An operator taking dirhams across the counter in Dubai at nine in the morning can promise the family in Manila their pesos by lunchtime. Whether it can keep that promise depends on the balance sheet, because somebody in Manila has to be holding those pesos already, in an account the operator funded weeks before this particular customer walked in.
That is one of two costs sitting underneath every corridor, and it is the one that rarely gets written about. The other is the price of the rail itself: what the correspondent chain charges to carry a payment, deducted long before the operator has quoted anybody a rate.
Neither of these is the fee the end customer sees. That one is the operator's own price, and setting it is the operator's business. These two sit further back, in the cost of goods underneath the route, which is where the margin on a corridor is decided and where a surprising amount of it goes.
Elytra is built for that layer. We sell to money transfer operators, payroll platforms, marketplaces and B2B payment businesses — companies that already hold the licence, the customer relationship and the payout partner, and that are running somebody else's rail underneath all three. What follows is what that rail costs, and what we built to replace it.
Follow the money that never moves
The account has a name older than almost everyone using it. A Nostro — Latin for "ours", as in our money held at your bank — is an account you maintain at an institution in another country, funded in advance, so that payouts in that market can happen at all.
It has to be funded before the first customer of the month arrives, and it has to stay funded through the busiest day, because a pool that empties on a Friday afternoon is a failed payout and a support queue. So it is never sized to average volume. It is sized to peak volume with a margin for error on top, and then it waits.
The correspondent chain multiplies the requirement. A payment does not travel from Dubai to Manila. It travels from a bank in Dubai to a correspondent that bank happens to have a relationship with, then to another correspondent that one has a relationship with, and eventually to a bank in Manila, because no institution holds an account with every other institution in the world. Each link in that chain wants a minimum balance of its own, and each one wants it from you.
The result is cash on your balance sheet that you cannot deploy anywhere else for as long as you want the route to stay open, in every market you serve. Nobody invoices it, so it is not a fee. Nothing is written off, so it is not a loss. It simply sits there, and it grows every time you add a country.
Count the rail as cost of goods
The parked cash is one half of what a route costs you. The other is what the chain charges to carry each payment, and it arrives in two very different forms.
Some of it is visible: the wire charge, the lifting fee each correspondent deducts as it passes the payment on, the SWIFT messaging. Those land on a statement and a finance team can add them up. The larger part arrives as an exchange rate. Every institution that converts currency along the way applies its own, and the distance between that rate and the market's is itemised nowhere, because in law it is a price rather than a fee. You are charged it all the same.
This is why the cost of a corridor is so reliably underestimated by the businesses paying it. You can only total the things that appear on a statement, and the largest line does not.
Most operators who work it out properly find their real cost of goods is two to three times what their internal model shows.
Ask which corridors never get built
So far this is one company paying more than it thought. The larger effect is upstream of that, and it decides which corridors exist at all.
Under the correspondent model, opening a new route means finding several million dollars of fresh pre-funding before a single transfer settles on it. That is not a product decision or an engineering one. It is a capital allocation, argued in front of a finance committee, competing against every other use of the same cash.
So the map of corridors on offer is not a map of where people need to send money. It is a map of where somebody could justify parking several million dollars indefinitely. Routes into large, liquid markets are well served and keenly priced. Routes into everywhere else are thin, expensive or missing, which means the corridor your customers keep asking for is often the one your balance sheet cannot open.
Let the transfer fund its own payout
The pool exists to cover a gap in time. Money leaves the sender today and reaches the destination three or four days later, so something has to stand in for it at the far end in the meantime. Close the gap and the standing pool has no job left to do.
That is what Elytra does. It is a non-custodial orchestration layer that replaces the correspondent chain with a single API call, and every transaction it routes has the same three-part shape regardless of where it starts.
You collect local currency on the licensed infrastructure you already run — an exchange counter, a domestic wallet, a bank transfer, a QR payment. Elytra has no role in this step, and you remain the custodian throughout it.
A licensed partner converts it to USD. The path differs by jurisdiction and is selected automatically, so you do not carry a separate conversion relationship for every market you serve. Elytra routes to the correct licensed entity for the corridor and instructs the movement.
The dollars move domestically inside the United States. From the moment USD is available at the licensed US partner, every corridor behaves identically: Fedwire or FedNow moves the funds into the destination institution's US-held Nostro, and a notification confirms it. Your receiving partner, already liquid in local currency, disburses on domestic rails.
The consequence is in that third step. The dollars never leave the United States, so there is no international wire and therefore no chain of correspondents, which leaves nothing for four separate institutions to each hold your cash against. Settlement runs end to end in under six hours, the domestic leg clears the same business day, and FedNow is instant where the receiving institution supports it.
Both costs from the top of this piece come out together. The correspondents that were pricing the route are no longer in it, and the pool that was covering the delay has no delay left to cover. A new corridor stops being a capital allocation and becomes a routing configuration.
Hand over the rail without handing over the money
None of that would be worth much if changing rails meant putting your licence behind somebody else's balance sheet. Replacing the settlement leg is a regulatory decision before it is a commercial one, because the obligation to your customers stays with you no matter whose infrastructure their money crosses.
So Elytra never holds the money. Not at any step, including the moment in the middle where holding it would be most convenient for us.
Value stays with a licensed institution the whole way. You hold the local currency you collected. A regulated on-ramp partner holds value through the conversion. A licensed US money services business executes the domestic transfer. Your receiving partner holds the dollars until disbursement. Elytra issues the routing instruction and nothing else.
This is worth being precise about, because it is the first thing a risk committee asks and the place a vague answer fails. It is not a policy we adopted and could revise once the volumes justify it: there is no step in the architecture at which taking custody would be possible. We earn on movement and never on float, which is also the only arrangement under which our incentives and yours point the same way.
The rest of what you have built stays where it is. Elytra plugs in at the settlement layer, so the customer relationship, the local licence and the payout partner all remain yours. We replace the leg in the middle that was never yours in the first place.
Run your own number
The exercise worth doing has three inputs: the route pair, the monthly volume, and your current all-in cost as you understand it today.
The third one is where it gets uncomfortable, because the honest version includes the fees you were invoiced, the spread the chain applied at each conversion, and the capital you keep parked at the far end to make the route function. Most operators find that total is materially larger than the figure their model carries.
In every engagement we have run this with, the fee saving is what makes an operator competitive on price, and the capital coming back off the balance sheet is what gets the contract signed.
We would rather work that out with you than quote you somebody else's savings. Tell us the corridor and the volume, and we will model what the route costs you today against what it costs on Elytra. That number is worth having whoever you end up running the corridor with.