What a border does to a payment
A cross-border payment is one where the payer and the beneficiary hold their accounts in two different jurisdictions, usually in two different currencies. The difficulty comes from how the monetary system is organized. There is no global central bank, and so no settlement asset shared by the euro and the won. A transfer from Paris to Seoul is never a single settlement; it breaks down into a series of domestic payments. Entries in the books of banks that extend credit to one another link each leg to the next. A currency conversion happens at one point in the chain, and every link runs a compliance check. This architecture creates the four problems of cross-border payments: high cost, low speed, limited transparency, and limited access. The G20 launched a roadmap to address them in 2020.
Correspondent banking: nostro, vostro, and the correspondent chain
Correspondent banking is a relationship in which one institution holds an account for another and makes payments on its behalf, in a currency or market where the second institution has no presence. The model rests on reciprocal accounts. A Korean bank that wants to pay in dollars opens a USD account with a US bank. That single account appears in both banks' books: as an asset at one, and as a deposit at the other. The money does not “travel” from one country to another; it changes hands in the correspondent's books. Final settlement takes place in reserves at the central bank that issues the currency. The BIS points out that this system descends directly from the networks of merchant bankers: a chain of bilateral relationships. Sequential compliance checks, each run by a link on its own behalf, have been layered on top ever since.
| End | Perspective | What it is | Example |
|---|---|---|---|
| Nostro | The bank that holds an account with the other | “Our” account, held with the correspondent in its currency, and therefore an asset | The Korean bank monitors its USD nostro at its New York bank |
| Vostro | The bank that holds the account for the other | “Your” account on our books: a deposit, and therefore a liability | The New York bank holds the Korean bank's USD vostro |
| Loro | A third party to the relationship | “Their” account, used when describing the relationship between two other banks | Common in correspondent documentation and payment instructions |
| Corridor | Statistical | A sending country → receiving country pair served by at least one active relationship | The unit the CPMI and the BIS count to measure banks' retreat |
Where the money goes, where the cost goes
The total cost of a cross-border payment has five components. The sender's bank collects the upfront fees, and the institution that converts the currency takes the FX margin. Intermediary charges and the beneficiary bank's fees are deducted further down the chain, and exception fees are added when a payment fails on the first attempt. The price quoted upfront is almost always the smallest component. Most of the cost sits in the exchange rate, so it stays invisible until the rate is compared with a market benchmark. Another part only becomes known after the fact, because it is taken by parties the sender did not choose. The FSB's indicators quantify this breakdown. In retail payments, FX accounts for 87.1% of total cost in B2B and 99.7% in P2B (FXC Intelligence, March 2025). A negotiation that covers only the fixed fees therefore leaves out almost the entire cost of the transaction.
| Item | Who takes it | Visible upfront? | Typical size (source) |
|---|---|---|---|
| *FX margin (spread)* | The bank or PSP that converts | Rarely: it is built into the rate | 1.4% of the amount in B2B and B2P, 1.9% in P2B, 1.4% in P2P (FXC Intelligence for the FSB, March 2025) |
| Upfront fees | The payer’s bank | Yes, per the fee schedule | The smallest item: 0.2% in B2B, 0.5% in B2P, ≈0% in P2B, 1.2% in P2P (same source) |
| *Intermediary charges (lifting fees)* | Each correspondent along the way | No: depends on the actual routing, which is not known in advance | Deducted along the way and recorded in the message's charge fields (:71F in MT, ChrgsInf in ISO 20022) |
