Building an International Banking Strategy

Most international banking structures are built in the wrong order. Institution first, because someone made an introduction. Jurisdiction second, because that is where the institution is. Function last, discovered retrospectively as the structure is asked to do things it was not designed for.

The correct sequence runs the other way. Function, then jurisdiction, then institution. The method below is what that looks like in practice.

Step one: map what exists

Nothing useful can be decided before the current position is written down. One table, one row per relationship, six columns: owning entity, institution, jurisdiction, currencies, average balance, function performed today. Add a seventh column for the function the relationship was opened to perform.

Where columns six and seven disagree, the structure has drifted. Where column six is blank, there is a relationship without a job. In group structures, run the same exercise against the legal entity register: entities with accounts and no activity, and entities with activity and no accounts, both appear immediately.

Step two: define the functions the structure must perform

Six functions cover most internationally active structures. Define each in terms of the business as it actually operates, not as an abstraction.

• Operating. Commercial receipts and payments, in the currencies genuinely traded, at the volumes genuinely processed.
• Treasury. Reserve and working capital, held apart from operating flow so the position is legible at any moment.
• Settlement. Cross-border execution, with named rails and named counterparty destinations.
• Trade. Documentary credits, standby letters of credit, demand and performance guarantees, sized to real transactions.
• Custody. Securities, funds and precious metals, segregated from transactional accounts.
• Structural. Holding companies, special purpose vehicles, escrow and fiduciary arrangements, banked separately so that ring-fencing is real rather than notional.

Two disciplines apply. State the requirement quantitatively where possible: payment volumes, currency pairs, instrument sizes, settlement deadlines. And mark which functions are genuinely critical, meaning the business stops without them, because that determines where redundancy is justified.

Step three: assign jurisdictions to functions

Each function requires a different combination of regulatory standard, legal certainty, infrastructure, reputational standing and stability. Custody is decided by enforceable property rights. Settlement is decided by correspondent depth. Trade is decided by whether counterparties accept the jurisdiction's paper. Operating is decided by proximity to the commercial activity and the payment systems that serve it.

Assign function by function, then look at the result as a whole. If every function has landed in one jurisdiction, either the business is genuinely single-market or the analysis has been performed under a constraint that was not stated.

Step four: select institutions

Only now does the choice of bank arise, and it is a procurement exercise against a written specification. Five questions separate institutions that can perform a function from institutions that will say they can.

1. Which payment rails do you reach directly, and which through intermediaries? Names, not categories.
2. What is the escalation path when a payment does not arrive, and who by name answers at each stage?
3. Which of my counterparties already accept instruments issued by you?
4. What is your periodic review cycle, and what will you ask for each time?
5. Under what circumstances have you exited clients in my sector or jurisdiction in the last three years?

The fifth question is the one that is rarely asked and most predictive. An institution that answers it directly is a different proposition from one that treats it as impertinent.

A bank is a component. The strategy is the assembly.

Step five: sequence the transition

A designed structure is reached by migration, not by announcement. The order matters more than the speed.

1. Open and test new capability before reducing anything. Test means a real payment to a real counterparty on a real deadline.
2. Migrate one function at a time. Settlement and payroll last, because they are the least forgiving.
3. Run old and new in parallel for at least one full reporting cycle, monitoring the old account for unexpected receipts.
4. Trace every inbound instruction before closing: standing orders, direct debits, counterparty payment files, merchant settlement, tax and payroll authorities.
5. Close deliberately, in sequence, with written confirmation and residual balances swept to a named destination.

Step six: put the structure on a review cycle

Architecture decays because the business changes and the banking arrangement does not. An annual review holds the design against reality, and specific events should trigger one out of cycle.

Documentation is part of the structure

A structure that exists only in one person's understanding is not a structure, it is a dependency. The written record should state, for each relationship, its function, its owning entity, its authorised signatories, its escalation contacts, and what happens to it on the incapacity or death of the principal.

The test is blunt. Could a competent successor operate this structure next week, using only the documentation, without telephoning anyone who has left? For most groups and most families the honest answer is no, and the gap is discovered during a transition, when it is expensive and public.