Tools for Cybersecurity

Practical guide · November 16, 2026

How to test backups with a documented restore exercise

Choose a business task, restore the data and dependencies it needs into an approved test environment, and have its owner verify that the task works. Record the elapsed time and the age of the recovered data. A successful backup job is useful evidence that a job ran; it does not demonstrate that the business can recover.

New Zealand’s NCSC recommends testing different recovery scenarios, including file, application, database and identity recovery. It also notes that some components need testing together. NCSC backup guidance

Define success before restoring anything

Ask the business owner two questions: How much recent work could we afford to lose? How long could we operate without this service? Those decisions inform the recovery point objective (RPO) and recovery time objective (RTO). Use the objectives to evaluate the exercise, rather than choosing a passing threshold after seeing the result. NCSC recovery objectives

Hypothetical exercise: A service company wants dispatch staff to recover tomorrow’s job list after losing access to its scheduling database.

Item Pre-agreed exercise condition
Business task Open and search the recovered job list, including attached site-access instructions
Acceptable data loss No more than four hours
Acceptable elapsed recovery time Two hours from the exercise’s declared start
Environment Restricted test environment with outbound customer messages disabled
Acceptance owner Dispatch manager
Stop condition Any unexpected connection to production or customer messaging

These are invented targets for the example. A different business may need different objectives or a wider recovery scope.

Run a bounded exercise

  1. Confirm authorization, scope, test destination and the person who can stop the exercise. Keep production recovery procedures intact.
  2. Select and record the backup’s timestamp and identity. Confirm that the team can obtain necessary access and decryption material through the approved process.
  3. Restore using the documented procedure. Time preparation, retrieval, restoration and validation; do not start the clock only after difficult preparation is finished.
  4. Have the business owner perform the agreed task. Check representative records, relationships and permissions, not only whether a file opens.
  5. Record failures, cleanup and follow-up work. Remove or retain the test copy according to the organization’s data-handling rules.

Use an environment that prevents restored jobs, integrations or notifications from acting on real customers. If you cannot isolate a realistic restore safely, resolve the test design before proceeding.

A completed record can show a partial failure

Hypothetical result: The exercise starts at 10 a.m. using an 8 a.m. backup. The database becomes accessible and the validation attempt ends at 11:35 a.m.—95 minutes after the declared start. The selected backup is two hours old at the incident boundary, within the four-hour data-age criterion; representative record checks must still establish its contents. The 95-minute technical milestone does not demonstrate the two-hour business recovery objective.

However, the restored application can search jobs but cannot display attached access instructions. The acceptance owner records failed business validation because those attachments are part of the agreed dispatch task. Business recovery remains incomplete, so RTO is not demonstrated. Record the time the complete task becomes usable in a later exercise; do not stop the recovery clock at database availability.

The follow-up is specific: identify where attachments are stored, include their recovery dependencies, update the procedure and repeat the failed portion. Preserve the original result rather than replacing it with a success label.

Decide what the test did not establish

A file restore does not establish full application recovery. A successful application exercise does not necessarily demonstrate recovery after losing the identity system, the backup administrator’s device or an entire location.

Plan the next scenario around the most consequential untested dependency. Set a repeat schedule based on your recovery objectives and changes to the system; this worksheet does not prescribe a universal monthly or quarterly requirement.

Research basis: NCSC guidance reviewed November 16, 2026. The procedure and completed record are hypothetical editorial examples, not results from testing a backup product.