How to negotiate better FX rates with your bank

The quoted number and the delivered number

A bank quote and a bank commitment are different objects. On a multibank platform, margins are typically quoted as indicative. Two large US banks, asked during a 2026 multibank review to guarantee their quoted FX sales margins, both declined in writing. One declined on stated policy grounds. The other quoted attractive numbers and reserved the right to adjust them on the platform where the client actually traded.

A negotiation can move the quoted number. Whether the delivered number moves with it is a separate question, and the closing paperwork does not answer it.

What a bank can and cannot guarantee

A disclosed sales markup agreement is a real compliance construct, not a brush-off. Bank policies that permit one typically carve it up the same way: allowed on spot, not beyond spot, and not on trades executed through a third-party platform.

That pattern matters for how you open. A demand for a locked spread on a multibank platform is an ask the bank has to refuse, and a refusal early in a review costs credibility on everything else in the sheet. Asking again at the same level gets the same answer.

Read a favourable-looking answer carefully. A bank that quotes tight numbers and then reserves the right to adjust pricing on the platform has given you a sheet you cannot rely on for the channel that carries your flow. That version is harder to spot than a flat refusal, because the numbers are on the page.

Ask for what bank policy permits

When a guaranteed margin is off the table, several structures are not. Each is something a bank can actually sign:

• a most-favoured-pricing commitment rather than a fixed number
• a scheduled pricing review with agreed remedies if realised margin drifts
• tiered pricing against awarded volume
• bilateral execution on a defined subset of flow, where disclosed markup is permitted
• disclosed markup on spot trades, the one place policy usually allows it

A refused guarantee closes an instrument and nothing more. In the 2026 review, the message that declined the guarantee also asked to be awarded additional flow. A bank that wants more volume has reason to keep negotiating the structures above.

Two of these are worth asking for in writing, because most treasury teams have never requested either.

Convert every pip quote to basis points before judging anything

Banks quote FX margins in pips. The cost of a pip depends on the level of the exchange rate it sits on, so one uniform pip quote is a different price on every pair.

For pairs quoted to four decimal places, a pip is 0.0001, and one pip expressed in basis points is 1 divided by the rate. JPY pairs are quoted to two decimals, so a pip is 0.01 and one pip is 100 divided by the rate. Check which convention your sheet uses before applying the arithmetic.

The arithmetic is checkable on any trade. One pip on a EUR 10 million EUR/GBP trade is EUR 1,176 of margin. The same one pip on USD/CNH at 7.20 is USD 139 per USD 10 million. A flat sheet quoting one pip everywhere charges 61% more on EUR/GBP than on USD/CAD, and more than eight times what it charges on USD/CNH. The largest flows on a sheet are not necessarily the best priced ones.

Basis points carry the comparison. For precision, ask for parts per million: 1 basis point equals 100 PPM, and at the low single-digit basis point margins typical of a large corporate book, PPM resolves differences that basis points round away.

Spot and forward margins are usually additive

Sales sheets quote a spot margin and a forward margin in separate columns. On most sheets a forward trade pays both, so the all-in forward margin is the spot pips plus the forward pips. Some sheets do quote an all-in forward margin instead. Establish which you are reading, because the difference is large.

The tell is a forward column that reads lower than its spot column. Forwards are normally the costlier leg, so a cheaper-looking forward is the signature of an add-on rather than a standalone price. In the 2026 review, one pair was quoted at 5 pips spot and 3 pips forward, and the all-in forward cost was 8. Read alone, that forward column showed under 40% of the true cost. How far a forward column understates the total depends entirely on the ratio between the two columns.

Ask for tenor alongside the margin. A 3-pip forward markup on a one-month is a different cost of carry from the same markup on a twelve-month, and a sheet that quotes forward margin without stating the tenor band is not comparable to one that does.

Do not ask the bank whether its columns are additive. That question lets the bank choose whichever reading suits it, and if it would have honoured the cheaper reading, you have invited it not to. Ask instead for the proposal restated as an all-in margin per pair.

Award per pair, not per bank

Separate the award decision from the pricing decision. The natural output of a multibank review is a per-pair allocation, and treasury teams sometimes signal this themselves: the client in the 2026 review told its banks it might direct specific currency pairs to a single bank.

The reason is in the shape of the numbers. A proposal that shows a net gain can still lose on most of its pairs. In that review, one bank's package produced its entire net gain on two currency pairs, while the remaining four were priced above the client's existing fills. Accepted whole, the package would have paid for the strong pairs with the weak ones. Treat every proposal as a menu, and award each pair to the bank that wins it.

Keep your current pricing out of the room

When a challenger's quote loses to your incumbent, do not say so. “Worse than we pay today” hands the bank your floor, and a bank that knows the floor can price to just beat it rather than to its best. State the outcome only: these pairs will not be awarded on this pricing. The effect is the same and nothing is disclosed.

Apply the same scrutiny to savings claims made to you. If a claimed saving is computed against your blended average cost, and your incumbent fills some pairs tightly, the blended average sits above those fills. Moving flow to the challenger can then raise your cost on exactly the pairs the incumbent was winning. Decompose any claimed saving by pair and by counterparty before acting on it.

Extend the discipline to everyone who corresponds with the bank. Explaining your own execution mechanics to a quoting desk gives it material to argue with, and the most common use is to contest the volume base the whole negotiation rests on. Agree internally what stays internal before colleagues are in direct contact.

Read the scope

Flow a proposal does not cover is unbenchmarked. One proposal in the 2026 review applied only to platform-executed trades at short tenors, for a defined subset of group entities. It excluded a payments portal and anything done direct with the sales desk, both of which remained priced one-off. List the exclusions before comparing anything.

Expect an early question about sole-bank status. In that review, a bank opened by asking the client to confirm it intended to stop trading on the platform and execute with the mandated bank instead. Sole supply removes the comparison that disciplines pricing, so close the question plainly and expect it in another form later.

Note the channel each answer arrives on. A reply sent privately to an advisor rather than into a thread with the client copied is often more candid, and easier to reopen without an audience.

Three more requests to put in writing

Each traces to a mechanism above, and each can go into an email as written.

None of these is an ask the bank must refuse. Each closes an ambiguity that otherwise stays open in the bank's favour.

Where the negotiation stops

Four of the mechanisms above have a negotiated fix. The unit can be restated. Additivity can be resolved. Per-pair award prevents strong pairs subsidising weak ones. Scope can be closed.

The fifth is different. A guaranteed margin is available bilaterally, and on spot, where policy permits disclosed markup. It is generally not available on a multibank platform. So a treasury team that wants a contractual guarantee has a real option: move the flow it wants guaranteed to bilateral execution, and keep the platform for the flow where competition matters more than certainty. That is a genuine third control and it belongs in the decision.

What it does not do is tell you whether the pricing you agreed is the pricing you received. On platform flow the quoted margin remains indicative. On bilateral flow the commitment holds only as far as its own carve-outs. In both cases the difference between the agreed margin and the delivered margin is invisible on a trade confirmation, which shows one all-in rate and no fee line.

That gap closes with measurement: every fill scored against the rate the bank could actually have traded at in that moment, at your size and direction. The reference decides whether the number survives contact with the desk. The interbank mid is a midpoint nobody deals at, so a margin measured against it is easy to dispute. A margin measured against the bank's tradable rate is a fact the desk can check against its own book.

Measurement is not enforcement. Your enforcement is the ability to reallocate flow. Measurement is what makes reallocation defensible, and what tells you it is needed. That is the measurement Just runs.