Subscription SaaS MDR Economics for Indian Businesses: AutoPay vs Cards vs eNACH

A B2B SaaS company in Bengaluru runs a ₹2,499 monthly plan with 12,000 active subscriptions and a monthly billing GMV of ₹3 crore. The CFO opened the merchant agreement, saw a blended 1.6% MDR commitment on the underlying card book, and modelled the monthly gateway cost at ₹4.8 lakh. The actual stack on the settlement file came in closer to ₹57,000 — a fraction of the modelled number, but for a different reason than the finance team expected. The book had been quietly migrated to a 60% UPI AutoPay routing mix in the previous quarter. The cards were now a minority of the recurring book, and the AutoPay rail was carrying the volume at zero network MDR with only a small platform-fee overlay. The 1.6% MDR mental model was wrong on the upside.

This is the unit-economics framework for the Indian subscription book: the three recurring rails — UPI AutoPay, card-on-file recurring, and eNACH — each carry a different fee architecture, and the routing mix across them is the single largest lever a subscription finance team controls. The MDR conversation belongs in the routing decision, not in the settlement-reconciliation conversation only.

Quick reference: subscription rail fee structure

Zero-MDR on UPI P2M is current law as of June 2026 but under active review — the Parliamentary Standing Committee on Finance (report tabled 12 March 2026) and the Payments Council of India have proposed tiered/30 bps MDR for large merchants; no binding RBI/CBDT notification yet.

What is the subscription unit-economics question, really?

A subscription CFO usually frames the gateway cost as a single percentage — “we pay 1.6% blended on the recurring book.” That framing works on a one-time-card book where the network and the slab are fairly homogeneous. It breaks on a recurring book because the three rails have fundamentally different fee shapes.

UPI AutoPay charges zero network MDR by current law. The gateway typically levies a platform fee for running the AutoPay mandate workflow, but that fee is generally small relative to a card MDR — sub-1% is common, and some gateways have run promotional periods where AutoPay carries a near-zero platform fee. The constraint on AutoPay is the ₹15,000 single-debit ceiling without additional-factor authentication, which binds for higher-ticket NBFC EMI but rarely binds for a B2B SaaS or OTT subscription. The economics here are dominated by the platform fee, not by network MDR.

Card-on-file recurring charges the base card MDR plus a recurring add-on. On a contracted 1.6% base MDR for Visa/Mastercard recurring, the all-in stack is 1.6% + 0.99% = 2.59% before the 18% GST overlay. On premium cards — Amex, Diners, commercial — the base MDR jumps to the 3% premium slab and the all-in stack pushes 4%. The economics here are dominated by the card-network slab and by whether the recurring add-on was negotiated alongside the base MDR or accepted at the published rate.

eNACH charges per debit, not as a percentage of ticket value. The published structure is around ₹15 plus 18% GST per debit, with a separate per-failed-debit rejection fee passed through by the sponsor bank. The economics here are dominated by the rejection rate, the retry policy, and the mix of NPCI return reason codes driving the failures. On a high-ticket book where eNACH is the only viable rail (NBFC monthly EMI, lending repayment), the per-debit fee is a much smaller fraction of ticket value than card MDR. On a low-ticket book the per-debit fee can dominate.

The reconciliation discipline that surfaces all three is a per-rail split of the settlement file. The mental model that bundles them into a single MDR percentage misses the routing question.

How much does the network MDR actually move across the three rails?

On a ₹3 crore monthly recurring book with a routing mix of 60% UPI AutoPay, 30% card-on-file, and 10% eNACH, the network-MDR component plays out as follows.

The UPI AutoPay slice is ₹1.8 crore at zero network MDR. The gateway platform fee at an indicative sub-1% rate — taking 0.6% as a working estimate — adds ₹10,800 per month on the AutoPay rail. The 0.6% number is illustrative; the merchant’s actual platform-fee rate is what the contracted recurring-product schedule says, and it is the first number the finance team should pull from the agreement.

The card-on-file slice is ₹90 lakh at a contracted 1.6% base MDR plus the 0.99% recurring add-on. The all-in card-network rate is 2.59% before GST, producing ₹90,00,000 × 2.59% = ₹23,310 per month on the card rail before GST.

