Backups that survive a deleted folder, a bad sync, and a cancelled subscription
Every agency has a version of the same story. A folder gets deleted, or a designer overwrites the final file, or a subscription lapses and takes the archive with it. The recovery goes fine or it does not, and the difference was decided months earlier.
What you are actually protecting against
| Failure | Beaten by | Not beaten by |
|---|---|---|
| Someone deletes a folder | Version history, trash retention | Sync alone |
| A file is overwritten with a worse version | Versioned history going back weeks | A single mirrored copy |
| Ransomware encrypts a share | Immutable or off-account copies | Sync, which replicates the encryption |
| An account is compromised | A copy outside that account's control | Backups stored in the same account |
| A subscription lapses or a vendor shuts down | A local or second-provider copy | Anything living only in that vendor |
| A laptop is lost or stolen | Cloud copy plus disk encryption | Hoping it turns up |
Notice that most rows are defeated by the same thing: having only one place where the data lives, even if that place is replicated.
3-2-1, translated for a small team
Three copies of anything you cannot afford to lose, on two different systems, with one copy off-site and outside the account holding the original.
- Copy one: the working files, wherever the team edits them.
- Copy two: a different service or an external drive, updated automatically. Automatic matters; manual copying stops the first busy week.
- Copy three: off-site, and importantly not accessible with the same login as copy one. This is your protection against a compromised account rather than a failed disk.
An external drive that lives permanently plugged into the same machine is one power surge or one ransomware run away from being useless. Rotate it, or use a service.
Decide retention before you choose a tool
Two questions set every configuration you will make:
- How much work can you afford to lose? If the answer is one day, you need daily backups. If it is one hour, you need something continuous.
- How far back do you need to go? Damage is often noticed weeks later, when a client asks for a file nobody has opened since. Thirty days of history is a reasonable floor; a project archive may need years.
Then check that your chosen service's default retention actually matches those answers. Defaults are usually shorter than people assume.
The things people forget to back up
- Website databases, separately from files. A files-only backup of a CMS restores an empty shell.
- Email. Hosted mail is durable but not immune to a compromised or mistakenly deleted account.
- The subscriber list. Export it regularly and store the export like any other asset.
- Invoices and accounting records, as dated PDFs independent of the software that made them.
- Design source files, not just exported deliverables.
- Credentials and recovery codes, with a documented path to recover the password manager itself.
- Client assets you were given, which are often irreplaceable and rarely re-supplied willingly.
- Configuration: DNS records, deployment settings, and environment variables. Cheap to export, painful to reconstruct from memory.
Test the restore, or you do not have backups
This is the step everybody skips and the only one that proves anything.
- Pick one real file and one whole folder, at random.
- Restore them to a scratch location, from the backup, not the trash.
- Open them. Confirm the contents and the version are what you expected.
- Time it. Knowing a full restore takes six hours changes what you promise a client mid-incident.
- Write down the steps you just performed, so the next person does not improvise.
Client work deserves its own rule
Deliverables you produced for a client are their business continuity as well as yours. Two practices avoid most disputes: hand over final files in an open, portable format at project close rather than keeping them hostage in your tooling, and keep your own archive of every delivered project for as long as your contract and local record-keeping obligations require. Say which one applies in the contract, so nobody is guessing two years later.
A setup that fits an afternoon
- Turn on version history and extend retention on the storage the team already uses.
- Add a second, automatic backup to a different provider or a rotated external drive.
- Schedule database and site backups separately, and confirm they include the database.
- Export subscriber lists, accounting data, and DNS configuration on a monthly reminder.
- Restore one file today. Put a quarterly reminder on the calendar to do it again.
None of this is interesting work. It is the difference between a bad afternoon and a lost client.
No company paid for placement in this article. Verify current prices and terms with each provider before buying.