Do you have backups? is the wrong question. Almost everyone says yes. The useful question is: how long would it take you to be running again, and how much would you have lost?
Two numbers worth knowing
Every recovery plan comes down to two figures, and most businesses have never put numbers on either.
How much data can you afford to lose? If your last backup was at 2am and the server dies at 4pm, you’ve lost a day of work. For some businesses that’s an inconvenience. For others it’s a catastrophe. The answer determines how often backups need to run.
How long can you be down? Restoring a full server from a cloud backup over a normal business internet connection can take a day or more. If you can’t be down that long, you need a different approach — a local copy that restores fast, or the ability to run temporarily from the backup itself.
Neither number is right or wrong. But if you haven’t chosen them deliberately, they’ve been chosen for you by whatever was set up years ago.
Where backups quietly fail
- Nobody is watching. Backup software reports failures. If those reports go to an inbox nobody reads, a job can fail for months.
- The scope drifted. Backups configured for a folder that’s no longer where the data lives.
- It’s on the same site. A backup drive next to the server protects you from disk failure and nothing else — not fire, not flood, not theft, not ransomware.
- Ransomware can reach it. If your backup is a mapped drive, an infection encrypts it along with everything else. Backups need to be isolated from the systems they protect.
- Cloud services aren’t backed up. Microsoft 365 and Google Workspace replicate your data — they don’t protect you from someone deleting it. Retention windows are shorter than most people assume.
The test that settles it
Pick a file from three months ago. Ask whoever manages your IT to restore it. Time how long it takes and note whether they can do it at all.
That single exercise tells you more than any amount of documentation. And if the answer is uncomfortable, better to find out on a Tuesday morning than during an actual incident.
What good looks like
Backups running on a schedule that matches what you can afford to lose. At least one copy off-site and out of reach of your production systems. Monitoring that alerts a human when a job fails. And a restore tested on a real schedule — not when we get a chance.