3 2 1 Backup Rule Explained for 2026

Keep at least three copies of important data, on two different types of storage, with at least one copy off-site or offline.

That's the practical starting point for the 3 2 1 backup rule, but location and isolation aren't interchangeable. A disk unplugged beside your computer can survive a network attack and still disappear in the same fire. A cloud copy can survive that fire and still be deleted through a stolen password.

The useful question isn't just whether a backup exists. It's whether you can recover the files, applications and settings you need when your normal equipment, premises or administrator account is unavailable.

Table of Contents

What the 3 2 1 Backup Rule Requires

The National Cyber Security Centre's Offline backups in an online world guidance, first published on 13 August 2019, defines the approach as at least three copies of important data, on two different devices or storage media, with at least one copy off-site.

Part of the rule What you keep Failure it survives What does not count
Three copies Live data plus two backups Loss of one file set Two folders on the same disk
Two media types Two devices or storage formats One disk or backup program failing, provided the copies are independent Two partitions of one drive
One off-site or offline copy Another premises, offline media, or isolated cloud Fire, theft or flood with off-site storage; site-wide malware with effective isolation A second server in the same office

Three copies includes the original. For example, your working accounts spreadsheet is the live copy. A recoverable version on a backup drive and another in an independent backup service complete the set. Duplicating the spreadsheet into another folder on its original disk doesn't protect it from that disk failing.

Different storage means meaningful separation. A computer's internal storage and a dedicated external backup drive provide separate devices. Splitting the internal drive into partitions doesn't. Equally, separate hardware won't protect everything if the same compromised backup account can erase every repository.

Off-site addresses the loss of a place. A drive held at another premises can survive an office fire. An unplugged drive in the office addresses network exposure instead, so don't treat offline storage alone as complete protection against building loss.

Separate failures, not just storage

The underlying test is simple: if one copy is compromised, another should remain intact. Buying more capacity doesn't achieve that if every copy relies on the same disk, login or building.

For websites, include both the site files and the database that holds content, orders or settings. WebinOne's website backup extension options are a useful resource when assessing that part of your coverage, but an extension still needs a deliberate storage destination and a restore procedure.

Practical rule: Count the ways your copies can fail together, not just the number of places where a filename appears.

Before choosing software, name your live data and backup destinations using the table's terms. If you can't explain which failure each destination protects against, you haven't finished designing the backup.

Why a Same Office Backup Still Fails

83% of firms performed some form of backup, but among organisations that backed up, 50% stored copies on servers or devices in the same premises as the original, according to a 2023 Opinium survey of 514 UK business leaders for Beaming.

That distinction matters more than a backup job's green tick. A server and a backup appliance can operate perfectly until someone steals both, water enters the cupboard, or an attacker reaches both through the office network.

Measure All firms
Firms that performed some backup 83%
Firms that kept no backups 17%
Backups stored in the same premises as the original, among backing-up firms 50%
Backups stored outside the office 35%
Backups kept at least 30 miles away 18%

The survey illustrates the gap between doing backups and arranging for recovery after a premises-wide incident.

An infographic titled Why a Same Office Backup Still Fails illustrating backup statistics for businesses.

The cupboard isn't a recovery site

Say a London office keeps its file server under a desk and its backup NAS, a network-attached storage device, in a locked cupboard. The lock helps against casual access. The NAS also gives a convenient way to recover from a failed server disk without downloading everything.

Neither advantage makes it a disaster-recovery location. If staff can't enter the building, they may have no access to either device. If the NAS accepts the same administrator credentials as the server, physical separation within the office won't prevent account-based deletion.

Keep the local copy for quick operational recovery, but don't give it a job it can't perform. The off-site destination must remain accessible without the office equipment, including any password vault or encryption key stored only there.

Geographical distance also requires practical judgement. Another room isn't another site, while a nearby property may share the same flood exposure or access restrictions. Choose somewhere with clearly different physical risks rather than measuring distance alone.

If the building were unusable for a week, which backup could you reach without entering it?

Write down that copy's actual location and who can retrieve it. If the answer is a drive in the office cupboard, your next task is separation, not another local backup job.

When Cloud Storage Counts as the One

A cloud copy can count as the off-site copy, but it needs to fail independently of the original. Cloud sync under the same login is a weak substitute for a separately protected backup.

Off-site means a different physical location. Offline means disconnected from the systems or networks that could attack it. Immutable means protected against alteration or deletion for a defined retention period, even though the storage may remain online.