| Beneficiary-side fees | The beneficiary's bank | No: outside the payer's contract | 0.1% to 1.3% of the amount sent, weighing more heavily the smaller the amount (anonymous FSB–World Bank survey, 262 responses from 48 jurisdictions, 2025) |
| Exception fees | Any link in the chain | No | Investigations, repairs, returns: the cost of payments that fail on the first attempt, which the BIS describes as common and expensive (Bulletin No. 87, 2024) |
# --- MT 103 (out of CBPR+ scope since 2025-11-22, but still read everywhere) ---
:20:REF20260715001
:32A:260715USD108250,00 <- amount SETTLED interbank: value date + currency + amount
:33B:EUR100000,00 <- amount INSTRUCTED by the customer, before conversion
:36:1,0825 <- rate applied. Compare with the day's reference rate: the gap = the margin
:71A:SHA <- charge bearer: OUR (payer) / BEN (beneficiary) / SHA (shared)
:71F:USD25,00 <- charges ALREADY DEDUCTED from the amount in transit (present if SHA or BEN)
:71G:EUR15,00 <- charges TO BE ADDED to the payer's debit (present if OUR)
:57A:BKKRSEXXXXX <- beneficiary's institution
:59:/1234567890 KOREAN SUPPLIER
:70:INVOICE 2026-0042 <- remittance info: this is what makes reconciliation possible
# --- pacs.008 (CBPR+): same information, named and structured fields ---
<IntrBkSttlmAmt Ccy="USD">108250.00</IntrBkSttlmAmt> <- formerly :32A
<InstdAmt Ccy="EUR">100000.00</InstdAmt> <- formerly :33B
<XchgRate>1.0825</XchgRate> <- formerly :36
<ChrgBr>SHAR</ChrgBr> <- formerly :71A: DEBT | CRED | SHAR | SLEV
<ChrgsInf> <- formerly :71F, but the charging agent is IDENTIFIED
<Amt Ccy="USD">25.00</Amt>
<Agt><FinInstnId><BICFI>...</BICFI></FinInstnId></Agt>
</ChrgsInf>
<PmtId><UETR>eb6305c9-1f7f-49de-aed0-16487c33b1e2</UETR></PmtId> <- end-to-end tracking
# HOW TO READ IT
# instructed amount - sum(ChrgsInf) - FX margin - receiving bank's fees
# = what actually lands in the beneficiary's account.
# With ChrgBr = SHAR, this result is NOT known when the payment is sent.MT vs. ISO 20022: where the migration really stands
CBPR+ (Cross-Border Payments and Reporting Plus) is the set of usage guidelines for ISO 20022 messages on the Swift network. It defines how those messages are used for cross-border payments and cash management reporting. A coexistence period allowed MT messages within that scope until November 22, 2025. Since then, these flows have been exchanged in ISO 20022 through the FINplus service, and the MT 1xx, 2xx, and 9xx categories no longer meet CBPR+ requirements. The main effect of the migration is on message content: party, purpose, and charges data become structured, and therefore machine-checkable. The FSB notes that speed improves mainly when both institutions in a corridor have completed their migration.
| Use case | Legacy MT message | ISO 20022 message | Practical benefit |
|---|---|---|---|
| Customer credit transfer | MT 103 | pacs.008 | Structured party data (name, address, identifiers), coded payment purpose, extended remittance information, UETR for tracking |
| Interbank payment | MT 202, MT 205 | pacs.009 | Clear payment chaining and an explicit cover payment instead of manual matching |
| End-of-day statement | MT 940, MT 950 | camt.053 | End-to-end references carried through to the statement: reconciliation can be automated |
| Intraday statements and debit/credit notifications | MT 941, MT 942, MT 900, MT 910 | camt.052, camt.054 | Real-time cash monitoring, with a breakdown of fees deducted |
The G20 targets: what was promised and where things stand
In October 2021, the FSB published quantitative targets endorsed by the G20, with a common deadline of end-2027. Each target sets a numerical level for cost, speed, access, or transparency, so progress can be measured year to year. Remittance costs are the exception: that target is aligned with Sustainable Development Goal 10.c, with a 2030 horizon. The FSB tracks the indicators every year. Its progress report of October 9, 2025, shows faster wholesale payments and flat costs. The FSB itself says satisfactory improvements are unlikely to be achieved globally by the 2027 deadline.
