A disaster recovery plan can look reassuring in a folder and still fail at the moment your business needs it. The common problem is not a lack of backups. It is discovering, during an outage, that a key system cannot be restored quickly, the right people cannot access their accounts, or nobody is sure who is authorised to make decisions.
Knowing how to test disaster recovery turns an assumption into evidence. For a small or medium-sized business, that evidence protects more than data. It helps protect customer service, staff productivity, cashflow, reputation and the ability to keep communicating when normal systems are unavailable.
Start with the business impact, not the technology
A useful test begins by identifying what must keep running, what can wait and what cannot be recreated if it is lost. Avoid treating every application, document and device as equally urgent. That approach creates an expensive, vague plan that is difficult to test properly.
Speak to the people responsible for operations, finance, customer service and IT. Ask practical questions: if email was unavailable for a day, what work stops? If the main line or cloud phone system failed, how would customers reach the team? If staff could not access Microsoft 365 or a line-of-business application, could they work from another location or device?
The answers establish priorities. A business may need customer records and phones back within four hours, while archived files can wait for two days. A school or healthcare setting may have stricter requirements around critical systems, records and safeguarding. The right recovery approach depends on the organisation, its obligations and the cost of downtime.
Set recovery targets people can measure
Two measures give a test direction. The recovery time objective, or RTO, is how long a service can be unavailable before the impact becomes unacceptable. The recovery point objective, or RPO, is the maximum amount of data the business can afford to lose.
For example, an RTO of four hours means the service should be available again within four hours of an incident. An RPO of one hour means restored data should be no more than an hour old. These are business decisions, not just technical settings. A system backed up nightly cannot meet a one-hour RPO, regardless of how well the restoration process is documented.
Choose a realistic disaster recovery test scenario
A full recovery exercise is valuable, but it is not the only way to test. Smaller, targeted tests can expose weaknesses without unnecessarily disrupting the working day. The key is to test a situation that could genuinely affect your organisation, rather than running a demonstration designed to succeed.
Common scenarios include a ransomware incident, accidental deletion of important files, loss of a server, broadband failure, a prolonged power outage, failure at a cloud provider, or an office that staff cannot access. For organisations with several locations, consider the loss of a single site and whether staff can continue working elsewhere.
Start with the scenario most likely to cause meaningful operational disruption. For many businesses, this is not a dramatic building-wide incident. It may be a compromised user account, failed internet connection or unavailable cloud application. Test the recovery process around that event, including the communications and decision-making that go with it.
How to test disaster recovery in stages
Testing should build confidence gradually. A plan that has never been reviewed should not begin with a high-risk live failover. Begin by checking the plan on paper, then progress to controlled technical and operational tests.
Review the recovery plan with the people who would use it
A tabletop exercise is a structured discussion of an incident. Give participants a scenario and work through what would happen in the first hour, the first day and the return to normal operations. Include management, IT support, department leads and anyone responsible for communicating with customers, suppliers or staff.
This quickly highlights gaps. Are emergency contacts current? Can a senior manager access the plan if normal systems are unavailable? Does the plan identify the correct suppliers, account details and escalation routes? Who decides whether to isolate devices, notify customers or invoke alternative working arrangements?
Do not assume that a document held in a shared drive is available during every type of incident. Keep a secure, accessible copy of critical recovery information outside the systems it relates to. It should include contacts, account recovery procedures, asset information, network details, licences and a clear order of action.
Test backup restoration, not only backup completion
A successful backup report does not prove that your data can be recovered. Backups can be incomplete, corrupted, encrypted by an attacker or too slow to restore within the required timeframe.
Select representative data and restore it to a controlled location. Check that files open correctly, permissions are appropriate and the restored information is complete. If you use virtual servers, test whether a server can be recovered and whether its applications start properly. If your critical system relies on a database, confirm that the database is usable, not merely present.
Record how long each stage takes. The clock should begin when the issue is recognised, not when an engineer starts the technical restoration. Time is often lost through diagnosis, approvals, locating credentials and deciding which systems to recover first.
Test people, access and communications
Recovery frequently fails at the human level. Staff may need multi-factor authentication but have lost their usual phone. Remote workers may lack the permissions, devices or connectivity to use an alternative system. A recovery inbox may exist, but customers may not know where to send urgent requests.
Test whether named staff can access essential systems from a safe alternative device or location. Check administrator accounts, password recovery processes and break-glass access, which should be tightly controlled and monitored. Confirm that key contacts can be reached outside normal working channels.
Also test the communications plan. Draft the messages you would send to staff, customers and suppliers, then agree who approves them. Clear, timely information reduces confusion and limits the number of people trying to resolve the same issue independently.
Test failover where it is justified
Failover testing moves a service to an alternative environment, connection or site. It can provide strong evidence that recovery arrangements work, but it needs planning. A poorly timed live failover could create avoidable disruption, especially where legacy applications, specialist equipment or third-party integrations are involved.
For critical services, agree a maintenance window, success criteria and a rollback plan. Test the service from the user’s perspective: can staff sign in, make calls, access files, process orders and use the information they need? Technical availability alone is not enough if the service performs too slowly or cannot support normal workflows.
A connectivity test might involve switching to a secondary broadband connection or mobile backup. A telephony test could check whether calls divert to mobiles or an alternative number when the primary service is unavailable. These smaller tests are particularly useful for businesses that depend on customer contact throughout the day.
Define success before the test begins
A disaster recovery test should produce evidence, not just a general feeling that things went well. Before starting, set clear pass criteria based on your RTO, RPO and operational requirements.
For example, the test may pass only if the finance system is restored within four hours, data is no older than one hour, authorised users can sign in, and the team can process a sample transaction. A communications test may pass when incoming calls reach an agreed alternative destination and staff know how to handle them.
Assign one person to record timings, decisions, issues and workarounds. Workarounds matter because they can reveal hidden dependencies. If recovery only succeeded because a particular employee knew an undocumented password or had a spare device at home, the process is not dependable yet.
Treat findings as improvement work, not a pass or fail exercise
Every test should end with a short review while the details are still fresh. Document what worked, what failed, what took longer than expected and what was unclear. Then assign each action to an owner with a realistic deadline.
Typical improvements include changing backup frequency, documenting a forgotten dependency, updating contact lists, providing an alternative internet connection or clarifying responsibilities during an incident. Some findings will require investment, while others are simple process changes. Prioritise them according to the risk they reduce and the business impact they address.
Retest the areas that failed. A plan is only improved when the change has been proven, not when an action is marked complete in a spreadsheet.
Build a sensible testing schedule
The right frequency depends on how quickly your systems, staff and risks change. Review the recovery plan after a major technology change, office move, new cloud application, merger, supplier change or cyber incident. These events can invalidate assumptions that were sound only a few months earlier.
As a practical baseline, review contacts and procedures quarterly, test backup restoration at least quarterly, and run a broader recovery exercise annually. Higher-risk or more regulated environments may need more frequent, formal testing. Smaller businesses can make progress by testing one critical service at a time rather than attempting everything in one day.
Technology should make recovery easier, but it does not remove the need for preparation. A carefully planned test shows whether your people, processes, connectivity and systems can work together under pressure. With experienced support and clear ownership, disaster recovery becomes a manageable part of keeping the business dependable – not an unpleasant surprise saved for the day something goes wrong.