Paying and collecting across borders. 7 chapters and a final quiz.

Chapter 1. Mapping your corridors and reading the chain.

A multi-country company does not have “international payments.” It has corridors. A corridor is defined by four variables: origin country, destination country, currency, and beneficiary type. Yet two neighboring corridors rarely share the same lead time, cost, or reject rate. So the treasurer’s first job is not to pick a provider. It is to inventory the company’s corridors, then rank them by annual value and by number of incidents.

The underlying mechanism fits in one sentence. A cross-border payment is a chain of domestic payments, stitched together by accounts that banks hold with one another. The sending bank holds an account with a correspondent, denominated in the destination currency. That account is its nostro, and the correspondent records it in its own books as a deposit, the vostro. Nothing crosses the border. Book entries change hands, and a full compliance check is rerun at every link in the chain.

• Measure the value date at the beneficiary, not the execution date your bank shows. They are two different dates, and only the first one is what the supplier experiences
• Count your links. Ask for the named correspondent chain on your top five corridors. A serious bank will provide it
• Date your cut-offs. Document each cut-off by currency and by channel, never in aggregate
• Rank your corridors by annual value, then by number of incidents. Where the two rankings meet, you find the two or three projects that pay off
• Reject the average. A two-day average lead time can hide 80% of payments arriving within an hour and 20% taking five days

Chapter 2. Sending a payment order that goes through the first time.

A rejected payment costs far more than an expensive one. You have to track it down, fix it, resend it, explain the delay to the supplier, and absorb the FX difference on the second attempt. The BIS describes these exception fees as frequent and costly (Bulletin No. 87, 2024). Yet almost every reject stems from data you control yourself, in your counterparty master data. Cleaning that data up pays off more than renegotiating your pricing.

What the message must carry

<CdtTrfTxInf>
  <PmtId>
    <EndToEndId>INV-2026-0042</EndToEndId>        <!-- YOUR reference: it comes back on the statement -->
    <UETR>eb6305c9-1f7f-49de-aed0-16487c33b1e2</UETR>  <!-- unique end-to-end identifier -->
  </PmtId>
  <ChrgBr>DEBT</ChrgBr>                            <!-- DEBT | CRED | SHAR | SLEV: see chapter 3 -->
  <Purp><Cd>SUPP</Cd></Purp>                       <!-- coded purpose: speeds up screening -->

  <Cdtr>
    <Nm>EXACT LEGAL NAME OF BENEFICIARY</Nm>       <!-- the name on the contract, not the trade name -->
    <PstlAdr>                                      <!-- STRUCTURED ADDRESS: dedicated fields -->
      <StrtNm>Jalan Tun Razak</StrtNm>
      <BldgNb>18</BldgNb>
      <PstCd>50400</PstCd>
      <TwnNm>Kuala Lumpur</TwnNm>                  <!-- town: mandatory field -->
      <Ctry>MY</Ctry>                              <!-- country as ISO 3166-1 alpha-2 -->
    </PstlAdr>
  </Cdtr>

  <CdtrAgt><FinInstnId>
    <BICFI>XXXXMYKLXXX</BICFI>
    <ClrSysMmbId>...</ClrSysMmbId>                 <!-- destination country's national code, when required -->
  </FinInstnId></CdtrAgt>

  <CdtrAcct><Id><IBAN>...</IBAN></Id></CdtrAcct>   <!-- or <Othr><Id> outside the IBAN zone -->

  <RmtInf><Strd><RfrdDocInf>                       <!-- STRUCTURED remittance info: the key to reconciliation -->
    <Tp><CdOrPrtry><Cd>CINV</Cd></CdOrPrtry></Tp>
    <Nb>2026-0042</Nb>
  </RfrdDocInf></Strd></RmtInf>
</CdtTrfTxInf>

• The exact legal name, as it appears on the contract and in the local business register. Not the trade name, not an internal abbreviation
• The address in its dedicated fields: town, and country as an ISO 3166-1 alpha-2 code, never copied into a free-text line
• The local institution code from the table above, validated with a regular expression at data entry, not at the moment of sending
• The coded payment purpose (Purp), so no analyst has to guess what the transaction is
• The structured remittance reference, so the beneficiary can match the invoice without writing to you
• The UETR, stored in your own system, indexed by invoice number. Without it, no claim can be investigated

One last rule is counterintuitive and pays off: never let your operations staff fix counterparty details directly on the payment screen. The fix has to go back into the master data. Otherwise the same error will come back on the next payment, and the one after that. A counterparty is created once, verified once, and paid a hundred times. Then track your reject rate by corridor every month. No metric moves faster once a team takes real ownership of it.

Chapter 3. OUR, SHA, BEN, and calculating landed cost.

Three letters decide who pays what and, above all, what the beneficiary actually receives. In ISO 20022, they have become four codes in the ChrgBr field. The choice is not a formality. It determines whether your supplier receives the exact invoice amount or an amount reduced by fees deducted along the way by banks the supplier never chose. A biller that expects an amount accurate to the cent cannot work with shared charges.

