Why customers leave their bank for international payments — and what it takes to win them back

Ask anyone inside a bank why cross-border volume and customers are drifting to specialist digital providers, and you will almost always hear the same answer: they’re cheaper. It’s a comfortable explanation, because it points at something outside the bank’s control: correspondent banking costs what it costs, and nobody wants to start a price war in a business that carries a meaningful share of payment revenue.

New research out of Harvard suggests that comfortable explanation is the wrong one.

The research paints a different picture

In February 2026, Wenxin Du, Catherine Huang and David Scharfstein of Harvard Business School published a working paper. It compares three routes for cross-border payments: banks, specialist payment providers and stablecoins. On cost, the finding is unambiguous: on retail transfers, bank foreign exchange markups run several times higher than what specialist providers charge for the same payment. That holds even for the most liquid currency pairs — the ones where market thinness is no excuse. At the top of the range, markups run up to ten times higher than at the competition.

What deserves attention is how the authors read this: they attribute the gap not to higher costs at the bank, but on the one hand to friction in correspondent banking, and on the other to pricing power. And that pricing power exists only because customers have — so far — stayed.

Another important sentence sits a few lines further on, and it has nothing to do with money:

“Sending an international FX wire often requires an in-person branch visit, and there is typically no mechanism to track the payment once funds have been debited from the sender’s account.”

Against that, here is what the alternative offers: fully digital interfaces; pricing communicated upfront; and payment tracking in real time. Read both findings side by side, and the migration stops looking like a pure price discussion: someone who has to visit a branch to send money abroad, and who then has no idea where that money is, isn’t comparing markups. They’re comparing whether the thing works.

The moment the decision actually happens

Here is what that looks like on a Tuesday morning, in a business that has nothing to do with payments. A supplier invoice arrives from Taiwan. Someone in finance opens the banking app or online banking, taps through to the international transfer, and is confronted with an enormous form: recipient name, full address across separate lines, account number, international bank code, a routing code whose format depends on the destination country, currency, purpose of payment, reference, and more besides.

Each of those values appears on the invoice the payment is based on. The app asks for all of it to be typed in any way. This happens a few times a year, never regularly enough to build a routine, yet always often enough to remember how tiring it was last time. Careful typing and repeated checking before submission — only to learn of an error days later, when the payment is returned as failed — depending on the institution, minus the charges incurred.

This is where the decision falls, not in the fee schedule and not at the rate on the confirmation screen. The form is where a bank is perceived as a reliable place for international payments, and that form is precisely the part most modernization programs never touched.

What is at risk — and what is not

Banks have not lost the international payments relationship. Roughly three-quarters of SME cross-border flows still run through banks. That said, between 35 and 50 percent of SMEs and mid-corporates, in Europe as in North America, already route part of that volume through fintech providers. What stands out is that the same figure appears independently at both Mastercard and McKinsey.

The word that matters is part: nobody moves everything at once. What typically happens is that one payment goes elsewhere because it was easier that day, and then the next one does too. Each of those partial migrations shifts margin out of one of the most profitable corners of the payments business and towards the competition. That margin is at risk; lost, however, it is not.

From the revenue side, McKinsey sharpens the picture. Lower-value cross-border payments account for around a tenth of transaction volume but deliver roughly a third of the revenue. So with every single decision taken against the house bank, a disproportionately profitable segment erodes a little further — precisely where it hurts most.

The problem is not the infrastructure

In payments, there is a reflex: every cross-border question gets an infrastructure answer. Modernize the rails, join a scheme, partner with someone who has already done it.

Except that the rails are, by now, largely fine. Since late 2025, cross-border payments have run on one structured standard across the network. Within an hour, the great majority of correctly initiated payments reach the recipient’s bank. The industry did that expensive, unglamorous work, and it worked.

The qualifier is what counts: correctly initiated. Every one of those investments improved what happens after submission; not one of them touched the screen where the payment is created. A payment system cannot repair data it never received correctly, because what arrives is what the customer typed.

That leaves the bank in an odd position: modern infrastructure, competitive capability, and a customer experience from the decade before last. This is not an infrastructure problem, it is a problem of the first mile and of user experience. And solving it costs a fraction.

Where the repair is needed

The fix is unglamorous, which probably explains why it went unnoticed for so long: customers should stop typing what is already in front of them. That is exactly what Cross-Border Photo Payment Initiation does.

The customer photographs the supplier invoice or uploads it — a photo, a PDF, even a screenshot from a messenger thread. Gini extracts the payment data and returns it into the transfer form: recipient, address, account details, bank code, amount, currency. And with Gini’s front end, entries are checked in real time, while the customer is still in the flow. An implausible bank code shows up immediately, not days later. The customer reviews, confirms, and moves on.

What reaches the bank is a complete, correctly structured payment order, handed to the international payment systems already in place. Behind the initiation layer nothing changes: no new rail, no repricing, no core banking project. Institutions already running Gini’s Photo Payment domestically know the route anyway. For them, cross-border is the extension of an integration their security, compliance and engineering teams approved long ago.

The strategic point

The uncomfortable reading of the Harvard findings is that banks have been competing on the wrong variable. The encouraging insight: precisely the variable customers care about is one they can still change, without touching pricing or infrastructure.

Banks own what no specialist provider can replicate — the primary account relationship, the trust, the rails, the regulatory standing. What has always been missing is the consistent step through to the UX layer: that moment when an invoice becomes a payment. That moment is no footnote to the international payments business; it is becoming the main driver of it.

Cross-Border Photo Payment Initiation is Gini’s product for banks that want to close the initiation gap in app and online banking. See how it works.

Sources

1. Du, W., Huang, C. & Scharfstein, D. (2026). Competing Rails for Cross-Border Payments: Banks, Fintechs, and Stablecoins. Harvard Business School Working Paper, 15 February 2026. https://www.hbs.edu/faculty/Pages/item.aspx?num=68673
2. Mastercard SME cross-border research, 2024.
3. McKinsey & Company, How banks can win back lower-value cross-border payments business.
4. SWIFT: ISO 20022 live since November 2025; 90% of correctly initiated payments delivered within one hour (SWIFT press release, October 2024).