Your Backup Probably Doesn’t Work

"Yes, we back up." I hear it on nearly every first visit. Then I ask when someone last restored a file from that backup, and the room goes quiet.

An untested backup is a guess. Here's how to turn it into something you can rely on.

The test you can run this week

Pick a file that matters, a client folder, last month's accounts, a project directory. Now restore it from your backup to a different location and open it.

That's it. If you can't complete that in under an hour, you don't have a working backup, you have backup software. The failures this exercise uncovers are consistently mundane: the job stopped running four months ago and the alert email went to a former employee; the external drive filled up; the sync only ever covered one user's Documents folder; the file restores but is corrupt; nobody knows the password to the backup account.

The 3-2-1 rule

The standard worth aiming at, and it's simpler than it sounds:

• 3 copies of your data, the live one plus two backups.
• 2 different types of storage, for example a local drive and cloud storage.
• 1 copy off-site, physically elsewhere, or in the cloud.

Why off-site matters in practice: fire, flood, theft and lightning damage don't discriminate between your server and the backup drive sitting next to it. In South Africa specifically, add power surges after load-shedding. I have seen a surge take out a server and the USB drive plugged into it in the same instant.

Ransomware changes the requirements

This is where most small-business backup strategies fall down. Modern ransomware doesn't only encrypt the machine it lands on. It deliberately looks for network drives, mapped shares and connected cloud sync folders, and encrypts those too.

So if your "backup" is a network drive that's always mounted, or a cloud folder that syncs continuously, it is inside the blast radius. The encrypted files sync up, overwrite the good ones, and your backup is now a backup of the ransomware's work.

What survives ransomware:

• Versioned backups where you can roll back to a point before the infection.
• Immutable or write-once storage that cannot be altered after the fact, even with admin credentials.
• Offline copies. A drive that is physically disconnected between backups. Unfashionable, and it works.

Decide what you can actually afford to lose

Two questions decide your whole design:

How much data can you lose? If backups run nightly and the failure happens at 16:00, you lose a day's work. For an accounting practice at year-end, that's serious. For a business whose files rarely change, it's nothing.

How long can you be down? Restoring 500 GB over a typical South African upload link takes considerably longer than most people assume. If being down for two days would be catastrophic, you need a local copy for speed as well as a cloud copy for safety.

Answer those honestly and the right setup becomes obvious, and usually cheaper than the over-engineered thing you feared.

What I usually recommend for a small SA business

1. Business cloud storage with version history for working documents, so a corrupted or encrypted file can be rolled back per-file.
2. An automated versioned backup of the machines that matter, to storage the machines themselves can't reach and overwrite.
3. Monitoring that reports failures to a person who still works there, checked at least monthly.
4. A restore test every quarter. Diarise it. Fifteen minutes, four times a year.
5. Documentation of what is backed up, where it goes, how to restore it, and who has the credentials, not held only in one person's head.

The uncomfortable point

Antivirus, firewalls and staff training all reduce the chance of a disaster. Only backups determine whether a disaster ends your business or costs you an afternoon. It's the least interesting item on any IT budget and the only one that decides whether you survive the bad day.

If you'd like your current setup checked properly, including an actual restore test rather than a look at a dashboard, that's part of IT support, and I'd rather find the problem now than during the incident.