# The percentages are MEASURED AVERAGES for the cross-border B2B segment
# (FXC Intelligence for the FSB, March 2025; FSB-World Bank survey, 2025)

sending_fees       100,000 x 0.2%           =    200 EUR    # visible: this is the negotiated rate
fx_markup          100,000 x 1.4%           =  1,400 EUR    # invisible: built INTO the rate
arrival_fees       100,000 x 0.1 to 1.3%    =  100 to 1,300 # outside your contract
lifting_fees       read from <ChrgsInf> in the pacs.008     # depends on actual routing

landed_cost = 1,700 to 2,900 EUR, i.e. 1.7% to 2.9% of the amount
price quoted by the bank = 0.2%

# WHAT THE CALCULATION SHOWS
#   The price you negotiate covers the SMALLEST item on the bill.
#   The two items that matter - FX markup and arrival fees - appear
#   in no price list. They have to be measured, corridor by corridor.

• Require the applied rate and a timestamp on every transaction advice, not just the amount debited
• Have `ChrgsInf` included in the data your bank sends back to you. In ISO 20022, the agent that deducted a fee is identified, which the MT format did not allow
• Isolate exception fees in a dedicated general ledger account: rejects, repairs, returns. If nobody tracks them, they disappear into general bank charges
• Compare total cost, not line items: a provider that is cheaper on fees but more expensive on FX costs you more
• Reserve `DEBT` for flows with a contractual net amount, and accept that it costs more to send. That is the price of certainty

Chapter 4. Tracking, chasing, and recalling: gpi and incident handling.

gpi is a service rulebook layered on top of existing correspondent banking. Participating banks commit to same-day credit, fee transparency, and passing on remittance data in full. The technical building block that makes all of this usable is the UETR, a unique reference carried in the message header since November 18, 2017, which survives the entire chain. Without it, a claim becomes an investigation by phone.

That leaves governance. A cross-border incident involves at least three teams (treasury, accounts payable, and procurement) and gets lost in the gaps between them. Appoint a single owner per corridor, then set an internal deadline for the first reply to the supplier, regardless of how long the bank takes to respond. Finally, keep an incident log with each root cause. After one quarter, three causes will account for three-quarters of the cases.

Chapter 5. Choosing between a bank, a fintech, and a regional rail.

No rail wins everywhere, so you choose corridor by corridor, not provider by provider. The same group can pay its European suppliers through its bank, its freelance contractors through a fintech with local accounts, and its intra-African trade over a regional rail. The question is never “which rail is best.” It is who actually credits the beneficiary on this specific corridor, how fast, and at what landed cost.

Chapter 6. Collecting, converting, and repatriating.

Collecting is a different job from paying. When you pay, you choose the rail. When you collect, your customer chooses, and will always take the path that is easiest for them. An Indonesian customer will gladly pay into an Indonesian account, but will pay poorly, late, or not at all into a foreign account that requires FX paperwork. So go get the money where it is, in the currency it is in.

• Map exchange controls country by country: prior approval, import documentation, deadlines for repatriating export proceeds, caps on dividends
• Date your trapped cash. Cash that cannot leave the country is not really cash, and group reporting must show it as such
• Look for natural hedging before converting: paying a local supplier out of local receipts eliminates two conversions and two markups
• Net intercompany flows where local law allows it, and document that permission before the first netting cycle
• Collect and pay in the same currency whenever you can, rather than converting in both directions on the same corridor

The scale of the problem is public. PAPSS puts the funds trapped for African airlines alone at more than $2 billion. It estimates the cost of routing through hard currencies across the continent at $5 billion a year (PAPSS / Afreximbank, July 7, 2025). These amounts are not bank fees. They are revenues that have been collected and booked, yet remain unavailable to the company that earned them. A treasurer who discovers this item at year-end close has already lost control.

Chapter 7. Handling sanctions and compliance without blocking your own flows.

A cross-border payment is not slow because the rails are slow. It is slow because every link repeats the same checks, one after another, on data it did not produce: sanctions screening, AML/CFT checks, beneficiary consistency checks. You cannot remove these checks or shorten the chain that runs them. All you can do is give them what they need to reach a decision quickly.

• Screen before you send, not after a reject. An internal check against your own lists takes a few seconds and spares you a case at the correspondent
• Keep ownership data up to date for your counterparties in high-risk jurisdictions, and refresh it on a fixed schedule
• Document every release. The file you build once serves for later cases at the same correspondent
• Give notice before the first payment on a new or unusual corridor. A bank that has been warned processes; a bank caught by surprise holds
• Never resend a blocked payment unchanged through another channel before you understand why it was blocked

One last point of method. The number of active correspondent relationships fell by about 30% between 2011 and 2022, even as volumes grew (BIS, Bulletin No. 87, 2024). On a fragile corridor, everything now goes through one or two intermediaries, which alone set the corridor’s price, lead time, and risk policy. A prudent treasurer knows their names before needing them.