The question rarely arrives on its own. It arrives with a date attached. The board meets on Thursday. Diligence closes Friday. The lender wants the schedule before month end. Somebody needs to know what currency did to the quarter, why, and what changes as a result.
Most teams can answer it. The constraint is rarely whether the number exists somewhere in the building. It is how long the answer takes to assemble, and whether it survives the first question asked of it.
Scale is a misleading word for this
It suggests the demands on treasury rise when the business does, and that a company holding revenue flat is holding its currency problem flat too. It does not. The work does not track growth. It tracks shape.
A business changes shape when it adds an entity, buys a company, moves production, switches a supplier, or sends a product to a new market. Some are growth. Most are not. All change which currencies it earns and spends in, and so change the question treasury must answer.
Three things have moved at once, and only one is inside the company.
Fig 1 · Three things moved
Three things have moved at once. Only one is inside the company.
Each is manageable alone, but together they produce one specific failure.
Source: WTO and IMF, July 2026; UNCTAD trade update, January 2026.
Before Anyone Asks
01The world became less predictable, and then stayed there
The WTO and IMF jointly index trade policy activity. They reported in July that activity over the first five months of 2026 ran at nearly twice its 2024 level, the highest in the series, driven by restrictive rather than facilitating measures.
The important number there is not the size of the move. It is that this is a level, not a spike, and treating it as weather to wait out has been the wrong call for two years running.
02The business is changing shape underneath you
UNCTAD’s January trade update notes that nearly two thirds of world trade moves inside value chains now being reconfigured into new hubs and routes.
That is about the world. Inside one company it lands far more specifically. A tariff makes a sourcing route uneconomic, so procurement moves to a supplier in another country. Production shifts closer to the end market. The revenue mix relocates, and the fastest-growing line is no longer the one the plan was built around. Cost bases end up in currencies the revenue line does not share.
None of that is growth. Each moves part of the business into a different currency, and none arrives with a note to treasury explaining what just happened.
This is why boundaries matter and volume does not. Doubling revenue inside one functional currency changes almost nothing. Adding one changes a great deal, because every question from then on must cross a boundary that was not there before.
Fig 2 · Boundaries, not volume
Boundaries matter. Volume does not.
Each change of shape moves part of the business into a different currency.
Before Anyone Asks
03The time to answer collapsed
Acquisitions close between planning cycles, bringing new entities and new pairs, and their questions arrive during diligence, on the transaction’s schedule rather than the close calendar.
Some of the delay is not treasury’s to fix. A position is sized against a forecast produced elsewhere, usually in FP&A, and when that forecast moves the gap lands against the program rather than the input. That deserves its own issue.
A practice that needs a month to produce a traceable exposure picture is not slow. It is not in the conversation.
Fig 3 · The clock
The question arrives with a date attached.
The dates arrive before a month-long answer does.
Before Anyone Asks
What the three do together
Each is manageable alone. Together they produce one specific failure: the natural offset.
An offset is two positions in different parts of the group that cancel out. A receivable in one entity against a payable in another. It is not a strategy but a property of the configuration, either there or not.
Not every pair that looks like one is one. Inside a single entity two positions net cleanly. Across entities keeping their books in different currencies they do not net at all.
Which is why a change of shape takes it away first. Change a supplier’s country and an offset that was there last quarter is gone. Add an entity and a new one appears. Neither announces itself, because an offset is a fact about the relationship between two entities rather than about either one, and the relationship is not what gets reported.
A process that assembles exposure the way the close is assembled, entity by entity, reports every position accurately and still cannot see the pair. Nobody made an error. There was no column for it.
Fig 4 · The natural offset
Every position reported accurately. Nowhere to see the pair.
An offset is a fact about the relationship between two entities, not about either one.
Before Anyone Asks
The hard part is not the number. It is the movement
The hard part is the question straight after the number. It moved. Which entity, which currency, which part of the business, and how much was the market rather than something the company did to itself. Producing the figure again does not answer it. It is answered by holding two points in time next to each other and seeing what changed in between.
A workflow built by hand cannot do that, and the reason is not effort or care. Anything built by hand is built toward a particular answer. The steps, the groupings, the layout, all of it takes its shape from the question that was live at the time. Ask a question shaped slightly differently and that work does not adjust to it. It has to be unpicked, several steps back before it can go forward, and the choices made along the way get reconstructed on the way through, because nobody writes down every one of them. Nobody could.
So the business changes shape, and the analysis cannot follow it.
That is what traceability is for, and it is not a record kept for its own sake. It is an analysis that can be asked a different question without being rebuilt. A figure tied back to the books can be taken apart into what produced it and put back together a different way. One that cannot be tied back can only be reported.
Fig 5 · The question after the number
It moved. What changed in between?
A figure that ties back to the books can be asked a different question.
Before Anyone Asks
The part worth saying plainly
You cannot build the ability to answer during the meeting in which you are asked.
That is the argument, and it is why this is not a growth story. Infrastructure for scale is usually justified by volume. The justification here is different: the shape can change without warning, and each change quietly invalidates part of the answer that was true last quarter.
A practice that can produce its exposure as it stands now, tie it back to the books, and ask it a different question without rebuilding it, is not a faster version of the same practice. It is the only one that stays correct while the shape moves underneath it.
None of that is a view about how much to hedge. It is the ability to say what is true, quickly enough for it to matter.
Saying it out loud
The question at the close is rarely the hard part. The hard part is the date attached to it, and that date keeps moving.
It used to be weeks. The question came at the board meeting, the answer in the pack that followed. Not now. The question arrives the day the number lands, and the answer is expected in the room, because everyone in it is already looking at the figure.
Fig 6 · The date keeps moving
It used to be weeks. Now the answer is expected in the room.
The question and the answer have moved into the same room.
Before Anyone Asks
The standard after that one is not faster reporting. It is understanding how the shape of the business is changing while it is still changing, and what part of that is currency, before it has finished arriving in the results.
None of that is a knowledge problem. It is a structural one, which means it closes the way structure closes: slowly, deliberately, and never in the week you need it.
Before Anyone Asks