A backup that has never been restored is a hope, not a safeguard. Five steps help non-technical founders test recovery on a spare copy, measure the result and leave a runbook behind.
Open your hosting dashboard and you will probably find a green tick beside the word "Backups". It is reassuring, and it answers a smaller question than the one that matters. The tick says a copy was made. It does not say that anyone has ever turned that copy back into a working system.
A backup restore test closes that gap. Someone takes a recent backup, rebuilds the system from it on a spare copy, times the job and checks that real work can be done in the result. Teams that run one often find missing files, an out-of-date copy or steps that only one person remembers.
Founders skip it for understandable reasons: the system works today, the person who set it up seems reliable, and the test sounds technical. It is closer to a fire drill than an engineering project, and most of it is deciding what 'recovered' means for your business.
Five steps follow, from listing what must be protected to writing the runbook. First, it helps to be clear about what a restore test is and how it differs from the backup itself.
What is a backup restore test, and why do untested backups fail?
A backup restore test is a planned exercise in which a recent backup is used to rebuild the system on a spare copy, then checked for completeness and timed. Untested backups fail because jobs stop silently, copies miss files or settings, access keys are lost, and the restore takes far longer than anyone assumed.
Three terms get mixed up in most conversations, and each answers a different question.
| Term | What it is | Question it answers | Typical gap |
|---|---|---|---|
| Backup | A stored copy of data and settings | Do we have a copy? | The copy sits on the same server, or is incomplete |
| Restore test | A rehearsal of rebuilding from that copy | Can we recover, and how long does it take? | Never done, or done once long ago |
| Recovery plan | A written guide for a real failure | Who does what, and in what order? | No owner, no contact list, steps in one person's head |
Most teams have the first, few have the second and fewer still have the third. The steps below build the second and third from the first.
For a SaaS product, backup is one of the early technical choices covered in ten architecture decisions to make before you build. This article covers the part most teams skip, which is actually restoring.
An illustrative business
Picture a composite business, invented for this article: a regional catering company whose order and invoicing system was built by a former contractor. A nightly backup runs, and the hosting dashboard shows green ticks. One morning the server disk fails.
The newest usable backup might turn out to be weeks old, because the job stopped after a password change and nobody saw the error. The five steps below are designed to find that kind of problem on a quiet Tuesday, not on the day the disk fails.
1. What exactly are you trying to get back?
Before testing a restore, list everything the system needs to run again: the database, uploaded files, settings, secret keys, the code itself and any data held by third-party services. A backup that holds only the database may leave the application unable to start. Define 'recovered' as a working system, not a restored file.
What to do
- Database: customers, orders, invoices and history.
- Uploaded files: documents, images and attachments.
- Settings and secrets: environment settings, API keys, certificates and encryption keys.
- Code and build steps: the repository and the instructions to deploy it.
- Third-party data: payment, email, SMS and CRM records held elsewhere.
- Access: logins for hosting, the domain and the backup storage.
What a pass looks like
A one-page list in which every item shows where it is stored, how often it is copied and who can reach it.
Watch out for
Uploaded files are often stored apart from the database, and secret keys sometimes exist only on the live server. Third-party services are another blind spot, because teams assume the provider 'just has it'. Ask the person who manages the system to confirm each item in writing.
Not sure your backups would survive a bad day?
If nobody has ever restored from them, the first job is finding out what they actually contain.
Review your backup setup2. How much loss and downtime can the business tolerate?
Two limits turn vague worry into a testable target. The first is how much recent data you could lose, such as the last hour or the last day of orders. The second is how long the business can run without the system. Technical teams call these the recovery point objective and the recovery time objective.
What to do
- Ask each team lead: if the system came back with yesterday's data, what would break?
- Describe the cost of downtime in plain terms: orders not taken, staff idle, customers waiting.
- Write both limits at the top of the inventory from step 1.
- Compare them with how often backups actually run today.
What a pass looks like
Two plain sentences, approved by someone with authority, that a technician can test against.
Watch out for
Limits that are implied but never stated leave nobody able to say whether a restore passed. Demanding near-zero loss for every system is the opposite mistake, because tighter limits cost more. Save them for the systems that earn them.
3. Can the latest backup be restored onto a spare copy?
A restore test rebuilds the system from the latest backup in a separate place, such as a temporary server, and never over the live data. Someone technical performs it while someone from the business watches and times each stage. The aim is to learn what actually happens, including every missing item and every step that needed guessing.
