What counts as a cross-border payment
A cross-border payment is not simply one where the buyer and the seller live in different countries. In card processing the phrase has a technical definition, and that definition carries a cost, a decline rate and a regulatory perimeter with it.
Three different businesses are sold under this heading. Pay-in is collecting from consumers abroad; pay-out is sending money to sellers, drivers or suppliers abroad; corporate cross-border is moving company money between entities and managing the currency exposure. dLocal and EBANX are built around the first two, Payoneer and Tipalti around the second, Corpay's cross-border arm and Airwallex closer to the third. A provider excellent at one is frequently mediocre at the others.
Most of this page concerns pay-in, because that is where money is lost quietly. Payout costs are visible — a wire fee, a conversion, a landing time. On the collection side the largest cost is revenue that never arrived because the transaction was declined, and no invoice records it.
Local acquiring versus cross-border acquiring
This is the distinction that decides whether an international expansion works. Everything else here is downstream of it.
Process a foreign-issued card through a home-country acquirer and several things happen at once, none helpful. The issuer's risk engine sees a foreign acquirer country and weights toward decline, because a disproportionate share of card fraud arrives that way. The transaction prices into cross-border interchange and scheme categories. Domestic-only card schemes become unreachable — Elo and Hipercard in Brazil, RuPay in India. Local installment products, which in Brazil are how a large share of considered purchases are paid for, cannot be offered at all. Local authentication requirements such as India's tokenization mandate may not be supported end to end.
The result is an authorization gap. Take an illustrative merchant with 100,000 monthly attempts averaging a hypothetical $60. If cross-border acquiring approves 72% and local acquiring in the same market approves 86%, that is 14,000 additional approved transactions a month — roughly $840,000 of illustrative recovered revenue. No fee negotiation moves that much money, yet procurement routinely spends six weeks on the processing fee and ten minutes on approval rates, because the fee appears on a document and the declines do not.
The mistake most businesses make here is assuming that pricing in local currency at checkout means the payment is acquired locally. Multi-currency pricing is a presentation feature; local acquiring is a licensing fact. The second mistake is trusting country counts: Nuvei publishes that it reaches more than 200 markets with local acquiring in 52 of them. Ask for the acquiring map, not the processing map.
Alternative payment methods, and why they are not optional
A card-only checkout in the Netherlands, Brazil, India or Mexico is not a checkout with fewer options; it is one most of the market cannot use. The mechanics differ in ways that reach into refunds, reconciliation and fraud rather than staying inside the payment page.
| Market | Method | What it is | What changes downstream |
|---|---|---|---|
| Brazil | Pix | Instant account-to-account transfer operated by the central bank | Irrevocable. No chargebacks, but no chargeback defense either: a refund is a new outbound payment and scam losses are unrecoverable |
| Brazil | Boleto bancário | A printed or digital voucher paid at a bank, ATM or banking app | Confirmation is delayed by days, so orders must be held and a meaningful share never gets paid |
| Brazil | Card installments | Purchase split over monthly installments on a domestic credit card | Requires local acquiring; settlement can spread across the installment schedule rather than arriving in two days |
| Mexico | OXXO | Cash payment at a convenience store against a generated reference | Same delayed-confirmation profile as boleto; reaches consumers with no card at all |
| Mexico | SPEI | Interbank electronic transfer system | Push payment, reconciled against a reference the consumer must enter correctly |
| India | UPI | Real-time account-to-account interface, dominant for consumer payments | Recurring billing needs UPI Autopay mandates rather than a stored card; local data and tokenization rules apply |
| Netherlands | iDEAL | Bank redirect with authentication in the consumer's own banking app | Strong authentication, effectively no chargeback path; refunds are outbound transfers |
| Colombia | PSE and Bre-B | Bank transfer and, more recently, an instant payment system | New rails need early provider support; EBANX was an early Bre-B implementer |
| Nordics, Poland, Belgium | Swish, MobilePay, Blik, Bancontact | Domestic wallet and bank-transfer schemes | Each has its own refund window, dispute model and settlement calendar |
| Sub-Saharan Africa | Mobile money | Carrier and wallet balances used as primary accounts | Often the only realistic collection and payout method; per-country licensing is the constraint |
Two consequences follow. Irrevocable rails move fraud risk rather than removing it: no chargeback also means no scheme process for recovering a payment sent under a scam, so the dispute lands in customer service and the refund budget instead. And recurring billing does not port — an APM subscription needs a local mandate construct such as Pix Automático, UPI Autopay or SEPA Direct Debit. Treat mandate support as a gating requirement, not a roadmap item.
FX markup and the four places it hides
The headline processing fee is often the smaller half of the cost. dLocal reports its economics as a blended take rate that deliberately combines processing fees and currency conversion revenue, and discloses that the blend varies materially by country and payment method. EBANX publishes no rate card at all. Payoneer publishes fees per market but states plainly that the conversion margin dominates for most users.
