Founding customer pilot now open in Addis Ababa·Prices in ETB, same at renewal·Join the pilot →

Disaster recovery for banks: what a plan that works looks like

A recovery plan is only real once it has been rehearsed. Here is how banks can set targets, design a second site and prove it works.

HuluBuild Cloud team3 min read
A technician working on server racks in a data center

Every bank has a recovery plan on paper. The harder question is what happens at 10 a.m. on a salary day when the primary site goes dark. Customers do not care which component failed. They care whether the mobile app, the ATM and the branch teller still work.

A plan that works begins with two numbers, a second site that is genuinely independent, and a habit of proving the whole thing on a schedule.

Start with two numbers

Recovery time objective (RTO) is how long a service can be down. Recovery point objective (RPO) is how much recent data you can afford to lose. Set them per service, because they drive almost every cost that follows.

ServiceTypical priorityWhat drives the target
Core banking and ledgerHighestCustomer balances must be exact
Mobile and internet bankingHighestThe channel customers notice first
ATM and card switchingHighCash access and merchant payments
Reporting and analyticsLowerCan usually wait for a rebuild
If nobody can tell you the last date recovery was tested, you do not have a recovery plan. You have a document.

Make the second site truly independent

A recovery site in the same building, on the same power feed or behind the same carrier protects you from very little. Independence is what you are paying for.

  • Separate location, far enough to avoid the same local incident
  • Separate power and a different network carrier where possible
  • Encrypted replication matched to your RPO
  • Enough capacity to run critical services, not just store copies
  • Keys and credentials that remain available if the first site is lost

Rehearse the failure

Recovery drills expose what diagrams hide: a certificate that exists only at the primary site, a firewall rule nobody documented, a batch job that assumes a hostname. Run drills on a schedule, record the actual recovery time, and fix what you find before the next one.

  1. Agree the scope and rollback steps with risk and operations
  2. Fail over a low-risk service first, then build toward core systems
  3. Measure actual recovery time and data loss against your targets
  4. Write down what broke and assign an owner to every fix
  5. Share dated results with auditors and the board

Quick answers

What is the difference between RTO and RPO?+

RTO is the maximum time a service can be unavailable. RPO is the maximum amount of recent data, measured in time, that can be lost. Each service should have its own targets.

How often should a bank test disaster recovery?+

On a regular schedule agreed with risk and compliance teams and any regulator, and after major system changes. Every test should produce dated, written results.

Share this article
Ready for the next step?

Talk to our team, or explore how HuluBuild Cloud can help.