Most people choose a payment gateway by putting three rate cards side by side and picking the smallest percentage. It is the wrong comparison, and the gap it optimises for is usually worth less than a tenth of a percent of revenue.
Four things matter more, and all four are knowable before you sign anything: when the money actually reaches your bank, what proportion of attempted payments succeed, whether the provider holds the RBI authorisation it needs, and how exposed your particular payment mix is to the change coming to UPI pricing.
That last one is not hypothetical any more. On 4 August 2026 the Taxation and Other Laws (Amendment) Bill, 2026 was introduced in the Lok Sabha, amending Section 10A of the Payment and Settlement Systems Act, 2007. Section 10A is the provision that has made merchant discount rate zero on UPI and RuPay debit since January 2020. The amendment does not impose a fee by itself, but it removes the blanket statutory bar on one.
Everything below is what we work through with a client before wiring a gateway into a store. Rates are as of August 2026 and you should confirm them against current rate cards before budgeting, because they move. The Shopify and WooCommerce comparison covers the related question of what Shopify and WooCommerce each add on top.
The advertised rate is not what you pay
What payment gateway charges actually amount to is never the advertised percentage. Every Indian gateway quotes a headline rate and then charges several things that are not in it. Working out the real number takes about ten minutes and it changes which provider looks cheapest more often than you would expect.
MDR, the part that is advertised
Merchant discount rate, taken as a percentage of the transaction. As of August 2026 the major domestic gateways cluster tightly: Razorpay's standard card rate is 2%, PayU quotes 1.95% to 2.25% on cards, and Cashfree sits around 1.90% to 2.10%. Netbanking and wallets land near 2% at all three.
GST on the MDR, which is not
Eighteen percent, applied to the fee rather than the transaction. A 2% MDR is therefore 2.36% in practice. This one line is larger than every difference between the three providers above put together, and it is the line people leave out of comparison spreadsheets.
Flat per-transaction fees
UPI carries no MDR today, but that does not always mean free. Cashfree and PayU both apply a flat charge of about two rupees per UPI transaction as a platform or technology fee. On a low average order value that flat fee is a larger percentage than a card MDR would have been.
Setup, annual maintenance and the rest
Setup fees, annual maintenance charges and technology fees appear on some plans and not others. They are negotiable, they are frequently waived, and you will not be offered the waiver unless you ask.
International cards are a separate rate and a much larger one. Razorpay quotes 3% on international cards and PayU around 3.5%, again before GST. If even a tenth of your orders come from outside India, that line deserves its own row in the model rather than being averaged into the domestic rate. None of this is a build cost, incidentally: it is a running cost, and it is the one people leave out when they read a quote for the ecommerce website itself.
The three main providers, side by side
| Provider | Domestic cards | UPI | International cards | Settlement |
|---|---|---|---|---|
| Razorpay | 2% | No MDR, platform fee applies | 3% | T+2 |
| PayU | 1.95% to 2.25% | No MDR, about Rs 2 flat | 3.5% | T+2 to T+3 |
| Cashfree | 1.90% to 2.10% | No MDR, about Rs 2 flat | Quoted on request | T+1 |
| All three | Add 18% GST | Add 18% GST | Add 18% GST | Faster is chargeable |
Rates as reported in August 2026, standard published plans, before negotiation. Read across rather than down: the card column is nearly identical between all three, and the columns where they genuinely differ are the last two.
Build the comparison on your own payment mix, not on a rate card. Take last quarter's orders, split them by method, apply each provider's rate to the actual volume in each bucket, then add GST at the end. A store that is 70% UPI and a store that is 70% cards should reach different conclusions from the same three rate cards.
The zero-MDR era is ending, and what to do about it
Since January 2020, merchant discount rate has been zero on RuPay debit cards and BHIM-UPI. That was not a commercial decision by the gateways. It was statutory, enforced through Section 10A of the Payment and Settlement Systems Act, 2007 and Section 269SU of the Income-tax Act, 1961. Six years of Indian ecommerce economics have been built on it.