What to do
- Create a spare environment, separate from the live one, with its own access.
- Switch off outgoing email, SMS and payment connections in the spare copy, so it cannot contact real customers.
- Pick the most recent backup, not a hand-picked good one.
- Follow the written steps if they exist, and write them down as you go if they do not.
- Record the start and finish time of each stage.
- Note every question, workaround and missing item.
- Delete the spare copy when you finish, because it holds real data.
What a pass looks like
The system starts on the spare copy using only the backup and the written steps, with no secret knowledge held in one person's memory.
Watch out for
Restoring over the live system 'to see if it works' is the most damaging mistake in this article. Using a freshly made backup instead of the one that would really be used skips the question of whether last night's job ran. Treat the spare copy as live data for privacy, including who can access it and when it is deleted.
4. Does the restored system actually work?
A system that starts is not a system that works. Verification means logging in, opening real records, completing one real task end to end and comparing totals against the live system. It checks that data is complete, files open, integrations connect and the latest records are present, which is where silent backup failures show up.
What to do
- Log in as an ordinary user and as an administrator.
- Compare record counts and a few totals, such as orders this month, against the live system.
- Open a sample of uploaded files and documents.
- Complete one real task, such as creating an order and its invoice.
- Check the date of the newest record against the data-loss limit from step 2.
- Confirm that outgoing connections are still switched off.
What a pass looks like
People who use the system confirm it with their own eyes, and the numbers match.
Watch out for
Verification by the technician alone usually proves that services start, not that the business can work. Checking only one table is another trap. A restore that is complete for customers but missing last week's orders will pass a shallow check.
5. Who runs this again, and where is it written down?
A recovery runbook is a short, plain document that lets someone other than the original builder restore the system. It lists the backup locations, access arrangements, the ordered steps, the timings you measured and who to call. Without it, a one-off test proves the system could recover once, not that it will recover again.
What to do
- Turn the notes from step 3 into numbered steps.
- Add where the backups are, who holds the credentials and how emergency access works.
- Add the measured times and the checks from step 4.
- Name an owner and a deputy.
- Set a test schedule, plus extra triggers such as migrations, provider changes, key rotations and major releases.
- Store the runbook outside the system it describes.
What a pass looks like
A colleague who was not involved can read it and perform the restore, and the owner's calendar shows the next test.
Watch out for
A runbook saved on the very server it describes is unreachable when that server is down. Credentials known to only one person are the other classic gap, and what happens if your lead developer resigns tomorrow shows how quickly one departure can lock a team out.
Where the effort goes, and what skipping a step costs
The table compares the five steps on the effort each takes and the risk you carry if it is skipped.
| Step | Effort | Risk if skipped | Usually owned by |
|---|---|---|---|
| 1. List what to protect | Low | High | You, with the person who manages the system |
| 2. Set loss and downtime limits | Low | Medium | You and your operations lead |
| 3. Restore onto a spare copy | High | High | A technical person, with you watching |
| 4. Verify the result | Medium | High | The staff who use the system |
| 5. Write the runbook | Low | High | A named owner and a deputy |
The restore itself takes the most effort, yet the cheap steps around it, setting limits and writing things down, decide whether the test is ever repeated.
Run your first restore test with a second pair of hands
Share how your app is hosted and backed up, and we will help you plan a safe test on a spare copy.
Plan your restore testPass, patch or escalate: reading the result of your first restore
Pass when the restored system works, meets your time and data-loss limits and someone other than the original builder can repeat the steps. Patch when it works but misses the limits or depends on one person. Escalate when the restore fails, data is missing or nobody can run it. Retest after every fix.
A backup is a hope until someone has restored from it.
When a light check is enough
A simple brochure website that rarely changes can usually be checked with one restore of files and database, perhaps twice a year, often through the hosting provider's own restore tool. Keep a note of the steps and the date.
When a full test is essential
Run all five steps when the system holds customer or financial records, takes payments, runs daily operations or would stop revenue if it were down for a day.
When outside help earns its place, and when it does not
A specialist helps when nobody on your side understands how the system is built, when data is regulated, when the first test fails for unclear reasons, or when the restore is too slow for your limits. It is not needed when someone technical already owns the system, tests regularly and shows you the results.
Outside help can run the test, but only you can decide what recovery must mean for your business.
