Restore Verification
Periodically performing a real restore from backup and validating the result, as the only evidence that a recovery capability exists.
Backup success metrics measure the backup job, not recoverability. Every recovery failure of consequence has featured green backup dashboards. The list of organisations that discovered this during an actual incident is long and includes companies that did not survive it.
What a real verification does: restore to an isolated environment on a schedule, from the retention tier you would actually use (a cold archive restores far more slowly than a recent snapshot), then validate the data — row counts, checksums, referential integrity, and an application-level check that the restored system serves correct results.
Measure the elapsed time and compare it to the stated RTO. The discovered figure is reliably larger than the documented one, and the gap is the finding.
The specific gaps this surfaces: encryption keys not backed up or not accessible from the recovery environment; schema or version mismatch between the backup and the current application; missing dependencies such as configuration, secrets and object storage; restore procedures requiring credentials only one person holds; and cloud API quotas that throttle a large restore.
Automate the verification, alert on failure, and treat an unverified backup as no backup — because operationally that is what it is.