A Backup You've Never Restored Is Just a Subscription: Running a 30-Minute Restore Drill

The green tick problem

Somewhere in your organisation there is a dashboard with a reassuring green tick on it. It says the backup ran at 02:00. It says zero errors. Everyone sleeps a little easier.

Then someone deletes a folder of tender documents the day before a deadline, and the conversation that follows is not about the backup at all. It is about who can click restore, which version to pick, how long it takes, and whether the last four hours of work are gone forever.

Backups are not a product you buy. They're a capability you own, and the only way to know you own it is to use it on purpose, on a quiet day, before you need it.

Start with the files that would ruin a week

Don't try to test everything. Pick five to ten items that would genuinely hurt:

• The shared finance folder in SharePoint.
• The customer database export or the accounts package data file.
• The managing director's OneDrive, which quietly holds the only copy of the board pack.
• The mailbox of whoever handles orders.
• A line-of-business system on the local server — the folder the software writes to, not the software itself.

Write the list down and pin it to the drill. The same items get tested every quarter so you can compare results over time.

Who can actually press restore?

This is where most small firms get a surprise. In Microsoft 365, recoverability is layered:

• Version history in SharePoint and OneDrive covers individual file edits, but only within the limits your tenant has configured.
• The recycle bin retains deleted items for a set period before they're purged — often 93 days, but check what yours actually says.
• Retention policies may keep more, or may be doing nothing at all if nobody set them up.
• Tenant-level recovery for mailboxes and sites depends on roles. A Global Administrator can do things a normal user cannot, and if that admin is one person on holiday, your recovery time just grew by a week.

So the drill asks a blunt question: name two people who can restore this file, right now, without reading a guide. If you can only name one, that's a finding.

Plain English RTO and RPO

You'll see these acronyms in every backup brochure. Ignore the jargon and translate them:

• Recovery point — how much data can you afford to lose? If backups run nightly, a file deleted at 4pm might come back missing a day's edits. Is that acceptable for the finance folder? For the shared drive? Be specific per item.
• Recovery time — how long until the file is back in someone's hands? Ten minutes for a single document is very different from a day to rebuild a server.

Write both numbers next to each item on your list. Rough figures are fine. The act of writing them turns vague confidence into a plan somebody can be held to.

The drill itself: thirty minutes, once a quarter

Book it like a meeting. Warn the person who owns the affected data. Then work through this:

1. Pick one item from the list. Single file from SharePoint is a good starting point.
2. Delete it deliberately — or use a test copy — and start a timer.
3. Have the nominated second person do the restore. Not the IT contact. Someone who would realistically need to do it.
4. Restore to an alternate location, not over the original. This avoids making a test into an incident.
5. Note the time it took, the version restored, and whether any content was missing.
6. Repeat the timing against your stated recovery target. Gap? Write it down.
7. Check the backup tool's own logs for the last 30 days. Look for skipped jobs and silent failures, not just red errors.
8. Confirm the offline copy exists. Ransomware hunts connected backups. If every copy is reachable from a logged-in desktop, you have one copy, not three.

For a local server, add a step: restore a single folder to a spare drive or a virtual machine and open two or three files to prove they're readable. A backup file that restores but won't open is a very expensive way to store nothing.

What the first drill usually exposes

Expect at least one of these:

• The person who 'does the backups' has never restored anything.
• The retention setting is shorter than the firm assumed.
• Nobody has admin rights on the right account, or the admin account is shared with three people and audited by none.
• The backup covers the mailbox but not the shared calendar someone relies on daily.
• The server backup runs but skips the folder the line-of-business app actually uses.

None of these are disasters. They're just cheaper to find now.

Put the next one in the diary

Thirty minutes a quarter, four items a year, one page of notes. That's the whole habit. It costs less than a team lunch and it turns a promise on a dashboard into something your business can actually stand behind.

If you take nothing else away: a backup you've never restored isn't insurance. It's a subscription with hopeful branding.