| Challenge | Target | 2023 | 2024 | 2025 |
|---|---|---|---|---|
| Retail payment cost | Average cost of 1% or less and no corridor above 3% by end-2027 | 23.7% of corridors above 3% | 24,1 % | 18.3%, with average costs of 1.6% B2B, 1.8% B2P, 1.9% P2B, and 2.6% P2P |
| Remittance cost | Global average cost of a $200 transfer at 3% or less by 2030, no corridor above 5% | 6,3 % | 6,4 % | 6.5%, with 19.3% of corridors still above 5% on the SmaRT average |
| Wholesale speed | 75% credited within an hour, the rest within one business day, by end-2027 | 53,8 % / 92,7 % | 50,6 % / 92,0 % | 54,6 % / 93,2 % |
| Retail speed | Same target for retail payment services | 34,2 % / 74,0 % | 33,5 % / 69,0 % | 35.4% / 67.3% (the second indicator is falling) |
| Remittance speed | Same target, in every corridor | 52,6 % / 76,6 % | 53,1 % / 76,6 % | 54,4 % / 76,3 % |
| Access | More than 90% of people who want to send or receive a remittance have access to a service, by end-2027 | 73.8% of adults have a transaction account | n/a | 78.7% (Global Findex 2025) |
| Transparency | Total cost, expected speed, tracking, and terms disclosed by all PSPs, by end-2027 | 54.5% of services disclose cost and speed | 55,6 % | 62,9 % |
Remittances: who sends, where, and at what real cost
A remittance is money sent by a migrant worker to a household in their home country. These are small transfers, and international indicators measure their cost on a standard $200 transfer. Low- and middle-income countries received $685 billion in 2024, up 5.8% year over year. That is more than foreign direct investment and official development assistance combined (World Bank, Dec. 2024). Five countries receive a large share of the flows: India ($129 billion), Mexico ($68 billion), China ($48 billion), the Philippines ($40 billion), and Pakistan ($33 billion). Relative to the recipient country's gross domestic product (GDP), remittances equal 45% in Tajikistan, 38% in Tonga, and 27% in both Nicaragua and Lebanon (same brief, 2024 data). The cost of the transfer comes out of the amount the receiving household gets, so every percentage point saved raises income in the destination country by the same amount.
| Receiving region | Average cost | Services crediting within an hour | What the detail shows |
|---|---|---|---|
| Sub-Saharan Africa | 8.8%, the most expensive region in the world | Sharply down in 2025, after being the fastest region to receive in 2024 | A joint IMF–World Bank analysis of one corridor in the region points to reliance on US dollar cash (transport, security, logistics), a local tax on digital transactions that discourages the shift to digital, and non-bank PSPs without direct access to payment systems |
| Europe and Central Asia | 7.9%, among the highest | 66.2%, the highest rate of the regions covered | Expensive and fast: pricing does not follow technology. Beware the sample: the RPW database dropped ten low-volume corridors (including Russia → Ukraine, Estonia, Latvia, and Lithuania) and added Poland → Ukraine |
| Latin America and the Caribbean | 5.7%, unchanged since 2023 | 60.6%, up 8 points in a year | Clear proof that speed and price are separate problems: providers can get faster without cutting prices at all |
| Middle East and North Africa | Close to the global average | 52.3%, up nearly 5 points | P2P transfers out of the region remain among the most expensive in the world |
| South Asia | Stable, below the global average | 50,1 % | Highly competitive Gulf corridors; speed slipped slightly in 2025, and B2B there is still the slowest of any region |
Alternative rails: gpi, cards, fintechs, linked fast payment systems, stablecoins
A cross-border payment rail is the infrastructure that carries instructions and funds between payer and beneficiary. Five families of rails coexist, and none has replaced correspondent banking: they speed up, bypass, or duplicate its chain. They differ in the currencies covered, the amounts handled, the ways funds are received, and the traceability they offer. Comparing prices between two rails only makes sense for the same corridor, amount, funding method, and payout method.
| Rail | How it works | Strength | Limitation or point to check |
|---|---|---|---|
| Correspondent banking + Swift gpi | Correspondent chain, pacs.008 message, end-to-end tracking by UETR, and status updates fed to the tracker | Universal reach: all currencies, all amounts, with traceability and clear accountability | Stacked fees and the last mile: only 54.6% of wholesale payments credited within an hour (Swift/FSB, 2025) |
| *Card networks (push payments)* | Visa Direct, Mastercard Move: funds are pushed to a card, account, or wallet | Reach: 195+ countries and 150 currencies claimed by Visa Direct, 200+ countries by Mastercard Move | These reach figures are published by the networks themselves (Visa Direct and Mastercard Move documentation, accessed July 2026) and are not audited; the final cost depends on the funding method and the local acquirer |
| Money transfer fintechs | Local accounts in each country and internal netting: the cross-border payment becomes two domestic payments | Price: Wise reports an average take rate of 0.52% on $243.5 billion in cross-border volume for the fiscal year ended March 2026 | Scope: mostly digital corridors and liquid currencies. Comparing that 0.52% with the 6.5% global average would be misleading |
