade convention, not measured data, and are flagged as such throughout.

About this article: Researched, written and edited by Umashankar Triplicane Dwarakanathan, with AI research assistance; every figure is meant to trace to the primary source cited. See the Editorial Policy for how sourcing, AI use and corrections work.