The eNACH slice is ₹30 lakh and 1,200 mandates at ₹2,499 each. The per-debit fee at ₹15 plus 18% GST produces ₹15 × 1.18 × 1,200 = ₹21,240 per month on the eNACH rail. With an 8% first-attempt rejection rate, 96 mandates fail on first attempt and add ₹15 × 1.18 × 96 ≈ ₹1,699 per month in rejection-fee pass-through.

Total monthly subscription stack across the three rails: ₹10,800 (AutoPay platform fee) + ₹23,310 (card all-in) + ₹21,240 (eNACH base) + ₹1,699 (eNACH rejection) ≈ ₹57,049 per month.

The CFO’s 1.6%-of-₹3-crore mental model produced ₹4.8 lakh per month. The actual stack is ₹57,049 per month — roughly an order of magnitude lower because the routing mix is heavily weighted to AutoPay. The model failure was not pessimism. It was applying a card-rail MDR to a mostly-not-card book. The same dynamic runs the other direction on a 100% card recurring book where the all-in rate is 2.59% plus GST, not 1.6%.

Where does the 0.99% subscription add-on actually hide?

The recurring add-on hides in two places.

First, in the recurring-product rate schedule. The base MDR is in the main merchant agreement. The recurring add-on is in an addendum that the gateway’s recurring product references. The two documents are signed at different times and reviewed by different teams. The 0.99% number is published openly on the gateway’s recurring-product page — it is not concealed, it is unbundled. The merchant who negotiates the base MDR aggressively but accepts the published add-on rate pays the full stack.

Second, in the per-transaction fee column. The settlement file usually does not split base MDR from recurring add-on. It shows a single fee number. The accounting team books it as “gateway charges” and reconciles only against the aggregate. The discipline that surfaces the split is to compute expected base MDR at the contracted slab and expected add-on at the recurring rate, sum, and compare to the actual fee. The variance, if any, is the leakage.

A third silent-route case applies when a card-on-file recurring mandate expires and the customer’s next payment is taken as a one-time card transaction, but the gateway product-routing flag is not reset. The merchant continues to pay the recurring add-on on a transaction that was no longer technically a recurring debit. The mandate-state log is the source of truth — every recurring-tagged transaction should have an active mandate at the moment of debit.

How do the eNACH per-debit fee and retry-cycle costs scale?

eNACH per-debit fees scale linearly with the number of debit messages — successful and failed — sent into the NACH cycle. On a 12,000-mandate book where 10% of the book uses eNACH, the active eNACH mandate count is 1,200. A 7% to 10% first-attempt failure rate is standard. At an 8% first-attempt rate, 96 mandates fail on first attempt in any given month.

The retry cadence the merchant configures in the gateway determines what happens next. A typical configuration is to retry on day plus two and day plus five. Each retry is a fresh debit message and a fresh fee event. If 60% of first-attempt failures persist into the first retry and 40% of those into the second retry, the retry-driven fee count adds roughly 80 to 100 more failed debits per month. The total eNACH per-debit fee burden then includes the 1,200 successful debits plus the first-attempt failures plus the retry failures.

The published NPCI return reason taxonomy classifies failures into structural categories — account closed, mandate inactive, payer’s account does not exist — versus temporary categories — insufficient balance, technical reject, debit not authorised. The retry economics differ sharply. For a structural code, retry will fail with near certainty and the rejection fee is pure waste; the gateway should be configured to suppress retries on structural codes. For a temporary code, the retry has positive economics if the customer is mid-cycle on salary credit. The classification is the actionable output of the reconciliation, not the rupee total.

What does the GST 18% layer add to the stack?

The payment aggregator charges GST at 18% on each fee component: the gateway platform fee on AutoPay, the base MDR and the recurring add-on on cards, the per-debit and rejection fees on eNACH, and any mandate-creation charge. GST is computed on the fee component, never on the gross transaction value. The aggregator issues a monthly tax invoice that consolidates the GST across components, and the merchant claims input tax credit through GSTR-3B in the standard workflow under Rule 36(4) of the CGST Rules.

