Fintech, dissected: FX and what actually happens when money changes currency

Last time I walked through auth vs settlement, and the gap between a card being approved and money actually moving.

This one is FX. Different mechanism, same shape. The moment your customer sees something happen is not the moment it happens, and everything expensive lives in between.

Ask a founder how their FX works and you almost always get one sentence: mid-market plus a margin.

That’s the price. It isn’t the product.

The product is a chain of four steps, and the margin only survives if you’ve made a decision at each one. Most founders have made a decision at exactly one of them.

A quote is a position you haven’t covered yet

Start where the customer does.

They see a rate. That rate came from your provider, streaming prices derived from the interbank market through one or more liquidity providers, with your margin already inside it.

Then they accept it.

At that moment you’ve committed to delivering a currency at a price you’ve already fixed, before you’ve necessarily bought it yourself. And between their click and your execution, the market keeps moving.

That window is the whole game, and there are only three honest answers to who carries it.

You pass it straight through. The trade is executed back to back the instant the customer accepts. Your window is close to zero, your provider carries the risk, and you earn the margin for the distribution rather than the position. Most fintechs should be here.

You hold the quote. You give the customer thirty seconds, or two minutes, or the length of a checkout flow. That is a real position and it is yours. Perfectly fine, as long as the margin is priced for it and somebody knows it’s happening.

You batch. You collect trades and execute the net position later in the day. Better pricing, and a longer window where the market can move against you.

None of those is wrong, but they’re three different businesses.

The version I see go wrong is the fourth one: nobody chose. The behaviour came from a provider default, and the product was priced as if the window were zero when it wasn’t. Someone had priced the product before anyone had priced the risk.

Quote, execute, net, settle

Founders see the first step. The other three are where the economics get decided.

Quote. The price you show, margin included.

Execution. Someone buys the currency from a liquidity provider. This may be instant and matched to the customer’s trade, or it may be later and aggregated.

Netting. At volume you don’t execute a hundred small euro to dollar trades one at a time. You net them and execute the position, which prices better and lengthens your exposure.

Settlement. The currencies actually move between accounts.

That last step is where mental models break, in exactly the way they break on cards. Spot FX conventionally settles two business days after the trade. Same day and next day exist for the major pairs, at a price, with earlier cut-offs.

So your app says converted. Your ledger says converted. The currencies land two days later.

Approved is not paid, and converted is not settled. It’s the same lesson wearing different clothes.

If your product tells the recipient they’re paid today, someone is fronting liquidity across that gap. It’s your provider, your sponsor, or you. It’s worth knowing which, because whoever it is has priced it.

The calendar is a product decision

Two things that look like operations and are actually product.

Cut-off times. Every currency, rail and correspondent bank has one. Miss it and the value date rolls to the next business day. A payout your customer believes is same day quietly becomes next day, and they hear it from their recipient rather than from you. For less liquid currencies the cut-off can land early in your morning.

Weekends and holidays. The market closes Friday evening and reopens Sunday evening. If your app converts at 2pm on Saturday, you’ve quoted a rate nobody can execute until Sunday night, so you’re carrying the weekend. That’s a legitimate choice and plenty of consumer products make it deliberately, pricing the risk into the spread. The problem is making it accidentally.

Holidays aren’t shared, either. A currency’s home market holiday moves value dates for that currency alone, so your dollar flows run clean while another corridor sits there looking broken to everyone except the person who knows why.

Design for these and they’re product rules. Ignore them and they’re support tickets.

Where the margin actually lives, and what eats it

FX revenue rarely appears as a fee. It lives in the spread, the difference between the rate you got and the rate you gave.

Two things worth knowing about your own economics.

Your all-in cost isn’t the spread your provider quotes. An FX trade is sandwiched between a payment in and a payment out, and both have costs on their own rails. Model the FX margin, ship it, and then discover the payment legs took a chunk you hadn’t counted. Which rail the money arrived on changes what you actually kept, and that is a subject deserving its own article.

FX pricing is tiered, and nobody tells you when you cross a tier. The rate you were offered at launch is not the rate available at scale. Providers are rarely in a hurry to volunteer that. Ask, and keep asking.

And decide honestly whether the FX margin is your product or your subsidy. Some businesses earn on the conversion. Others use it to fund something else and price it near cost. Both work. Confusing the two is how you end up defending a number in a board meeting that was never designed to be defended.

FX as a feature, or FX as a product

This fork decides how much machinery you need.

FX as a feature. The customer holds one currency and pays out in another. You convert at the point of payment and send. No foreign balances, no overnight positions, no treasury function. The exposure window is short and most of it can be passed through. Most fintechs should start here and many should stay.

FX as a product. Multi currency balances. The customer holds euros and dollars and chooses when to convert. Perhaps a rate they can lock. Now you hold positions that outlive the transaction, you need somebody whose actual job is treasury, and you need a ledger that carries balances in multiple currencies cleanly. That last one is a larger piece of work than it sounds.

The trap is drifting from the first into the second without deciding to. It starts as a reasonable request: could customers keep their euros rather than converting straight away? Of course. Six months later there are overnight positions across four currencies and no name against them.

If you’re going there, go deliberately.

And remember the conversion doesn’t happen in a vacuum. Under a sponsored model the FX executes under your sponsor’s permissions with their FX partner, and the payment legs either side are theirs too. So your FX design lands in the flow of funds you hand them, and it gets read closely. Where the currency changes, who executes it, whose money it is at that moment, and when it settles. The same questions as always, asked in a second currency.

The question worth asking your team

Between your customer accepting a rate and you owning that currency, how long is the gap, and who is carrying it?

The good answers are specific. Sub second, passed straight through. Or: up to two minutes, we carry it, it’s in the margin. Or: we net at 4pm and treasury owns the position until then.

The bad answer is a pause, because a pause means the provider’s default is your risk policy and nobody has read it.

Currency conversion looks like arithmetic. It’s a series of decisions about time and about who holds risk. The arithmetic is just where those decisions surface.

One final point. If you’re building on a mature platform, a lot of this is already handled, because somebody has been through the cut-offs, the value dates and the reconciliation before. That doesn’t mean you get to ignore it. It means you can spend your time deciding what your product should do rather than discovering what the market does.

At Integrated Finance, multi currency and FX sit on the same ledger and the same orchestration layer as everything else, because a currency conversion is just another movement of money that has to be recorded properly. The value isn’t that the maths is clever. It’s that the four steps, and the gaps between them, are already someone’s job.