| Interlinked fast payment systems (FPS) | Two domestic real-time systems connect directly (PromptPay–PayNow since 2021, UPI–PayNow since 2023) or through a multilateral hub (Nexus) | On the Singapore–Thailand corridor, combined fees and FX margin fell from 13% of the amount before the link to less than 3% (FSB, 2025) | Volumes are still modest (just over 870,000 transfers in 2024, +16%), with daily limits and mostly bilateral links. Nexus Global Payments, a company set up in Singapore by five central banks (India, Malaysia, the Philippines, Singapore, Thailand) and announced on April 3, 2025, aims to go live in 2027 (FSB, Oct. 2025) |
| Stablecoins | Token backed by reserves, overwhelmingly in US dollars, transferred from wallet to wallet on a public blockchain | Available 24/7 with no intermediary: on sub-Saharan corridors, a $200 transfer would cost about 60% less than through traditional channels (Chainalysis 2024, cited by the BIS, Annual Economic Report 2025) | The BIS finds that stablecoins fail the three tests of money (singleness, elasticity, integrity); validation fees are volatile, pseudonymity is at odds with compliance, and the travel rule burden falls on platforms |
Compliance slows everything down: sanctions, AML/CFT, the travel rule
The time a cross-border payment takes is driven by compliance checks, not by the speed of the rails. Each link reruns the same checks, one after another, on data it did not produce. These checks cover whether the message is complete and consistent and whether the amounts are permitted. They also confirm that the beneficiary is not on a sanctions list, and they include AML/CFT screening. The BIS describes this setup as integrity safeguards layered on a process built on bilateral trust, with a large amount of duplicated effort. A problem caught late in the chain forces the earlier links to unwind what they have already done.
- Travel rule. FATF Recommendation 16, revised in June 2025, requires originator and beneficiary information to travel with the payment from end to end, rather than being split across a series of separate domestic payments.
- A threshold and a format. The standardized information requirements apply to cross-border payments above a threshold that cannot exceed USD/EUR 1,000, and the data must be structured according to messaging standards, ISO 20022 in particular. This is where compliance and technical migration meet: the same data serves both.
- Beneficiary consistency check. The receiving bank must detect misdirected or fraudulent payments, through after-the-fact verification, anomaly monitoring, or upfront verification such as confirmation of payee. In the EU, verification of payee (VOP) has been mandatory since October 9, 2025, under the Instant Payments Regulation (EU) 2024/886.
- Net settlements and batches. The FATF has clarified that intermediaries do not have to “unbundle” a net settlement and that aggregated transactions can be handled the same way, which is good news for internal netting models.
- Crypto-assets. In the EU, Regulation (EU) 2023/1113 has extended the travel rule to crypto-asset transfers since December 30, 2024, with no minimum amount and enhanced checks above €1,000 for self-hosted wallets (EBA guidelines EBA/GL/2024/11).
- Timeline. Full implementation of revised Recommendation 16 is expected by end-2030; the implementation guidance went out for public consultation in June 2026, with comments closing on August 21, 2026 (FATF).
Pitfalls in practice and a practitioner's checklist
The operational problems with cross-border payments show up after go-live, and service agreements do not describe them. The table below maps each symptom to the technical cause behind it and the fix to request from the bank or payment service provider (PSP).
| Symptom | Root cause | Fix |
|---|---|---|
| The beneficiary receives less than the amount sent | ChrgBr = SHAR (formerly :71A:SHA): each intermediary takes a fee along the way, and the receiving bank adds its own | Use DEBT / OUR when the commercial contract requires a guaranteed net amount, and budget for receiving-side fees (0.1% to 1.3% of the amount) |
| The actual cost is several points above the quoted price | The FX margin built into the rate: 87% to 99% of total cost depending on the use case | Require the applied rate and that day's benchmark rate, track the margin in basis points by corridor, and include it in the annual pricing review |
| No one can tell where a payment is | No end-to-end identifier stored on the customer side | Store the UETR for every payment and make it available to support: that is what lets a non-bank team use gpi tracking |
| Waves of rejects after a format change | Free-text address lines, truncated names, country missing from its dedicated field | Clean up counterparty master data before Swift’s new deadline (pushed back past November 2026): country (ISO 3166-1 alpha-2) and town in their own fields, not repeated in a free-text address line |
| Sent Friday, credited Tuesday | Missed cut-off, local holiday at the destination, and a beneficiary leg processed manually | Document the cut-off and holiday calendar for each corridor, then measure the actual value date yourself, not the promised one |
| Reconciliation breaks on returned payments | A return comes back net of fees deducted along the way, under a new reference, sometimes at a different exchange rate | Match returns by UETR or original reference, and book the difference as fees and FX differences, never as income |