Backing Up a Working Computer

No affiliate links on this site, and nothing here is ranked.

What the task actually is

Being able to carry on working after a machine is lost, stolen, dropped, encrypted or flooded.

Not "having a copy somewhere." The test is what you would do on the morning after, and how long it would take.

Two copies, in two places, one of them not in the building. That is the requirement, and everything else is implementation.

What the operating system does

Automatic backup to an external drive, configured once. Both major systems include this and it works.

It covers the commonest failure by a distance: the machine breaks and the files are on a drive in the drawer.

Where it stops: the drive is in the same building as the machine, connected to it, and subject to the same fire, flood, theft and encryption.

A local backup is half the answer and it is the cheaper half.

Why a cloud drive is not a backup

Because deletion synchronises.

A file removed on the machine is removed in the drive. Ransomware encrypting local files encrypts what synchronises.

Most services offer version history and a deleted-items window — commonly thirty days — which covers accidents and does not cover anything discovered later.

A cloud drive is an excellent way to have your files on several machines. It is not a backup, and the distinction is the one most commonly got wrong.

What an offsite backup adds

Distance. The copy is not where the machine is.

Versions over a long period, so that a problem discovered in March can be recovered from February.

And independence. A separate account, a separate provider, so that losing access to one thing does not lose both copies.

This is the subscription worth having, and it is one of very few on this site recommended before something has gone wrong — because the failure it protects against is total and arrives without notice.

What to check

What is actually included. External drives, network locations, and the folders you assume are covered but are not.

How restoration works, and whether you can restore one file or only everything.

How long it takes. A restore that takes four days is a plan with a four-day gap in it, and knowing that in advance changes what you do on the day.

And the retention period. How far back you can go, which is the difference between recovering from an accident and recovering from something that went unnoticed.

The thing that makes it real

Restore something, once, deliberately.

Pick a file, delete it, restore it from the backup. Twenty minutes.

A backup nobody has ever restored from is an assumption, and the failure modes — a drive that was never actually writing, a schedule that silently stopped, an account whose payment lapsed — are all invisible until the day they matter.

Repeat annually, and note the date. It is the only test that distinguishes a working backup from a configured one.

What to back up, and what not

Everything that would take real time to reproduce.

Documents, records, photographs of work, the accounting file, anything a customer sent.

Not applications, which reinstall. Not the operating system, unless a full-image restore is the plan and you have tested it.

And not things that live somewhere else already — records inside a tool are the tool's problem until you export them, and the answer there is a periodic export into the backed-up folder rather than a second backup product.

That last point catches people out. A firm with excellent machine backups can still lose everything in a cloud tool it never exported.

The offsite copy nobody thinks of

A drive kept somewhere else.

Not elegant, entirely adequate, and free after the drive. Two drives, alternated, one at home and one at the workshop, swapped monthly.

It is worse than a service — manual, monthly rather than continuous, dependent on somebody remembering — and it is enormously better than the local-only arrangement most small firms actually have.

Where a subscription is genuinely not wanted, this is the version to use rather than concluding that offsite is out of reach.

Encryption, briefly

A backup drive holding customer records is customer records sitting in a drawer.

Both operating systems can encrypt an external drive with a few clicks, and doing so converts a theft from a disclosure into an inconvenience.

Note the passphrase somewhere that is not the machine being backed up, which is the failure that turns an encrypted backup into no backup.

The three failures that produce total loss

Worth naming, because each defeats a different half of the arrangement.

Fire or flood defeats the local copy, because the drive is next to the machine.

Ransomware defeats the synchronised copy, because encryption propagates.

And account loss — a forgotten password, a lapsed payment, a provider closing — defeats the cloud copy while leaving the machine untouched.

Two copies in two places covers all three only if the two are independent, and a local drive plus an offsite service is the arrangement that satisfies it.

A local drive plus a synchronised cloud folder does not. It looks like two and behaves like one against the second failure. For another point of comparison, consult Asana.

The short version

• The requirement is two copies in two places, one not in the building, and the test is what you do the morning after
• Both operating systems include automatic local backup, which covers the commonest failure and is the cheaper half
• A cloud drive is not a backup: deletion synchronises, and version windows are commonly thirty days
• Offsite backup adds distance, long version history, and independence from a single account or provider
• Check what is included, how restoration works, how long it takes, and how far back it goes
• Restore something deliberately once a year — a backup nobody has restored from is an assumption