The one afternoon a year that turns a hope into a fact.

01The difference between a backup and a fact is a restore

You have set up a backup. Files copy nightly; the log shows green; the drive or cloud bucket fills with data. You feel, reasonably, that you are covered. You are not. What you have is a hypothesis: if something goes wrong, this will probably work. The word "probably" is doing a lot of quiet work in that sentence.

Backups fail in ways that never announce themselves. The destination ran out of space six weeks ago and the job has been silently overwriting the same partial snapshot. The cloud client lost authentication and has been logging "success" against a local cache it never uploaded. The archive is intact but encrypted under a key you rotated and no longer have. The external drive was bumped, the filesystem is corrupt, and every file inside it is unreadable. None of these failures show up in a status light. They show up the first time you try to restore — which is either a planned test or a crisis.

There is only one way to turn a hypothesis into a fact: pick a file, pick a folder, and actually restore it from the backup to a different location. Not a preview. Not a file listing. A real restore, opened, verified, compared against the original. If the backup covers a system volume, go further: restore to a spare machine or a virtual machine and confirm it boots. Thirty minutes of boredom now is several hours of blind panic later.

Do this on a schedule you write down. Once a year is the floor, not the ceiling. Twice a year is better. A small business holding client data or anything legally significant should test quarterly. Every time you change the backup software, the destination, the retention policy, or the machine being protected, test again immediately — a configuration change is exactly the moment a silent failure slips in.

While you are there, check the oldest files the backup holds. Retention settings have a way of being shorter than assumed, and knowing how far back you can actually go is the real question your backup needs to answer. Check the log for errors you scrolled past. Verify the destination still has the space the next six months of growth will need.

None of this is technical. It takes an afternoon, one calendar reminder, and the willingness to find out rather than hope. The people who lose irreplaceable data are not only the people who had no backup. They are often the people who had a backup they never tested.

A green status light means the job ran. Only a restore tells you whether it worked.

Thirty minutes of boredom now is several hours of blind panic later.

  1. Pick something realA folder you would actually miss, not a test file you created for the drill.
  2. Restore from the oldest copyThe off-site or offline one — the copy you have never used.
  3. Open the filesNot a directory listing. Open them, in the applications that made them.
  4. Write down what went wrongA missing key, an expired credential, a forgotten exclusion — that is the finding.
  5. Once a yearThat is the whole discipline. An afternoon turns a hypothesis into a fact.
1real folder is enough to test a restore properly
0errors is the target, and the only number worth recording