- Blended into the take rate. One percentage covers acceptance and conversion together. Ask for the components separately, and ask which reference rate and timestamp are used.
- At settlement rather than at the transaction. The fee is quoted on the local-currency amount, but conversion into the settlement currency happens later, at a rate never quoted in advance.
- In dynamic currency conversion at checkout. The cardholder is offered their home currency; somebody earns a margin on it and it is rarely the merchant. It also depresses conversion, because the displayed price rises at the worst moment.
- In the payout leg. Platforms that collect in one currency and disburse in another pay a spread twice. BVNK makes the same point about stablecoin rails: price the round trip, because single legs always look cheap.
The practical test is arithmetic, not argument. Take a real week of your own volume, ask each shortlisted provider what the merchant would have received in its settlement currency for that exact set of transactions, and compare against the mid-market value of the same volume. That single number is the all-in cost.
Settlement currency, timing and trapped money
Cross-border settlement is where a payments decision turns out to have been a working-capital decision. Three variables define it: which currency you are paid in, which legal entity pays, and on what calendar.
Currency. Settling locally keeps FX risk and timing with the merchant, which is right if there are local costs to pay or a treasury function to hedge. Settling in the home currency hands the conversion to the provider, which is convenient and is where the spread lives. The choice should be deliberate and priced, not inherited from a default.
Timing. Settlement periods differ by market and method far more than merchants expect. An instant rail such as Pix or UPI confirms in seconds, but the payout calendar is a separate commercial term. Brazilian card installments can pay out across the installment schedule, so a merchant selling in twelve installments finances the customer for a year whether it meant to or not.
Repatriation. Some currencies are not freely convertible or exportable, and no provider can contract that away. Capital controls and central bank approval regimes mean money collected in certain markets sits there until the process allows it out; dLocal's own commentary has singled out Argentina and Nigeria as places where volume and currency dynamics move together. Model cash that is earned but not yet repatriable as a separate line.
Get three answers into the contract rather than an email: in which currency and from which legal entity do we settle in each market; on what calendar, by payment method; and who carries the FX risk between authorization and settlement.
Licensing, entities and what money cannot buy
Cross-border providers exist because several large markets require a licence, a local entity, or both, to acquire domestically — and obtaining them takes years. What a merchant buys from dLocal, EBANX, Airwallex or Adyen is not software. It is somebody else's regulatory perimeter.
Those perimeters differ, and are worth reading rather than assuming. dLocal holds an EU electronic money institution licence, a UK payment institution licence approved by the FCA in December 2024, FinCEN money services business registration in the US, and licences or approvals across Brazil, Mexico, Argentina, Nigeria, Kenya, South Africa, the Philippines and the UAE among others. EBANX obtained a Major Payment Institution licence from the Monetary Authority of Singapore in 2025 but publishes no consolidated list of its entities. Airwallex reported around 80 licences and permits by the end of 2025. Adyen holds a Dutch banking licence supervised by De Nederlandsche Bank; Checkout.com is a UK electronic money institution and a principal scheme member; BVNK holds a Maltese crypto-asset service provider licence under MiCA.
Three further constraints bind merchants directly: data and tokenization rules dictating where card data may be stored, notably in India; indirect tax registration for digital goods across a long list of jurisdictions; and sanctions and AML screening obligations attaching to whoever moves the money, which affects how fast payouts to new beneficiaries release. The tax burden is why the merchant of record model exists: Paddle contracts with the end customer as the legal seller and carries the VAT and GST registration and remittance itself, a genuine transfer of operational burden paid for with a discount off gross revenue and the loss of the direct customer relationship.
The four models you can actually buy
Despite the number of logos in this category, there are only four structural answers to the cross-border problem. Matching the model to the business matters more than choosing between vendors inside a model.
| Model | How it works | Examples | Right when | Wrong when |
|---|---|---|---|---|
| Local-methods pay-in specialist | The provider is the licensed local party, offers local cards, instant rails and cash methods, and settles to the merchant abroad | dLocal, EBANX | Meaningful volume in emerging markets where local methods dominate and acquiring is licence-gated | Volume sits in the US, UK and western Europe, where the specialist adds a second reconciliation for little gain |
| Global acquirer with local licences | One contract and one integration, with local acquiring where the acquirer holds permissions | Adyen, Checkout.com, Nuvei, Worldpay | Volume sits in markets the acquirer is genuinely licensed in and consolidated reporting is worth money | Target markets are on the processing map but not the acquiring map |
| Merchant of record | The provider is the legal seller and carries indirect tax registration and remittance, paying the vendor a net amount | Paddle | Digital goods and software sold into many jurisdictions by a company that cannot staff global tax compliance | The business must own the customer relationship, the checkout data and the refund policy |
| Multi-currency account and payout platform | Local receiving accounts, in-account conversion and outbound payment to beneficiaries abroad | Airwallex, Payoneer, Corpay, Tipalti, BVNK | The problem is paying people and suppliers abroad, or holding balances in several currencies | The problem is consumer collection, which these are not built for |
Plenty of businesses deliberately run two of these: a global acquirer for the core markets and a local specialist for one or two hard ones. What is not defensible is arriving there by accident, one country at a time, until finance reconciles five settlement files against three ledgers.