How KSoft Technologies Helps With Backup and Recovery
We offer maintenance and support for custom software, and backup and disaster recovery are part of that work. A first restore test is a natural place to start, because it shows what is really being protected before anyone changes the setup.
Depending on your system, we can help you list what needs protecting, agree recovery limits, run a restore on a spare copy and write the runbook. If your hosting provider or in-house developer already tests restores and can show you the results, we will say so.
If you would like to talk through your recovery options, bring your hosting details and the date of your last known restore. More from our team is on the KSoft Technologies YouTube channel.
Make recovery a habit before it becomes an emergency
A backup is a promise from your past self to your future self. A restore test is how you find out whether the promise holds, while there is still time to keep it.
The five steps are modest. List what must come back, state how much loss and downtime you can live with, rebuild from the latest backup on a spare copy, check that real work can be done, then write it down and schedule the next test.
The sensible next step takes one afternoon. Ask whoever manages your system for the date of the last restore and the location of the written steps. If either answer is vague, book a test for this month, with someone from the business in the room.
A first test often finds something uncomfortable, and that is the point. Every gap found on a quiet day is a gap that will not surprise you on a bad one.
Write the runbook before you need it
A tested restore and a one-page runbook turn a bad day into a manageable one. Talk to us about setting up both.
Start your recovery planFrequently Asked Questions
What is a disaster recovery plan, and do small businesses need one?
A disaster recovery plan is a short written guide for getting a system back after a serious failure. It names who acts, in what order, which backups to use and how to confirm success. Small businesses need one because a single failed disk, deleted account or ransomware event can stop trading, and improvising under pressure wastes hours.
Why do backups fail silently?
Backup jobs usually fail quietly because nobody is watching them. A changed password, full storage, expired access key or moved server can stop the job, while the schedule still looks fine. Without alerts and a regular restore test, the failure surfaces only on the day you need the copy.
Why is keeping backups on the same server risky?
If backups sit on the same server or account as the live system, one event can destroy both: hardware failure, a deleted account, malware or an accidental command. A safer setup keeps at least one copy in a separate location under separate credentials, so a single mistake or attacker cannot reach everything.
How often should I test restoring my backups?
Test at least once a quarter for systems that hold customer data or revenue, and at least once a year for simple ones. Add a test whenever something major changes, such as a migration or a new hosting provider. Frequency should match how much change and loss your business can tolerate.
What is the 3-2-1 backup rule?
The 3-2-1 rule is a common guideline: keep three copies of your data, on two different types of storage, with one copy held off-site. It protects against a single point of failure. It is a starting point, not a guarantee, so you still need to test that the copies restore.
Should I rely on my hosting provider's automatic backups?
Provider backups are a useful layer, but do not assume they cover everything. Check what is included, how long copies are kept, whether files and settings are covered, and who can restore them. Add an independent copy under your own control, and test a restore from whichever source you plan to depend on.
Can I run a restore test without taking my live system offline?
Yes. Restore onto a separate spare copy of the system, such as a temporary server or test environment, and leave the live system running. Never restore over live data as a test. Treat the spare copy with the same care as live data, because it holds real information, and delete it afterwards.
When should I run an extra restore test outside the schedule?
Run an extra test after a migration, a change of hosting provider, a major release, a password or key rotation, or a change in who manages the system. Each can silently break a backup or its restore steps. A test within days of such a change catches the break while it is cheap to fix.
How do I start a restore test this week?
Pick your most important system, list what it needs to run, and ask whoever manages it to restore the latest backup onto a spare copy while you watch. Time it, then log in and complete one real task. Write down every surprise. That single session will tell you more than the backup dashboard.
How long should a restore take?
It depends on data size, storage type and how the system is set up, so measure it instead of guessing. Compare the measured time with the downtime your business can tolerate. If the restore takes much longer than that, change the design, such as keeping smaller, faster backups or a standby copy.
What does backup and recovery testing cost?
The main cost is a few hours of technical time plus temporary hosting for the spare copy. No standard figure exists, because it depends on system size, complexity and who does the work. Ask your provider whether restore testing is included in your plan before paying for it separately.
Should I hire a specialist for backup testing or do it in-house?
Do it in-house if someone technical owns the system and has time to repeat the test. Bring in a specialist when nobody on your side knows how the system is built, when data is regulated or high-value, or when a first test fails and the cause is unclear. Either way, keep the runbook on your side.
