The 3-2-1 Backup Rule, Applied to a Real Home-Office Computer
The 3-2-1 rule is the most repeated backup advice on the internet and the least implemented, because it is usually explained as an abstraction rather than as a decision about your actual machine and data. This version assumes one home-office computer — from what counts as a copy to the restore test that proves you have one.
What 3-2-1 actually asks for
Three copies of anything you would hate to lose — the live data plus two independent backups.
Two different kinds of media, so a single failure mode cannot take both: an internal disk plus an external drive, not two identical drives from the same batch.
One copy somewhere else — a different building: a drive at a relative's home, or a cloud account not used for everyday file sharing.
The idea is usually credited to Backblaze; see its explanation of the strategy. A rule of thumb rather than a standard — the point is not the digits but that redundancy and physical separation solve different problems.
First, the trap: sync is not backup
Most home-office "backups" are synchronisation, which fails in four predictable ways:
- Deletions propagate. Delete a folder by accident and it vanishes from every connected device and the cloud copy.
- Bad edits propagate. Wrong program, lossy re-save, and the corrupted version is now authoritative everywhere.
- Encryption propagates. Ransomware on the synced machine encrypts what it can reach, including the cloud folder, unless the service keeps independent versions.
- Access is not ownership. A billing failure, lockout, or policy dispute can cut you off from everything in one account.
Version history helps — see Google Drive vs Dropbox vs OneDrive for how the three handle conflicting and restored versions. But a rolling 30-day window is not an archive of a project you touched last March, and a lockout blocks the history along with the files.
A concrete plan for one machine
Assume a laptop with roughly 500 GB of data: documents, a photo library, and one client project folder.
| Copy | Where it lives | How it is made | What it protects against |
|---|---|---|---|
| 1 — live | Internal drive | Nothing; this is the original | — |
| 2 — local backup | External SSD or HDD attached most days | Time Machine on macOS, File History plus a system image on Windows | Accidental deletion, bad saves, a failed disk, most ransomware while the drive is unplugged |
| 3 — off-site | Cloud backup account, or a second drive rotated out monthly | A dedicated backup client, or a scheduled rotation | Fire, theft, flood, or anything taking the laptop and dock drive together |
The local backup should be a backup application, not a folder you drag things into.
Capacity and hardware reality
Buy a destination at least twice the size of the data you protect: backup software keeps multiple versions, and incrementals after the first full copy fill the drive. Rankings for the best external SSDs, the best external hard drives, and the best NAS devices cover the trade-offs; our backup-focused external drive list is closest to this use case. A NAS can be a local target, but a NAS inside the same network is not your off-site copy.
Turn on the automation you already have
Neither platform needs third-party software for copy 2.
macOS. Time Machine keeps hourly, daily, and weekly versions to an attached disk or supported network volume; steps are in Apple's Time Machine user guide. Two things people miss: Time Machine does not back up the macOS system files themselves, so a full restore means reinstalling the OS and pulling data back; and iCloud Drive and iCloud Photos already live with Apple, so decide whether you want them inside the local backup too.
Windows. File History saves repeat copies of items in your libraries and chosen folders; setup and restore steps are in Microsoft's Backup and restore with File History. It is not a whole-disk image, so pair it with a system image or treat reinstalling Windows as part of your recovery plan. If your machine runs a release whose free updates have ended, that is a reason for a working backup, not to postpone one.
Add copy 3 with a purpose-built service if the volume justifies a subscription; our comparison of the best online backup services and the best cloud storage explains which keep independent historical versions rather than mirroring your current state.
The offline copy is the one everyone skips
Three copies mean little if all are always connected, because malware, a compromised account token, and a confused sync client reach everything plugged in. The fix is cheap: one destination that is deliberately not always available. Unplug the Time Machine drive after a backup weekly. Keep the cloud backup in an account with its own password and second factor, not mapped as a drive letter. Rotate a physical drive to another location. An offline copy turns "we got hit by ransomware" into an inconvenience instead of a negotiation.
Run a restore drill (this is where the plan becomes real)
Schedule it quarterly, working through these in order:
- Single file, from last month. Restore one document you edited last month from the version history. This proves the history exists, not just the latest mirror.
- A whole folder to a different location. Restore your most important project folder to the Desktop, not over the original. This proves you can recover without risking the live data.
- Off-site pull. Restore something from copy 3 specifically. Off-site copies silently stop updating — often because a client uninstalled itself after an OS upgrade.
- Annually: full-system feasibility. Confirm you can get from a dead machine to running again — which installer, which key, which account, which drive. You need not wipe a laptop; write down the path and check the pieces are where you think they are.
Keep a one-page note beside the backup drive with three lines: where the three copies are, how to start a restore, and which account unlocks each one. In a real incident the person recovering is often not the person who designed the plan.
Common failure patterns
- One drive doing both jobs. One external disk serving as both your local and off-site copy is one failure from zero copies.
- Backup software left pointed at a stale path. The classic silent failure: the destination drive was renamed, reformatted, or replaced, and nothing was written for months. Check the date of the newest backup, not the app icon in your menu bar.
- Protecting files but not settings, licences, or credentials. A restored machine with no key files, 2FA seeds, or server details is a slow, expensive rebuild.
- Skipping the media difference. Two drives from the same batch can share a defect. Different media types, or at least different brands and purchase dates, spread that risk.
- Never checking the drive's health. Backup destinations are ordinary disks and fail accordingly. Our guide to reading drive health covers what SMART does and does not tell you.
FAQ
Is cloud sync the same as a backup?
No. Sync keeps locations matching, so deletions, bad saves, and encrypted files spread to everywhere that is connected. Version history helps, but a rolling window is not an archive and an account lockout reaches both.
Does a second computer in the same house count as a different medium?
Only weakly. Same power supply, same flood zone, same theft, often the same shared storage. A drive in another building, or a separate cloud account, is a genuinely different risk pool.
How often should I test a restore?
Quarterly is a sensible default, and any cadence beats none. Test one file from last month, one whole folder restored somewhere new, and one pull specifically from the off-site copy.
What about data that only exists in a web app?
It needs an export plan, because your disk backup never sees it. Schedule exports of mail, contacts, notes, and cloud-only documents into a folder that your Time Machine or File History destination actually covers.
Related: our cloud storage ranking and backup-focused external drives cover hardware and services; how we test explains the editorial process.