I spent the start of last week at TMANY’s New York Cash Exchange. I was there to present in Treasury Tank, which we won, and that is the last you will hear about it here.
The useful part was the agenda. It was billed as treasury technology. Sit through it and the subject underneath almost every session was narrower, and a good deal older, than the billing: how do you build a treasury practice that holds when something arrives you did not plan for. Tariffs and transfer pricing had their own session. Geopolitics ran through the lunch keynote and again through a session on managing risk in a fractured world. The economic outlook panel spent its time on tariff-driven pressure and a dollar under strain.
None of it was framed as prediction. All of it was framed as readiness.
And underneath readiness, session after session, sat the same question wearing different clothes. Somebody in the business looks at a currency number and asks: how much of that was the market, and how much was us?
It is a fair question, and it is usually not an accusation. It is also, in most finance functions, unanswerable in the time you get to answer it.
Here is why. None of it is anyone’s oversight.
01“How much” and “where from” come off different data
Almost every treasury can produce the first number quickly. It sits on a report that was built to produce it.
The second number is not on that report and never was. Knowing where a currency movement came from means holding two things against each other: what the rate did, and what the underlying balance did over the same window. The first is a market series anybody can get. The second is a description of the business changing shape, and it lives in the ledger, in intercompany, in whatever the operating side did that quarter with pricing and terms and suppliers.
Nobody built a report that puts those side by side, because the reporting was built to close the month. It does its job. It simply cannot answer a question it was never designed around.
Fig 1 · Two numbers
The second number is not on that report and never was.
Nobody built a report that puts those side by side, because the reporting was built to close the month.
02The business changes shape without announcing it in currency
This is the part that catches people, and it came up repeatedly in the tariff sessions without once being named as a currency problem.
Take a supplier who reprices because trade policy moved. Your cost base moves. Your margin moves. Your exposure moves. And if that supplier invoices the buying entity in its own functional currency, nothing anywhere in your books turns a different colour. There is no foreign-currency transaction to catch, because there is not one. The exposure changed economically and stayed invisible in the place anyone would look for it.
That is not a hedging failure and it is not a reporting failure. It is a category problem. A real change in currency risk arrived through a door marked procurement.
Fig 2 · Procurement
A real change in currency risk arrived through a door marked procurement.
The exposure changed economically and stayed invisible in the place anyone would look for it.
03Nothing underneath knows the sources are describing the same thing
Answering “where from” means reading across systems that were never built to agree with each other. The entity master, the ledger, the trade file, the intercompany schedule, whatever the operating side keeps in its own format. The same customer is spelled differently in each. The same entity carries different codes. The same balance is stated on bases that do not line up.
A person can reconcile that. A person does, every quarter, slowly, and usually the same person. What nobody has is something underneath that knows which of those sources are describing the same object in the first place, and where, in a business of this particular shape, a balance of this particular kind tends to sit.
Fig 3 · Same object
The same customer, entity and balance, stated differently in each.
What nobody has is something underneath that knows which of those sources are describing the same object in the first place.
04This is the gap AI is currently being asked to cover, and the honest sessions said so
There was a great deal of AI at NYCE. The sessions worth sitting in were the ones about what has to be underneath it.
One roundtable was built on a question I have not stopped turning over: do treasury teams trust automated numbers enough to send them to the CFO without re-checking them first? Its premise, printed in the agenda, was that most still quietly re-check.
That is not a capability problem. A model can categorise, match and forecast perfectly well. You re-check a number because you cannot trace it. And you cannot trace it because the thing producing it does not know what its own inputs mean: which sources are the same object, which balance belongs to which entity, where exposure hides in a business like yours.
Put a model on top of that and what you get is a faster answer you still have to check by hand. Which is the same answer, later.
Fig 4 · Underneath
A faster answer you still check by hand is the same answer, later.
You re-check a number because you cannot trace it.
Why the split matters more than it sounds
Leave the number undivided and what gets discussed is the net outcome. A net outcome has a rate path inside it, and a rate path was nobody’s to know. A view still gets formed on it, about the program, and through the program about the people running it. Nobody intends that. It is what happens to a number that arrives whole.
Split it, and the subject of the conversation changes. What is left to discuss is what the business actually did: where the balance grew, which entity, which supplier or customer relationship changed, what was priced in at the time it was priced in. Those are things somebody chose, and things that can be taken into account when the strategy is set next time. A rate path cannot.
Fig 5 · The split
Split it, and the subject of the conversation changes.
What is left to discuss is what the business actually did.
Which part of that remainder was knowable at the moment each decision was made is a further test, and a sharper one than it looks. It has an issue of its own coming.
Saying it out loud
The version that survives the room is short.
This much of the move was the currency. This much was the business changing shape, and here is where. The first part was never ours to decide. The second part is the conversation worth having.
Because the question is rarely did you get this right. It is can you tell me what this is. Only one of those has an answer, and it happens to be the one being asked.