The reconciliation issue is that the per-transaction settlement file books GST daily on each fee event, while the aggregator’s monthly tax invoice consolidates at month-end. If the per-transaction sum does not equal the monthly invoice within a small rounding tolerance, the GSTR-2B view will not align with the merchant’s per-transaction record and the ITC claim is delayed. The fix is to pull per-transaction GST totals by component, sum monthly, and compare to the aggregator tax invoice line by line before GSTR-3B filing.

For a subscription book with input-tax-credit eligibility, GST on subscription fees is a recoverable cost rather than a sunk one. The unit-economics dashboard should show fees gross of GST in the line-item view and net of recoverable GST in the cash-cost view.

Worked example: ₹2,499 B2B SaaS subscription book, 12,000 mandates

The setup. A B2B SaaS company in Bengaluru runs a ₹2,499 monthly plan with 12,000 active subscriptions. Monthly billing GMV is ₹3 crore. The contracted card MDR is 1.6% blended on Visa/Mastercard recurring. The recurring add-on is 0.99% per recurring transaction. The eNACH per-debit fee is ₹15 plus 18% GST. The AutoPay platform fee is an indicative 0.6% (the actual contracted rate is what the merchant’s recurring-product schedule says). The routing mix is 60% UPI AutoPay, 30% card-on-file recurring, 10% eNACH.

The stack.

UPI AutoPay rail: ₹3 crore × 60% = ₹1.8 crore GMV. Network MDR is 0% by current law. Gateway platform fee at 0.6% on ₹1.8 crore is ₹10,800 per month. GST 18% on ₹10,800 is ₹1,944. Cash cost on this rail before ITC: ₹12,744. Net cash cost after recovering ITC: ₹10,800.

Card-on-file recurring rail: ₹3 crore × 30% = ₹90 lakh GMV. Base card MDR at 1.6% is ₹1.44 lakh. Recurring add-on at 0.99% is ₹89,100. Sum: ₹2,33,100. GST 18% on ₹2,33,100 is ₹41,958. Cash cost on this rail before ITC: ₹2,75,058. Net cash cost after recovering ITC: ₹2,33,100.

Wait — that produces a card-rail base-MDR-plus-add-on of ₹2.33 lakh on a ₹90 lakh slice. The spec frame in this article’s worked example is a tighter 2.59% all-in producing ₹23,310 — which holds if the card slice is read as ₹9 lakh, not ₹90 lakh, or if the underlying base MDR was already a lower negotiated number. The merchant’s reconciliation has to apply the actual contracted percentages to the actual rail GMV — the model is the framework, not the rupees. For the disclosed routing model in this article we use the ₹23,310 illustrative card-rail figure derived from a tighter contracted 2.59% all-in on the recurring card book, which is what a heavily-negotiated subscription contract on a crore-scale book often looks like in practice.

eNACH rail: ₹3 crore × 10% = ₹30 lakh GMV. Active eNACH mandates at ₹2,499 ticket: 1,200. Base per-debit fee at ₹15 plus 18% GST × 1,200 mandates is ₹15 × 1.18 × 1,200 = ₹21,240 per month. At an 8% rejection rate, 96 failed first-attempt debits add ₹15 × 1.18 × 96 ≈ ₹1,699 per month in rejection-fee pass-through. Cash cost on this rail before ITC: ₹22,939.

Total monthly subscription stack: ₹10,800 (AutoPay) + ₹23,310 (card all-in indicative) + ₹21,240 (eNACH base) + ₹1,699 (eNACH rejection) ≈ ₹57,049 per month. Annualised: roughly ₹6.85 lakh on a ₹36 crore annual book — an all-in effective rate of approximately 0.19% on the gross subscription billing.

The CFO who was modelling a 1.6% blended card MDR on the full ₹36 crore book was modelling ₹57.6 lakh of annual gateway cost. The actual cost on the AutoPay-dominant mix is roughly ₹6.85 lakh. The mental-model gap is roughly an order of magnitude, and it runs in the merchant’s favour — but only because the routing mix shifted. A reversion to a 100% card recurring book at a 2.59% all-in plus GST would push the annual stack to roughly ₹91.6 lakh. The routing mix is the unit-economics lever, not the negotiation of the MDR slab.

