Correspondent Banking: Why a Cross-Border Payment Takes Days and Arrives Short

Correspondent banking is easiest to understand through a payment gone wrong.

A supplier in Lagos invoices you $5,000. You send $5,000. Eleven days later, they receive $4,955 and ask why you underpaid.

You didn't. Neither did your bank.

The missing $45 was taken in pieces by banks you have no relationship with and can't even identify. The eleven days were spent sitting in queues at institutions that never show up on your statement.

This is correspondent banking working exactly as designed — and understanding that design is the only way to have a real conversation about it, whether with your bank or with a supplier who thinks you shortchanged them.

Nostro/vostro: two accounts, and the whole system

Banks don't have accounts with every other bank. There are tens of thousands of banks, so a full mesh is impossible. Instead each bank holds accounts with a small number of others, and payments hop between them.

An account your bank holds at a foreign bank, denominated in that foreign currency, is a nostro account. From Latin, "ours": our money held by you. The mirror image, an account a foreign bank holds at your bank in your currency, is a vostro account. "Yours": your money held by us.

The same account is a nostro to one party and a vostro to the other. That's it.

Every cross-border payment is a series of debits and credits across these accounts. Every complication downstream comes from how long that series is.

A US bank with a nostro at a Nigerian bank can pay a Nigerian beneficiary in one hop. A US bank without one has to find a bank that does have that relationship, which usually means routing through a large institution in London or New York first.

The number of hops is the whole story.

Each one adds a queue, a cutoff time, a compliance check, and an opportunity to deduct a fee.

Serial payments and cover payments

There are two ways to structure the journey, and the difference matters because it determines what information travels with the money.

In a serial payment, the instruction and the funds move together down the chain. Bank A instructs Bank B, which instructs Bank C, which credits the beneficiary. Each bank sees the full payment details and each can take a fee from the amount passing through.

In a cover payment, the instruction goes directly from the sending bank to the beneficiary's bank, while a separate message tells the correspondents to move the cover funds between themselves. The beneficiary bank knows what's coming before the money arrives.

Cover payments were historically the place where information got lost, because the settlement leg between correspondents didn't have to carry the originator and beneficiary details. The dedicated cover message, MT202COV and now pacs.009 COV, exists to close that gap, which is why compliance teams care about a distinction that sounds purely technical.

For a payer the practical consequence is simple.

Ask your bank whether payments to a given corridor route serially or by cover, and how many intermediaries are involved. A bank that can't answer quickly is a bank that hasn't looked, and asking is often enough to make someone look.

Where the international wire transfer fees go

Three fee conventions exist, and choosing between them is the single decision a payer actually controls.

Where no instruction is given explicitly, SHA is commonly what applies, which is what happened to the Lagos supplier. Each intermediary deducted what's often called a lifting fee, taken from the principal rather than billed separately.

The problem with SHA isn't the cost.

It's that the cost is unknown until after the fact, which makes reconciliation on the receiving side impossible to automate. The supplier cannot match $4,955 against a $5,000 invoice without a human deciding that is what happened, and that human then has to decide whether to chase the $45 or write it off, every month, for every payment.

OUR costs more and is usually cheaper overall, once you account for the queries, the write-offs of small differences, and the relationship damage from a supplier who thinks they're being short-paid. If you pay the same foreign suppliers monthly, this is worth changing once and never revisiting.

Days, when the message takes seconds

The message reaches the far side almost immediately. The money does not, and the gap has four causes.

The screening row is the one that surprises people. It isn't a database lookup that either passes or fails: a partial name match against a sanctions list produces an alert, and an alert is cleared by a person reading it. That person works office hours in a particular time zone, and there may be three of them in the chain.

None of these is fixable by the payer.

That's the honest answer to "can we make this faster", and it's worth giving plainly rather than promising to look into it. What is fixable is expectation. A payment initiated at 4:00 p.m. on a Thursday before a bank holiday is a Tuesday payment, and saying so up front converts a complaint into a plan.

The SWIFT payment message changed, and it matters

For decades the SWIFT payment instruction for a customer credit transfer was an MT103, and the bank-to-bank leg was an MT202. Those messages were compact, positional, and limited: fixed field lengths, minimal structure for names and addresses, and no room to carry rich remittance information.

That era ended. The ISO 20022 coexistence period for cross-border payments closed on 22 November 2025, after which MT payment instructions were no longer supported by Swift's FIN service for interbank cross-border flows. ISO 20022 messages travel on FINplus instead. MT103 is replaced by pacs.008 and MT202 by pacs.009.

The cut was not total, and the detail matters if your bank still talks about MT. Five types were removed outright and are rejected on sight: MT 102, MT 102 STP, MT 103 REMIT, MT 201 and MT 203. Seven others, including MT 103 and MT 202, are still accepted and put through contingency processing that converts them, and that conversion became chargeable from 1 January 2026.

What this changes for a payer is worth stating plainly, because vendors have oversold it.

It does not make payments instant.

The hops, the cutoffs, and the screening are all unchanged. A new message format does not shorten a chain of bilateral relationships.

It does give the payment structured party data and a proper remittance field.

That reduces the number of payments stopping for manual review because a name was truncated to fit a fixed-length field, and it makes it possible for a supplier to receive a payment with the invoice number attached to it rather than jammed into a free-text line and hoping. For anyone reconciling incoming payments, that's the difference between a match and a phone call. The same problem, and the identifiers that solve it, show up on the crypto side too, in crypto reconciliation.

If your bank still asks you to squeeze invoice references into a short fixed-length field, it's passing a legacy constraint on to you. Worth asking why.

Correspondent banking has been shrinking for a decade

This is the structural trend behind the corridor problems, and it's worth citing carefully because the good data has an end date.

