The Backup Nobody Tested: A December Restore Drill

Ask whether the backups work and you will hear that they complete every night. That is the backup job reporting it ran. It is not evidence that anything can be brought back, which is the only thing a backup is for.

The gap between "backups ran" and "we restored" is where recovery plans go to die. December's quiet weeks are the best window all year to close it: one hour, one system, one actual restore.

Key takeaways

• A completed backup job is not evidence that anything can be restored.

• The only proof is a restore: pick one important system and actually bring it back.

• Measure how long it takes; the honest number is the one the plan needs.

• The drill finds the small breakages: the missing credential, the skipped folder, the expired key.

• An hour of drill in December beats learning the answers during an incident.

Step 1: Pick the system the business would miss first

The drill targets one system whose loss would hurt: the file share, the line-of-business application, the accounting data. One system, fully restored, teaches more than a dashboard of green checkmarks for fifty.

Choose with the business, not just IT. The system the business would miss first is sometimes not the one the backup report leads with.

Step 2: Restore it for real, to somewhere safe

The drill restores the latest backup to a test location and then proves it: the application opens, the records are current, the files read. Half-restores and "the files copied" do not count; the test is whether the system actually works from the backup.

Record the time from starting the restore to a working system. That number, not the vendor's marketing, is your real recovery time.

Step 3: Write down what the drill found

Every drill finds something: the service account password nobody current knows, the folder excluded from the backup set two years ago, the documentation that describes a server retired in the spring.

This fictional example shows the record. It is an illustration, not a prescription.

None of these are disasters. All of them were about to be, on the wrong afternoon.

Step 4: Fix the findings while they are cheap

The findings from a drill are the cheapest fixes in the whole recovery program: an exclusion corrected, a credential updated, a step documented. The same findings discovered mid-incident cost the outage's duration plus the fix.

The recovery scorecard is the next level up from this drill, and the first 24 hours after a critical supplier goes quiet covers the day you need the restore for real.

Step 5: Put the next drill on the calendar

One drill is a project; a drill every six months is a control. Schedule the next one before this one's findings are even fixed, and rotate which system gets restored. In two years you will have honest recovery times for everything that matters.

Frequently asked questions

Our IT provider says restores are tested. Is that enough?

Ask what was restored, when, and to where, and whether the working system was verified. Many providers test that backups complete. A drill where you watch a system come back is a different and stronger evidence.

How long should a restore take?

There is no universal right answer; there is your honest number and whether the business can live with it. The drill exists to replace the assumed number with the measured one.

Should the drill include the cloud systems?

Yes, with the provider's own recovery options. Cloud data needs the same question answered: if this account or tenant were lost or locked, what would we restore from, and how long would it take?

Who should run the drill?

Whoever would run the real restore, watched by someone who asks naive questions. The naive questions are where the findings come from.

A Defensible Next Step

Book one December hour: one important system, restored for real, timed and written down. Book a Decision-Clarity Call if you want someone in the room asking the naive questions.