Yusr completes multi-region resilience programme

RIYADH, 4 March 2026 — Yusr Bank today announced the completion of an eighteen-month programme to rebuild how its banking platform survives failure. The work began in September 2024, before the app opened to the public, and finished with a full rehearsal in February in which the entire customer-facing platform was moved to the bank's secondary region and back again while customers were using it.

Yusr's primary systems run in-Kingdom, in the cloud provider's Riyadh region. That has been true since day one and has not changed. What the programme added is a properly engineered second place to run from: a secondary region in Frankfurt, Germany, held for continuity and disaster recovery, kept continuously in sync with Riyadh rather than held cold and reconstructed from backups after the fact.

The distinction matters more than it sounds. A cold standby is a promise about a rebuild — someone restores last night's backup, someone else checks it, and the clock runs the whole time. A warm, continuously replicated region is a switch. Yusr's databases stream changes to Frankfurt as they are committed, the same application versions run in both places, and the configuration that ties them together is generated from one source rather than maintained twice by hand.

Riyadh runs it. Frankfurt is ready for it.

Under normal conditions every customer request is served from Riyadh. Frankfurt sits behind the same platform, receiving the replication stream, running the same builds and answering the same health checks, but taking no customer traffic. It exists so that a regional outage — a network partition, a power event, a failure inside the provider's own infrastructure — becomes an inconvenience measured in minutes instead of an outage measured in a working day.

The programme also tightened what "recovered" means. Not every service has to come back at the same speed, and pretending otherwise produces plans nobody can execute under pressure. Yusr now publishes internal objectives per service tier, and tests against them.

Objectives are targets for planned and unplanned failover, measured from the point of declaration. Card authorisation on the mada network continues to be processed domestically in both normal running and failover.

Rehearsed, not assumed

February's exercise was the first end-to-end run. Traffic was moved region to region during a Tuesday evening window, held there while the team worked through a scripted set of checks — a live card payment, a Yusr Send transfer, a new account application, an in-app chat session, a statement download — and then moved back. Total elapsed time from declaration to full service in the secondary region was eleven minutes. Customers were notified in advance through the status page and the exercise was logged there.

Three findings came out of it, all of them small and all of them now closed: a monitoring alert that pointed at the wrong dashboard, a batch job that needed a manual nudge to resume after the switch, and a runbook step written clearly enough for the person who wrote it and nobody else. Yusr's view is that the value of an exercise is precisely the list of things that were not quite right. A rehearsal that finds nothing was not a rehearsal.

A regulated bank does not get to say it will probably be fine. You either know how long it takes to come back, because you have done it, or you are guessing. We can now move the whole platform between regions in under fifteen minutes, and we do it four times a year in front of an audience that is looking for problems. That is the standard the job deserves.

— Omar Al-Zahrani, Chief Technology Officer, Yusr Bank

What changes for customers

Nothing, on a good day — which is the entire point. There is no setting to change and nothing to install. On a bad day, the app should keep working while an engineer somewhere is having a considerably worse afternoon than the customer is.

Yusr publishes real-time availability for every customer-facing service, along with the history of past incidents and scheduled exercises, on its system status page. More detail on how the platform is built is on the Yusr engineering blog.

— ENDS —