Why your backups probably would not save you, and the ten minute test that proves it
26 July 2026 · 6 min read · Optimum IT Solutions
Almost every business we speak to says it has backups. Far fewer have ever restored one. That gap is where most data loss actually happens, and it is cheap to close.
There is a comfortable moment in most owners' heads where backup stops being a worry. Someone set it up, there is a green tick somewhere, and the subject is closed. It is a reasonable place to stop thinking, and it is also the exact point at which the problem starts, because from then on nobody checks.
A backup is not a thing you own. It is a claim you are making about the future: if the worst happens, we can get our data back within a time we can live with. Like any claim, it is either true or it is not, and the only way to find out is to test it. Almost nobody does, and that is why so many businesses discover the truth on the one day they cannot afford to.
What actually goes wrong
When a backup fails to save a business, it is almost never dramatic. It is one of a small number of dull, entirely preventable things, and we see the same ones over and over.
- The job quietly broke. Someone changed a password, a folder was renamed, a drive letter shifted, and the backup started failing months ago. The alerts went to an inbox nobody reads, or to a member of staff who has since left.
- The backup is incomplete. It covers the file server everyone thinks about and misses the database behind the software the business actually runs on, or the mailboxes, or the one laptop where a partner keeps everything.
- The only copy is in the same place as the original. A backup drive plugged into the server it is backing up will be encrypted along with it by ransomware, and stolen along with it in a break in, and burnt along with it in a fire.
- Cloud storage is being mistaken for backup. Syncing is not backup. If a file is deleted or encrypted on the machine, the sync obediently pushes that change everywhere. The bad version becomes the current version.
- Nobody knows how to restore. The data exists, and the one person who understood the process is not answering their phone, so the recovery that should take two hours takes two days of reading documentation under pressure.
The ten minute test
You do not need a project or a consultant to find out where you stand. You need ten minutes and a willingness to be told something you would rather not hear. Do this yourself, or sit with whoever looks after your IT and watch them do it.
- Pick a file at random from a shared folder that matters. Not a test file you created for the purpose. Something real, ideally a few weeks old.
- Ask for it back as it was thirty days ago, restored to a different location so nothing live is overwritten. Time how long it takes from asking to having the file open in front of you.
- Do the same with one email. Pick a message from a mailbox and ask for it to be recovered from a month ago. Microsoft 365 and Google Workspace do not do this on their own past their own retention window, so this question often produces a very informative silence.
- Ask when the last successful backup of your main business system finished. Not the last attempted one. The last one that completed without errors, with the date and time on screen.
- Ask where the second copy lives. If both copies are in the same building or the same cloud account, you have one copy in two places, not two copies.
- Ask the last question plainly: if the main server or the main system died this morning, how many hours until we are trading again, and how many hours of work would we lose?
How to read the result
If the file and the email come back inside ten minutes and the answers are specific, you are in better shape than most. Write the answers down, put a note in the calendar to repeat the test in six months, and get on with your day.
If the answers are vague, that is the finding. Phrases like it should be fine, I think that is covered, or we would have to look into it are not reassurance. They mean nobody has checked, and nobody checking is the condition under which every one of the failures above develops quietly over months.
If nothing comes back at all, you have just had a very cheap piece of bad news. Finding out on a calm Tuesday with the business still running is enormously better than finding out at nine on a Monday with everyone stood around and a customer waiting.
What to fix first, in order
If the test went badly, resist the urge to buy something immediately. The fix is usually sequencing, not spending.
- Get one known good copy off site or into a separate account first. Separation matters more than sophistication.
- Widen the scope to include the things that were missed: mailboxes, the line of business database, and the machines where individuals keep work that never made it to the shared drive.
- Point the failure alerts at a person who still works for you, and agree that a red alert gets acted on rather than filed.
- Restore something real once a quarter. Fifteen minutes, diarised, done by someone in the business rather than only by a supplier.
- Write down the recovery steps in enough detail that a competent person who is not you could follow them. A plan that depends on one individual being reachable is not a plan.
The difference between a backup and a recovery
The ten minute test tells you whether files come back. It does not tell you whether the business comes back, and those are different questions. Getting a spreadsheet restored is not the same as rebuilding the server, the applications, the licences, the connections between systems and the access everyone needs to work.
That larger question is what a proper recovery test answers: rebuild the systems from the backups onto real hardware or a real cloud instance, measure how long it actually took, and measure how much data would have been lost. Until you have measured them, both numbers are guesses, and businesses tend to guess optimistically.
Most firms do not need to do that every quarter. Most firms do need to do it once, so that the numbers they are quietly relying on are real ones. If you would like a hand running either the ten minute version or the full one, ask us. If the test goes well, we will tell you that and leave you alone.
Want a straight answer on this?
Tell us the job that is costing you the most time. We will look at it and tell you honestly what, if anything, is worth doing about it. The first conversation is free.
Get a quote