The Committee on Payments and Market Infrastructures, working with Swift, tracked correspondent relationships for years. On the 2020 data, published in December 2021, correspondent banking relationships had declined by about 25% between 2011 and 2020, with a fall of 4% during 2020 alone.

The counterintuitive part: in that same year the volume of cross-border payments grew 2% and their value grew 7%. Fewer relationships, more traffic through them. That's concentration, not a decline in use.

The series has an end date, and quoting it without one is where commentary goes wrong. Chartpacks and underlying data run to end-2022, after which the CPMI and Swift had completed their commitment to the Financial Stability Board and the annual review stopped. So a figure for 2026 does not exist. Give the period with the number and let the reader judge how much has changed since.

De-risking, and who it lands on

The retreat wasn't random. Banks withdrew from relationships where the compliance cost exceeded the revenue, and that calculation goes badly in exactly the places that most need banking: smaller economies, jurisdictions with weaker AML supervision, corridors with high remittance volume and low transaction values.

The result is a two-tier system. A payment between two G7 countries is usually one or two hops, priced competitively. A payment into a smaller market can be four hops, priced by whoever holds the only remaining relationship into that corridor.

That pricing power is why the same $5,000 transfer costs a predictable amount into Frankfurt and an unpredictable one into Lagos.

The sender's negotiating position is weakest in precisely the corridors where the fees are highest. That isn't a coincidence, it's the mechanism.

What a payer can change about correspondent banking

Five things, in rough order of how much they help.

Set the charge instruction deliberately. OUR for suppliers you pay repeatedly. It removes the entire category of short-payment queries.

Ask your bank for the routing. Number of intermediaries per corridor, and their typical cutoffs. Some banks will move a corridor to a shorter route if asked, because they have more than one option and default to the cheapest for them.

Batch by corridor, not by date. Fewer, larger payments into a corridor cost less in aggregate than many small ones, because lifting fees are largely fixed per transaction.

Send early in the week and early in the day. This sounds trivial. Against a three-hop chain with three cutoffs, it's often two days.

Pick the right rail for the job. Not every cross-border payment needs a wire, and not every domestic one does either. The full map is in payment rails, and the domestic comparison most AP teams get wrong is in wire transfer vs ACH.

Give the supplier the reference in a form they can use. With structured remittance data now available, the invoice number can travel with the payment rather than in a separate email.

None of that changes the architecture.

It reduces the number of times the architecture surprises you, which in a finance team is most of what improvement looks like.

The route that skips the chain

Stablecoin settlement removes the intermediaries rather than optimising them. There's no nostro chain, no lifting fee, and no cutoff, because the network doesn't close.

It also introduces its own set of problems, which are real and different: the recipient needs somewhere to convert into local currency, that conversion has its own cost and its own queue, and the payment is final on confirmation with no recall. The cost of the off-ramp is usually where the saving goes, and that's covered honestly in stablecoin remittances. The broader comparison against wires sits in crypto for cross-border payments.

The contractual side, meaning when the obligation is actually discharged and what to write into the agreement, is separate again and sits in B2B crypto payments. For a company paying the same foreign suppliers every month, the question isn't which rail is better in the abstract. It's which suppliers can actually receive and convert, and what the total cost is at their end rather than at yours. Where the answer works, VaultNow handles the payout side with screening and an address book attached, so the operational discipline matches what the bank run already has.

Frequently asked questions

What is correspondent banking?

An arrangement where one bank holds an account with another so it can make payments in a currency or country where it has no branch. Cross-border payments move as a series of debits and credits across these accounts, and the number of banks in the chain determines the speed and the cost.

What is the difference between a nostro and a vostro account?

They are the same account seen from two sides. A nostro is an account your bank holds at another bank in that bank's currency, "our money with you". A vostro is an account another bank holds at your bank in your currency, "your money with us".

Why does an international transfer arrive short?

Because the payment was sent under a shared charge instruction, so each intermediary bank deducted a fee from the amount in transit. Sending under an OUR instruction, where the sender pays all charges, means the beneficiary receives the full invoice amount.

What is a cover payment?

A structure where the payment instruction goes directly from the sending bank to the beneficiary's bank while the funds move separately between correspondent banks. It's faster to advise the beneficiary bank, and it historically created gaps in the party information travelling with the settlement leg, which rules have since closed.

Do banks still accept MT103?

Not as the standard. Coexistence ended on 22 November 2025 and MT payment instructions are no longer supported on Swift's FIN service for interbank cross-border flows, with ISO 20022 carried on FINplus instead. MT103 itself is still accepted and automatically converted through contingency processing, which has been chargeable since 1 January 2026. Some other types, including MT 102 and MT 103 REMIT, were removed outright.

Is correspondent banking declining?

Correspondent banking relationships fell by about 25% between 2011 and 2020 on CPMI data, including a 4% fall during 2020, while in the same year the volume of cross-border payments grew 2% and their value grew 7%. The published data runs to end-2022 and the annual review was discontinued in 2023, so any figure quoted for a later year is an estimate rather than a measurement.

How can a company reduce cross-border payment costs?

Set the charge instruction to OUR for regular suppliers, ask the bank how many intermediaries a given corridor uses, consolidate payments into fewer larger transfers per corridor, and initiate early in the day and the week to avoid cutoff-driven delays.

Correspondent banking is slow and lossy for structural reasons: a chain of bilateral relationships, each with its own queue, its own compliance function, and its own fee. Most of what a payer can do is reduce the number of surprises rather than the number of hops.

The rest is a question of whether a given supplier has an alternative that works at their end. Where they do, VaultNow covers the payout side with address screening and permissions built in.

General information, current as at 31 August 2026. Not legal or financial advice. The CPMI figures cited carry their reporting periods.