How to Choose a Payment Gateway in India

Payment gateway selection in India usually happens once, early, on the basis of a percentage in a pricing deck, and then never gets revisited. That is backwards. The percentage is the least important number on the page.

Success rate beats headline pricing, by roughly twenty times

Run it on a brand doing Rs 1 crore a month of attempted prepaid volume. Negotiating the card rate from 2.0 per cent down to 1.8 per cent saves Rs 20,000 a month. Moving payment success rate from 84 per cent to 88 per cent adds Rs 4 lakh a month in captured revenue. Same effort, twenty times the return.

Success rate is successful transactions divided by initiated transactions, and it has to be read per instrument. The blended figure in a sales deck hides everything that matters.

• UPI intent, where the payment app opens through a deep link, typically runs 10 to 20 points better than UPI collect, where the buyer must leave your flow and find a pending request. If your mobile checkout still defaults to collect, that is the highest-value fix available to you today.
• Card success varies sharply by issuer. Some bank authentication pages degrade under load, which is precisely when your festive traffic peaks.
• Netbanking success at smaller banks can sit below 70 per cent. If netbanking is 8 per cent of your mix, that is a visible and fixable leak.

Ask every gateway on your shortlist for last-30-day success rates by instrument and by top issuer, for merchants in your category. A provider that cannot produce that report cannot help you improve the number either.

Pricing, per instrument, after GST

Published domestic rates cluster between 1.75 and 2 per cent plus 18 per cent GST. Razorpay lists 2 per cent as its standard domestic rate. Cashfree publishes rates in the 1.75 to 1.95 band, with an annual maintenance charge on some plans. PayU largely negotiates. UPI is currently zero-rated across all of them. Remember that GST turns a 2 per cent rate into an effective 2.36 per cent.

• Flat-fee card pricing, a fixed rupee amount per domestic card transaction, is available and beats percentage pricing above a break-even ticket. Against 2 per cent, a Rs 9 flat fee breaks even at Rs 450. At a Rs 1,800 AOV that is a saving of about 1.5 points on every card order.
• International cards run 3 to 4 per cent plus a fixed component. If exports are 5 per cent of orders they can be 15 per cent of total payment cost. Price them as a separate line.
• EMI, pay-later and BNPL carry subvention or a higher rate. Confirm in writing whether you or the lender absorbs it before switching the option on at checkout.

Then ask for the effective blended rate against your actual instrument mix from last quarter. Not the card rate. Not the best-case rate.

Settlement, reserves, and refunds

Standard settlement is T+2 working days. New merchants frequently start at T+7, and in categories a gateway scores as elevated risk, including supplements, nutraceuticals, claims-heavy cosmetics and electronics, expect a rolling reserve of 5 to 10 per cent held for 30 to 180 days. Raise reserves during the sales conversation, not after your first settlement lands short.

T+1 and on-demand settlement exist for an extra 0.1 to 0.2 per cent. Price that against your cost of capital rather than your anxiety. At 18 per cent a year, one day of money is worth about 5 basis points. Paying 15 basis points to pull cash forward by one day is a bad trade. Paying the same 15 basis points to pull it forward five days is a clear win. Do that sum before you buy the feature.

Working days matter more than the T+n label implies. A Friday evening sale settles Tuesday. A long festive weekend can stretch that to five calendar days, at exactly the point your ad spend is peaking and your inventory reorders are due.

On refunds, establish three things in writing. Whether refunds draw from your settlement balance or are debited separately. Whether the original processing fee is returned, which it usually is not. And how quickly the customer actually sees money back, because slow refunds convert directly into disputes and support tickets.

Routing across two gateways

The case for a second gateway is uptime and issuer-level performance, not price. One outage during a festive hour costs more than a full year of fee difference. A secondary also lets you route traffic to whichever acquirer performs better for a given issuing bank.

The cost is reconciliation. Two settlement formats, two refund states, two webhook streams. You need idempotent order identifiers, webhook deduplication, and one system of record for refund status. Do not attempt multi-gateway routing without a named owner for reconciliation. Also check that splitting volume does not push you below the negotiated volume tier on both contracts at once.

Start with failover, not weighted routing. Primary gateway, secondary triggered on failure or timeout. Add rules later, once your instrument-level data is trustworthy enough to route on.

What matters at each stage

• Under Rs 50 lakh a month. One gateway, published pricing, T+2. Spend your time on checkout, not contracts. Get UPI intent working on mobile, card tokenisation saved for repeat buyers, and a clean retry screen after a failed attempt.
• Rs 50 lakh to Rs 3 crore a month. Negotiate. Card rates of 1.6 to 1.8 per cent are realistic at this volume. Push for T+1. Ask for instrument-level dashboards and a named support contact with a response SLA during sale events.
• Above Rs 3 crore a month. Two gateways with failover, negotiated rates, issuer-level reporting, and a monthly review of success rate by instrument and issuer. At this volume, one point of success rate outweighs the entire fee negotiation.

Subscription brands add one requirement at every stage: e-mandate coverage across banks and cards, with a compliant pre-debit notification at least 24 hours before each charge. Mandate success rate is a separate metric from checkout success rate, and it decays quietly over time. Review it every quarter.