The Taxation and Other Laws (Amendment) Bill, 2026, introduced in the Lok Sabha on 4 August 2026, amends Section 10A. Reporting at the time was consistent on two points: the bill does not itself impose any charge, and it removes the legal provision that made charging impossible.
| What has been discussed | What was reported |
|---|---|
| Rate under consideration | Up to 0.4% on UPI and RuPay debit |
| Transaction threshold | Only on payments above two thousand rupees |
| Who it would apply to | Large merchants only |
| Small merchants | Exempt under every proposal so far |
| Already in force | MDR on RuPay credit via UPI above two thousand rupees, from 1 June 2026 |
| Status as of 24 August 2026 | Enabling law amended, no rate notified |
Treat the figures in that table as reported proposals rather than settled rules, because that is what they are. What is settled is the direction, and the direction is the part you can plan around.
What this actually means for your store
If you are a small merchant, probably nothing for now, and every proposal on the table says so explicitly. If you are large enough for it to reach you, the exposure is straightforward to calculate: take the share of your revenue that arrives by UPI or RuPay debit on orders above two thousand rupees, and apply 0.4% to it. For a store where most orders are UPI and most baskets clear two thousand rupees, that is a real number appearing in a line that has read zero for six years.
Two practical consequences. First, if you have been steering customers to UPI purely because it is free to accept, understand that the reasoning may expire. Steer them there because it converts well and settles cleanly, which are better reasons and will remain true. Second, if you are signing a multi-year gateway contract now, ask explicitly how any notified MDR will be passed through, and get the answer in writing rather than in a sales call.
The other thing worth doing is making sure you can see the number when it arrives. A store that cannot break settlement down by payment method will not notice a new charge for a quarter, which is the same reconciliation problem that courier and accounting integration exists to solve.
Settlement cycle, which usually matters more than the rate
How long the gateway holds your money before releasing it to your bank account. Standard cycles as of August 2026: Cashfree settles at T+1 business day, Razorpay at T+2, PayU at T+2 to T+3. Faster settlement is offered by most providers as a paid add-on, and it matters most to a store funding stock before it is paid for, which is most of what selling products online in India feels like in the first year.
One business day of difference sounds trivial. Run it against a real working-capital position and it usually is not. If you are buying stock on a short cycle, or paying suppliers before customer money lands, the gap between T+1 and T+3 is two days of your revenue sitting somewhere other than your account, permanently, on a rolling basis. That is a fixed loan you are extending to your payment provider, and it costs you whatever your own cost of capital is.
Thin working capital, stock bought on short credit, or a business where a delayed supplier payment costs you a discount. In those cases a faster cycle is cheaper than the overdraft it replaces.
Comfortable cash position, long supplier terms, or made-to-order goods where you are paid before you spend. Take the standard cycle and put the money into the rate instead.
Whether the provider holds a rolling reserve against chargebacks, how much, and for how long. A reserve is a second, quieter delay on a slice of every settlement, and it is not always on the rate card.
What happens to settlement during a dispute, a KYC review or a sudden volume spike. Providers can and do pause settlements. Find out the policy before you need it, not during.
Check the RBI authorisation first
Any non-bank entity that collects money on behalf of merchants in India is a payment aggregator, and needs a Certificate of Authorisation from the Reserve Bank of India to operate as one. This is not a formality, and it belongs on the pre-launch list in the build guide rather than in a review after you have integrated.
More than sixty entities held in-principle or final payment aggregator authorisation as of 2026. The rules also had teeth: entities that failed to submit an application within the prescribed timeline were required to inform their bankers and wind up their payment aggregator business by 28 February 2026. A provider that missed that deadline is not a provider you want holding your settlements.
Are you RBI approved?
Everyone says yes. The phrase is vague enough that it is almost never a lie, and almost never informative either.
Which authorisation do you hold, in which entity name, and is it in-principle or final?
A specific answer you can check against the RBI's published list. A vague one tells you something too.
Cross-border collection is a separate licence again. Payment aggregator cross-border, or PA-CB, authorisation is what a provider needs to handle money coming into India from abroad. By January 2026 nineteen entities were fully authorised as PA-CBs, with Cashfree Payments, Amazon Pay India, Adyen India and BillDesk among the earlier ones and Razorpay in a later wave. If you sell internationally, confirm the provider holds the cross-border authorisation specifically, not just the domestic one.
Success rate is worth more than the rate
The percentage of attempted payments that complete. Nobody advertises it honestly, everybody has one, and a difference of two points in success rate dwarfs a difference of two basis points in MDR. A failed payment is not a discounted sale. It is usually no sale at all, and it belongs in the same category as an over-long checkout: revenue lost without ever appearing in a report.