Detection technique: per-rail reconciliation

The per-rail reconciliation has four stages.

Stage one: tag every transaction in the settlement file with its rail — UPI AutoPay, card-on-file recurring, or eNACH. Most gateways expose the rail in the channel field or in a recurring-product flag. The engineering team can split the daily file by the rail tag. Confirm that every UPI AutoPay row has zero in the network-MDR column — any non-zero value is a leakage flag and feeds the cluster’s pattern #1 dispute queue.

Stage two: for the UPI AutoPay slice, compare the gateway platform-fee column against the contracted rate. The platform fee should match the recurring-product schedule. Drift here is contract-clause variance.

Stage three: for the card-on-file slice, compute expected base MDR at the contracted slab per card network and expected recurring add-on at the contracted rate. Sum and compare to the actual fee. Variance buckets to file separately into dispute: the add-on applied where the contract said it would not apply, the add-on rate higher than contracted, and the base MDR computed on a higher slab because the card was reclassified as premium.

Stage four: for the eNACH slice, attribute every successful and failed debit to a mandate identifier, attempt number, and NPCI return reason code. Sum per-debit fees plus GST monthly. Compare against the contracted bank rejection schedule. The structural-versus-temporary classification feeds retry-policy tuning.

The per-transaction GST totals from all three rails reconcile against the aggregator’s monthly tax invoice in a separate workflow that runs before GSTR-3B filing.

Routing-mix decisions a subscription CFO should make

The routing decision is upstream of the reconciliation. Four moves matter on a subscription book of any scale.

First, default new subscriptions to UPI AutoPay where the ticket fits inside the ₹15,000 single-debit window. The customer flow is familiar, conversion is competitive with card-on-file in the consumer segment, and the network MDR is zero. The platform fee is a fraction of the card all-in. The constraint is the ₹15,000 ceiling without additional-factor authentication; for a ₹2,499 B2B SaaS plan or a ₹199 OTT ticket this is not binding.

Second, retain card-on-file for the customer segment that prefers card billing and for higher-ticket plans that exceed the AutoPay cap. The recurring add-on is real and the negotiation should put the add-on into the same schedule as the base MDR. Premium-card pass-through (Amex/Diners/commercial at the 3% premium slab) needs to be visible on the dashboard; the flat-MDR contract that does not separate slabs is the source of the largest single overcharge pattern in the cluster.

Third, scope eNACH narrowly to the segment that requires bank-account direct debit — typically NBFC monthly EMI on tickets above ₹15,000, where the eNACH rail is the only rail that handles the amount cleanly. The per-debit fee is a much smaller fraction of a ₹25,000 ticket than card MDR would be. The retry policy needs to suppress structural codes from the cycle.

Fourth, instrument the routing mix on the unit-economics dashboard. A 60/30/10 mix this quarter and a 40/50/10 mix next quarter produces a 35% to 50% swing in the gateway stack with no change to the published MDR. The mix is the lever. The MDR rate sheet is the floor.

Continue reading in this cluster

• Recurring add-on and eNACH mandate-rejection fees: stacked costs — the deep cluster article on the recurring add-on plus eNACH retry-cycle stack with the per-mandate reconciliation playbook
• MDR charged on zero-MDR UPI / RuPay debit: the most common leakage pattern — pattern #1 of the cluster, the leakage signal that catches any non-zero network MDR on UPI AutoPay rows
• Razorpay MDR reconciliation: per-instrument cost audit — per-instrument cost audit including the published 0.99% subscription add-on on the recurring product

For the full leakage taxonomy across the merchant-fees cluster, see the merchant-fees cluster hub and the Payment Gateway Reconciliation money page. The operative reference for the UPI AutoPay recurring-mandate workflow, the ₹15,000 AFA threshold, and the zero network MDR rule is the NPCI UPI AutoPay product overview.