There is a genre of fintech writing that treats local payment rails as the future and correspondent banking as a fax machine that refuses to die. It makes for confident copy and it is not quite true. The two do different jobs, and knowing which job you have saves a lot of wasted integration.
What each one is good at
Correspondent banking moves large amounts between institutions, in any major currency, to virtually anywhere, with a documentary trail that auditors and regulators recognise on sight. For a seven-figure payment to a supplier, this is not a compromise — it is the correct instrument.
Local rails move small amounts to individuals, fast, inside one country, in that country’s currency. For paying two thousand people a hundred dollars each, this is the correct instrument, and correspondent banking would be an expensive mistake.
The confusion arises because the second use case grew enormously in the last decade while the first stayed roughly the same size.
The honest comparison
On speed: local rails settle in seconds; correspondent chains take days, and the variance is worse than the average.
On cost per payment: local rails are near-free domestically; correspondent hops each take a cut, which is irrelevant on large amounts and fatal on small ones.
On reach: correspondent banking goes almost anywhere; local rails go exactly one country each, and only if someone has built the connection.
On amount ceiling: correspondent banking has effectively none; local rails have practical limits, because they were designed for domestic retail.
On predictability of failure: local rails fail fast and clearly — a bad account number is refused in seconds. Correspondent chains fail slowly and vaguely, sometimes days later, sometimes with a fee already deducted.
How to tell which you need
Three questions usually settle it.
Who is the recipient? A business with a treasury function and an IBAN, or an individual with a phone? Institutions are comfortable on correspondent rails. Individuals are not — they want the money in the account their salary arrives in, addressed the way their country addresses accounts.
How many payments, how small? Fixed costs per payment decide this. A ten-dollar fee is invisible on fifty thousand dollars and ruinous on fifty.
How much does timing matter? Not the average — the worst case. If a delayed payment means an angry driver, a missed shift or a support ticket, days of variance is not a cost you can absorb.
Most real businesses need both
The company paying suppliers in three countries and contractors in twenty does not choose. It uses correspondent banking for the suppliers and local rails for the contractors, because those are different problems that happen to share a finance team.
The mistake worth avoiding is forcing one tool to do the other’s job — either grinding thousands of micro-payments through a correspondent chain, or trying to move a seven-figure supplier payment through a retail instant system that was never built for it.
Pick per flow, not per company.