Success rate is not a single number either. It varies by payment method, by issuing bank, by time of day and by whether a particular bank's netbanking is having a bad afternoon. What you are really buying from a good provider is intelligent routing: the ability to retry a failed transaction through a different acquiring path rather than returning an error to the customer.
Success rate broken down by method
Ask for UPI, cards and netbanking separately, for merchants of roughly your size and sector. A blended figure hides the case that matters to you. If a provider will not break it down, that is an answer.
What happens on failure
Whether the gateway retries through an alternative route automatically, how many attempts, and what the customer sees while it happens. Silent retry is worth real money. A raw bank error page is worth negative money.
The checkout on a mid-range Android phone
On a real device, on mobile data, on the UPI app your customers actually use. Not on the demo. A checkout that works flawlessly on a laptop and drops the UPI intent handoff on a three-year-old phone is a success-rate problem you caused, not one the gateway did.
Your own numbers after ninety days
Attempted payments against completed ones, from your own store data rather than the provider's dashboard. If it is materially below what you were told, you have a specific conversation to have, with evidence.
Saved cards, tokenisation and anything recurring
Two RBI frameworks shape what your checkout is allowed to do, and both catch people out when they assume an ecommerce pattern from another market will work here. Both are worth settling before the ecommerce website is built rather than after.
Card-on-file tokenisation
Since 1 October 2022, merchants and payment service providers in India may not store customer card credentials. Only the issuing bank and the card network may. What sits in your system instead is a token, and tokens are issued per card per merchant, so the same card at a different store is a different token.
The practical consequences are worth knowing before you design a checkout around them. Returning customers get a genuinely fast repeat purchase once a card is tokenised, which is good. But the token belongs to that merchant relationship, so it does not follow the customer anywhere, and it does not survive a migration to another platform in the way a database column would. Ask a prospective provider how tokenisation is presented at checkout and what happens to existing tokens if you leave.
Recurring payments and e-mandates
Recurring card payments run under the RBI e-mandate framework, which imposes an additional factor of authentication above a threshold. The general cap has been fifteen thousand rupees, raised to one lakh without additional authentication for specific categories including mutual fund subscriptions, insurance premiums and credit card bill payments. The merchant must also notify the customer at least twenty-four hours before each scheduled debit, stating amount, merchant name and date, and the customer can cancel the mandate through their bank without contacting you.
If your business model depends on subscriptions, silent renewals or auto-debit above the threshold, confirm exactly how your gateway implements e-mandates before you build anything. This is the single area where an assumption imported from a non-Indian playbook turns into a rebuild.
If you sell outside India
Three things change, and each of them is larger than the domestic decision you have just made. If exporting is the plan rather than an occasional order, it also affects which platform shape suits you.
The rate roughly doubles
International card rates run around 3% to 3.5% before GST against roughly 2% domestic. On thin margins that is the difference between an export order being worth taking and not.
You need the cross-border authorisation
PA-CB, not just the domestic payment aggregator licence. Nineteen entities held full PA-CB authorisation as of January 2026. Confirm your provider is one of them for the corridor you actually sell into.
Documentation becomes a real workflow
Inward remittance for exports carries paperwork that domestic collection does not. Ask the provider what they generate automatically and what your accountant will have to produce by hand.
Currency display is a conversion decision
Showing prices in the buyer's currency lifts conversion and introduces a conversion spread somebody pays for. Decide deliberately who that is rather than discovering it in a settlement report.
What a gateway cannot fix
It is worth being clear about the boundary, because gateway conversations often drift into problems the gateway has nothing to do with. Cash on delivery is the main one. ET Prime Research put COD at 60% to 65% of Indian ecommerce orders in 2024, rising towards 75% in smaller cities, while metros have moved past 70% prepaid.
None of that money passes through your gateway. It is collected by your courier and remitted to you later, on the courier's cycle, which means COD introduces a second and entirely separate reconciliation problem: matching remittances back to individual orders so you can tell whether you were actually paid for a delivered parcel.
A gateway will not help you there. The controls that shift COD volume towards prepaid, such as a discount for paying online, a COD handling fee or a value cap, live in your store rather than in the payment provider, and how easily you can build them depends on whether you are on Shopify or WooCommerce. Getting remittance status back into the order record is a courier integration.
If most of your revenue arrives as cash through a courier, spend your attention there before optimising ten basis points off the MDR on the minority that goes through a gateway. The larger number is almost always the one nobody is looking at.