How to run the evaluation
A cross-border selection that ends well looks much the same regardless of industry, and the work is front-loaded.
- Start from your own volume map — country, method, average value, issuing BIN country, current approval rate. Without it every quote is unverifiable.
- Demand the acquiring map country by country, and ask which legal entity acquires in each.
- Ask for approval rates for your merchant category in your top five markets, and ask what the denominator is. Approval on first attempt and approval after retries are different numbers, and the flattering one is usually quoted.
- Price the round trip in your own currency from a real week of transactions, so FX spread and processing fee are compared as one figure.
- Get the settlement calendar per market and method in writing, with reserve terms, which are common in cross-border and rarely mentioned before underwriting.
- Test the refund path on irrevocable rails before launch. A Pix refund is an outbound payment with its own failure modes.
- Ask what happens on exit — how long funds are held and whether local mandates are portable. Card vaults usually are; local mandates frequently are not, and a subscription business that cannot move its mandates is not free to switch.
Be willing to conclude this category is not for you. If ninety-odd percent of volume is domestic, a cross-border specialist buys a second contract and a second reconciliation in exchange for a rounding error. If a software business sells globally at modest scale and cannot staff indirect tax compliance, a merchant of record beats a direct acquiring stack even though the headline discount looks worse, because getting VAT registration wrong is not on the same cost scale as the pricing difference. The right answer here is almost never the lowest quoted percentage.
Frequently asked questions
What is the difference between local acquiring and cross-border acquiring?
In cross-border acquiring the acquirer is domiciled in a different country from the card issuer, so the card networks treat the transaction as cross-border and price it into cross-border interchange and assessment categories. In local acquiring the transaction is acquired by an entity licensed in the same country as the issuer, so it is classified as domestic. The commercial consequence is that local acquiring typically produces materially higher authorization rates, reaches domestic-only card schemes and installment products, and settles on local terms.
Why do international card payments get declined more often?
Issuer risk models weight a foreign acquirer country heavily toward decline, because a disproportionate share of card fraud arrives through cross-border transactions. Domestic-only card schemes and local installment products also cannot be reached through a foreign acquirer at all, and local authentication requirements such as domestic 3-D Secure flavors or tokenization mandates may not be supported end to end. The gap between cross-border and local approval rates in the same market is usually worth far more than any difference in processing fees.
What are alternative payment methods, and do I really need them?
Alternative payment methods are consumer payment methods that are not international card schemes: instant bank transfers such as Pix in Brazil and UPI in India, bank redirects such as iDEAL in the Netherlands, and cash vouchers such as boleto bancário and OXXO. In several of the largest e-commerce markets these are the dominant way consumers pay, so a card-only checkout excludes most of the market. They also behave differently from cards — instant rails are irrevocable with no chargeback path, and voucher methods confirm after a delay — which changes refunds, fraud handling and order management.
How do cross-border payment companies make money on foreign exchange?
Through a spread between the rate they apply and the interbank mid-market rate, rather than through an invoiced fee. Several report the two together: dLocal discloses a blended take rate combining processing and currency conversion revenue that varies by country and method, and Payoneer states that the conversion margin is the dominant charge for most users. The only reliable comparison is to price a real week of transactions end to end and compare the amount actually received in the settlement currency against the mid-market value of the same volume.
Do I need a local entity to accept payments in another country?
Often not, because that is exactly what cross-border providers sell. Companies such as dLocal, EBANX, Airwallex, Adyen and Checkout.com hold the local licences, entities and clearing relationships, and the merchant contracts with them instead. What cannot be outsourced is a merchant's own indirect tax obligations in some markets, data localization obligations in others, and the fact that money collected in a market with capital controls may not be freely repatriable regardless of who processed it.
What is a merchant of record, and when should I use one?
A merchant of record is a provider that contracts with the end customer as the legal seller, so it — not the vendor — issues the invoice, appears on the card statement, and carries the sales tax, VAT and GST registration and remittance obligation. Paddle operates this way for software and digital products. It suits companies selling into many jurisdictions that cannot staff global tax compliance, and it is the wrong choice for a business that needs to own the customer relationship, the checkout experience and the underlying transaction data.
How long does cross-border settlement take?
It varies by market and payment method far more than domestic settlement does, and the payout calendar is a negotiated commercial term rather than a property of the rail. An instant rail such as Pix confirms in seconds while the merchant's payout may still be next-day or slower; Brazilian card installments can settle across the installment schedule; cash-voucher methods confirm after several days. Any single blended assumption applied across markets will be wrong, so get the calendar in writing per market and per method.