Last Monday was a bank holiday in the United States. The Friday before, in the morning, a SWIFT had been sent to us from Switzerland: not at the cut-off, not in a rush, with hours to spare in the business day. The weekend went by, Monday’s holiday came, and the correspondent bank didn’t process the transaction either. The transfer landed on Tuesday at 3 in the afternoon, when the FX market in Colombia had already closed for the day. I had to wait for it to open on Wednesday, and on Wednesday it opened lower, near 3,100, well below the 3,150 of the previous Friday.
Our counterparty had nothing to do with it, and the transfer wasn’t sent late either. That is simply how the system works: a holiday in a country with no involvement in the transaction freezes two things at once, when the money arrives and at what rate it converts, no matter what time of the business day it was sent. In almost every blog in this industry the argument is the same: correspondent banking is slow and stablecoins are the future. That’s true, but it stopped being interesting a while ago.
What actually weighs is not how long the payment takes, but what happens to the company while it waits. It can end up carrying an FX position that nobody explicitly took on and that nobody is measuring. And in 2026 that happens to any treasury relying solely on correspondent banking for its dollar payments into or out of Colombia, more often than anyone would care to admit.
Let’s do the math
In 2026, the US Federal Reserve observes 11 bank holidays. Colombia has 18, set by Law 51 of 1983, known as the Emiliani Law. Of those 11 US holidays, only three overlap with a Colombian one: New Year’s Day, Columbus Day and Christmas. Those are no problem, because both countries are already closed. The Fourth of July doesn’t count either: in 2026 it falls on a Saturday and, on top of that, the Fed keeps its operating banks open the Friday before.
That leaves seven: Martin Luther King Jr. Day, Washington’s Birthday, Memorial Day, Juneteenth, Labor Day, Veterans Day and Thanksgiving. On those seven days Colombia usually operates normally, while the dollar leg of the payment can freeze simply because it is a holiday in Washington.
And there’s one more detail that makes it worse: five of those seven fall on a Monday or a Friday, right up against a weekend. When that happens, the freeze doesn’t last a day, it lasts three. That is exactly what just happened to me. SWIFT sent on Friday, a weekend in between, a holiday on Monday: four full days between sending and settling. Adding the five three-day holidays and the other two, worth one day each, the total comes to 17 calendar days a year on which the dollar leg of a payment into or out of Colombia can sit frozen purely because of the US calendar. Any treasury can run the same calculation for its own corridor.
What a holiday costs in a local-currency corridor
In a dollarized corridor, such as El Salvador or Panama, a holiday costs only time: the money arrives late, but it arrives in the same currency it left in. In a local-currency corridor, such as Colombia or Bolivia, it costs something more.
And I don’t need a hypothetical example, because I have just lived through one. The transfer landed on Tuesday afternoon, with the FX market already closed, so the rate that mattered was not Tuesday’s but Wednesday’s, when the market reopened. On the USD 150,000 sent to me from Switzerland, the gap between Friday’s official rate (near 3,150) and Wednesday morning’s (near 3,100) is more than COP 7 million. Nobody decided to put that money at risk and nobody hedged it, but there it sat, exactly as it sits on the balance sheet of any company that depends on correspondent banking to settle in pesos.
That uncertainty shows up in no risk report, because nobody decided to take it on: it was imposed by a holiday in a country that isn’t even part of the transaction.
This isn’t a technology problem, it’s risk governance
It’s worth asking how much of a company’s FX exposure is created without treasury ever consciously deciding to take it. If a treasurer walked into their risk committee with an open FX position, no defined size, no closing date, and explained that they took it because it was a holiday in Washington, nobody would approve it. And that is precisely what can happen, quietly, up to seven times a year, and up to 17 calendar days in total, at any company that relies solely on correspondent banking to pay in dollars into or out of Colombia.
Nobody calls it that because nobody measures it. In this industry we say “the payment is in transit”, as if it were merely a matter of time, when in reality there is unmanaged FX exposure. If it appeared on the balance sheet under its real name, somebody would have to explain it to the audit committee. As long as it keeps showing up as “payment in transit”, who is going to dare ask?
Closing that gap isn’t a matter of solving the two problems separately, but of solving them at the same time: settling without depending on a third country’s banking calendar, and fixing the conversion rate at the moment the payment is executed, rather than leaving it to the always uncertain moment when the correspondent decides to release the funds.
Three questions for your next payment “in transit”
The next time one of your payments sits “in transit”, ask your team three questions: how many of those days are down to the calendar of a country unrelated to the transaction, how many of them are attached to a weekend, and at what rate it will settle when it finally arrives. If nobody can answer all three precisely, or if nobody even has a way to calculate them, you have just found your treasury’s first blind spot.
Gabriel Ramírez is CFO of efy Technologies. He is writing this because last Friday a SWIFT was sent to him from Switzerland that arrived four days later, with the FX market already closed, so he had to convert it the following day at a worse rate than he expected.