Switching later, and running two
You are not married to the first choice, and the switching cost is lower than people assume for most stores, and certainly lower than a full redesign and migration. What makes it painful is not the integration. It is tokenised cards that do not transfer, active e-mandates that have to be re-collected, and any reporting your accounts team has built on the old provider's exports.
Running two gateways at once is a legitimate strategy and more common than it looks. The usual reasons are redundancy, so an outage at one provider does not stop the store, and routing, where one provider is better on UPI and another on cards. The cost is a second reconciliation, a second dashboard and a second set of settlement reports, so it earns its place at volume and gets in the way below it.
On WooCommerce, adding a second gateway and routing between them is straightforward because payment method availability is filterable in code. On Shopify it is more constrained, and it interacts with the platform's own transaction fee, which we covered in the Shopify vs WooCommerce comparison. Either way this is ongoing maintenance rather than set up once and forget.
How to decide this week
- Split last quarter's orders by payment method and apply each provider's rate to the real volume in each bucket. Add 18% GST at the end, not to each line.
- Calculate what a 0.4% MDR on UPI and RuPay debit above two thousand rupees would cost you, so you know your exposure before it is notified.
- Ask each provider which RBI authorisation they hold, in which entity name, and check it.
- Ask for success rates broken down by UPI, cards and netbanking for merchants your size.
- Compare settlement cycles against your own working capital rather than in the abstract, and ask about rolling reserves.
- If you sell abroad, confirm PA-CB authorisation specifically and price international orders at the international rate.
- If anything recurring is planned, confirm how e-mandates are implemented before a line of checkout code is written.
- Negotiate. Setup and maintenance fees are frequently waived and the rate itself moves at volume, but only if you ask.
For most stores below serious volume the honest conclusion is that the three main domestic providers are close enough that the rate should not decide it. Choose on settlement cycle if cash is tight, on success rate if you can get real numbers, and on support quality if you cannot, because at some point you will need to ask somebody why a specific customer's payment failed and you will want a useful answer.
Send us a quarter of order data split by payment method and we will tell you what each provider actually costs you, what the UPI change would add, and whether the settlement cycle is worth paying to shorten.
Questions we get asked about this
Which payment gateway is best in India?
For most stores, whichever one settles fastest into your bank and gives you the best success rate on the payment methods your customers actually use. Razorpay, PayU and Cashfree all cluster around 2% on domestic cards as of August 2026, which is close enough that the rate should not be the deciding factor. The differences that matter are settlement cycle, routing on failed payments, and how quickly you get a useful answer when something breaks.
Is UPI still free for merchants to accept?
MDR on UPI and RuPay debit has been zero since January 2020 under Section 10A of the Payment and Settlement Systems Act, 2007. The Taxation and Other Laws (Amendment) Bill, 2026, introduced in the Lok Sabha on 4 August 2026, amends that section and removes the statutory bar. No rate has been notified. Discussions reported so far point to up to 0.4% on transactions above two thousand rupees for large merchants only, with small merchants exempt. Some gateways also already charge a flat platform fee of about two rupees per UPI transaction, which is separate from MDR.
What are payment gateway charges in India, really?
Roughly 2% MDR on domestic cards, netbanking and wallets, plus 18% GST on that fee, which makes the effective cost about 2.36%. International cards run 3% to 3.5% before GST. Some providers add flat per-transaction fees on UPI, and setup or annual maintenance charges appear on some plans. Rates are as of August 2026 and worth confirming against current rate cards.
Can I store my customers' card details to make repeat purchases faster?
No. Since 1 October 2022 only the issuing bank and the card network may store card credentials in India. Your store holds a token instead, issued per card per merchant. Returning customers still get a fast repeat purchase, but the token is specific to your merchant relationship and does not transfer if you change provider or replatform.
Do I need a different gateway to accept international payments?
Not necessarily a different one, but you do need one holding cross-border payment aggregator authorisation, known as PA-CB, which is separate from the domestic licence. Nineteen entities were fully authorised as PA-CBs by January 2026. Confirm your provider holds it for the corridor you sell into, and budget international orders at the international rate rather than the domestic one.
Does the gateway handle cash on delivery?
No. COD money is collected by your courier and remitted to you on their cycle, so it never passes through the gateway at all. That creates a separate reconciliation problem , matching remittances back to individual orders, and it is where most stores lose money without noticing.