In Sophos's 2026 UK ransomware survey, 66% of attacks encrypted data, while 78% of affected organisations used backups to recover encrypted data. Recovery depends on having something the attacker hasn't encrypted or deleted too.

Copy design Off-site Protected from network deletion Main weakness
Encrypted disk rotated off site Yes Yes, while unplugged Easy to forget the rotation
Cloud sync with the same login Often No One password or sync operation can delete both
Immutable cloud or object lock Yes Yes, for the correctly enforced lock period Restore time and egress cost
Separate cloud tenant and admin account Yes Partly Still online and needs its own checks
Managed backup repository Yes, when hosted elsewhere Usually, if access and deletion permissions are separate You must still test the restore

Choose protection you can maintain

An encrypted removable disk is straightforward, but someone must move it, check the job and keep its recovery key available. Encryption protects confidentiality if the disk is lost. It doesn't make the disk immune to deletion while connected.

Immutable object storage reduces reliance on drive rotation. However, confirm that the lock settings prevent early deletion, rather than merely hiding a delete button from ordinary users. A separate cloud tenant adds administrative separation, but shared recovery credentials can undermine it.

Before selecting a cloud destination, record:

Don't assume Microsoft 365 or another hosted application automatically supplies an independent backup. Identify the recoverable copy, its retention and who controls it.

I would favour a layered arrangement: an operational backup for convenient restores, an isolated or immutable copy, and separately governed recovery access. Mark every destination that a stolen everyday password could still delete.

How to Place the Three Copies

Start with the data and recovery limits, not a shopping list. The NCSC's guidance on mitigating malware and ransomware attacks supports automated backups, isolated copies and regular restoration tests; its recovery-point example explains why daily backups are inadequate if losing four hours of transactions is unacceptable, because the potential gap is up to 24 hours.

This layout is a useful starting point for mapping the working data and its backup destinations.

An infographic showing the 3-2-1 backup strategy process for managing and protecting digital data effectively.

Build around what you need back

  1. List the data you can't replace. Include accounts, email, family photos, CRM records, SQL databases and line-of-business files. Record where they live, including cloud applications and files saved outside shared folders. For databases, use an application-aware backup rather than assuming a copy of open files will be usable.

  2. Set the acceptable data-loss gap. This is your recovery point objective, or RPO. Choose the backup frequency from how much recent work you could bear to recreate. Check successful completion times, not merely scheduled start times.

  3. Keep the working copy where it belongs. Leave live files on the PC, server or hosted application used for normal work. A backup destination shouldn't become another everyday working folder, because staff may edit the very copy you expect to recover.

  4. Automate the local backup. Use a dedicated disk or NAS rather than spare space on the source drive. Turn on failure notifications and give someone responsibility for investigating missed jobs.

  5. Create the remote recovery destination. Choose rotated encrypted media, immutable cloud storage or an independently secured repository. Keep older versions long enough to cover damage you might discover late. Don't overwrite your only clean recovery point at every run.

  6. Separate administrative access. Give backup administration its own account, restrict deletion rights and enable multifactor authentication where supported. Store emergency access details securely somewhere that remains reachable when the primary computer fails.

  7. Assign recovery authority and a deadline. Name who may approve and perform a restore. Define the recovery time objective, or RTO, as how long you can wait before the required service is working again.

Make a copy map

Keep a concise record with each copy's name, device or service, physical location, deletion authority and maximum acceptable age. Include the encryption-key location without exposing the key in an openly shared document.

Don't count duplicate backup folders on a removable drive as independent destinations. Likewise, multiple folders within a cloud account still need an assessment of shared account and service risks.

For a visual explanation to accompany your own copy map, use the video below.

Worked Setups for Home and Office

A household in Wickford

Consider a household where a laptop holds family photographs and a home-worker's client documents. The household uses a dedicated external drive for backup and disconnects it when the job finishes.

For the remote destination, they choose either a versioned, encrypted cloud backup under separate credentials or an encrypted drive kept at a relative's house. A cloud folder that merely mirrors current files wouldn't provide the same ability to recover earlier versions.

The household must also respect any restrictions on storing client information. A work laptop may need the employer's approved backup service rather than a personally purchased account.

This kind of mixed home and working environment needs a plan that someone can maintain without specialist equipment.

A cozy home office desk with a laptop overlooking a harbor, with a second office visible through a doorway.

The tempting shortcut is a NAS left connected to the home network. It's convenient, particularly when several people need automatic backups, but it doesn't replace the remote destination.

The remaining gap is key recovery. If the laptop contains the only saved cloud password and disk-recovery key, the household may have protected copies it can't access. Keep recovery access separately, then prove that another suitable computer can retrieve a file.

A small office in Chelmsford

Consider a Chelmsford office whose server holds shared documents, a SQL database and Exchange email. A scheduled backup writes to a NAS in a locked cupboard, while an immutable cloud repository holds retained recovery points.

The NAS is the convenient source for a routine hardware replacement. The remote repository is reserved for incidents where the building or local systems can't be trusted.

The database needs more than the presence of backup files. Its recovery procedure must account for the application version, required credentials and any supporting services. Mail recovery similarly needs an actual mailbox-restoration procedure, not just a server image nobody has opened.

Keeping email available during server downtime is useful continuity, but it isn't itself a historical mailbox backup.

Setup Task still to prove
Wickford household Recover an encrypted file without the original laptop or its saved passwords
Chelmsford office Bring the database and dependent application back into usable service

The office's remaining gap is dependency recovery. Restoring shared documents won't prove that staff can enter orders, open customer records or use email. Both setups remain provisional until someone completes the relevant restore and records what happened.

How to Prove a Backup Will Restore

31% of organisations relying on backups couldn't recover all their data or failed to recover it at all, according to a 2025 survey of 200 UK IT security decision-makers reported by Data Centre News.

That's a stronger reason to test your existing copies than to add another untested destination. A successful job notification tells you the software completed its task. It doesn't prove that a restored database starts, a mailbox contains the expected messages or a recovery point predates the attack.

Use a manageable testing schedule

I recommend the following schedule as a practical starting point. Increase the scope or frequency when applications change or recovery requirements become tighter.

What you restore How often A pass looks like What does not count
A random recent file to another folder Monthly The file opens and matches the original A job that only reported success
One mailbox or database in a test area Quarterly The data is usable inside the application Restoring over the live system
A rehearsal with the office unusable Yearly A named person restores within the agreed time A written plan nobody has tried

Use an isolated test location so a rehearsal doesn't overwrite live data or introduce suspect files into production. Check permissions as well as content. A restored folder that nobody authorised can open isn't a completed business recovery.

Keep the testing habit visible to whoever owns the backup.

An infographic titled How to Prove a Backup Will Restore, providing essential steps for testing data backups.

Test the incident, not just the file

A file-level pass is useful but narrow. Full recovery may involve Microsoft 365, servers, websites, business applications, phone-system configuration and network-device settings.

Rehearse distinct failures using the destination designed for each:

A restore is successful when the intended person can use the recovered service, not when the progress bar finishes.

Record the date, copy used, recovery point, elapsed time, missing items, and person who approved the result. Include the recovery scenario so the log shows which risks you've tested.

If the recovered data is too old, change the backup schedule. If recovery takes too long, investigate download speed, replacement equipment and application dependencies. Another copy won't automatically fix either problem.

Next Steps for Your Backup Review

If you can name your independent copies and produce a recent successful restore record, keep the copy map and test log together. Review both when you add an application, move data, change administrators or need faster recovery.

For a home setup, make this week's task concrete: run the backup, safely disconnect the removable drive, place a protected copy away from the property and restore a chosen file into a new folder. Open it in the application you normally use, rather than accepting that the filename has reappeared.

Close the specific gap

Don't buy another disk until you know what's missing. Your problem might be location, shared credentials, missing application data or a recovery process that depends on the failed computer.

For a London or Essex business that can't yet prove recovery, a backup and disaster recovery review is the sensible next step. We work with customers to create a plan that fits their existing infrastructure and budget, then carry out the remaining work.

Bring your copy map, recent job results and the recovery limits agreed by the people who run the business. That gives the review something concrete to measure.

Whichever arrangement you choose, require a dated evidence log before treating the backup as ready. Recoverability is the result you're buying, not storage space.


Networking2000 offers a free back-up and disaster recovery audit carried out by our engineers, helping you identify where your copies share risks and what you need to prove through restoration. Our managed back-up includes off-site back-up, and we can build a recovery plan around your existing infrastructure and budget. Visit Networking2000 to request an audit and put your backup arrangements to a practical test.