A backup plan is a written routine that makes at least two copies of your important data on two different types of storage, keeps one copy offsite, and keeps one copy offline or immutable. You create it by inventorying what you can't lose, choosing storage that satisfies that four-part rule, scheduling the copies automatically, and testing a restore before you trust it. The plan isn't the schedule. It's the schedule plus the evidence that the schedule works.

Most people who ask how to create a backup plan already own a backup. They just haven't tested it, and untested backups fail at a rate no storage marketing page will mention. If you get two things right, get the copy count right and the restore test right. Everything else is tuning.

Inventory data by rebuild cost before buying any storage

Start with a time estimate, not a terabyte count. Walk through every file that matters and write down how long it would take to recreate from scratch: tax returns, the family photo archive, the client project folder, the password vault. A spreadsheet you can rebuild in an afternoon needs a different treatment than a decade of unedited video footage that no longer exists anywhere else.

Group findings into three tiers. Tier one is irreplaceable β€” photos, legal documents, keys, source code, and any file whose original source is gone. Tier two is business-critical but regenerable at a cost, like cloud project configs, email archives, and design files. Tier three is convenience data β€” a media library, game saves, downloaded installers β€” where re-downloading is annoying but survivable. Your backup plan's budget should follow the tiers, not the folder structure you inherited.

Size the job honestly. A typical home user has 50–200 GB of genuinely important data once photos and documents are separated from entertainment, while a freelancer with client assets can cross 2 TB. Those numbers change the storage math significantly: 1 TB of cloud backup at current consumer rates costs more per year than a 2 TB external drive that lasts five. Note your tier-one figure before you shop.

Write the inventory down and date it. A plan that lives in your head fails the moment you forget a folder, and people routinely back up the Documents directory while their tax PDFs sit in a random download folder or a second drive. Set a recurring calendar reminder β€” quarterly is reasonable β€” to re-walk the list and catch new projects, new devices, and anything you've stopped using.

The 2-2-1-1 rule and what each number buys you

The classic 2-2-1 rule says two copies of your data, on two different types of media, with one copy offsite. A hardened extension adds a fourth digit: one copy offline, immutable, or air-gapped. That fourth copy exists because ransomware and accidental deletion propagate to connected drives and synced cloud folders within minutes, so a mirrored copy connected to the same machine isn't a backup in the sense that matters.

Two copies means one working file plus one backup, at minimum. In practice, tier-one data deserves three: a local backup, an offsite backup, and the original. Each of the four positions defends against a different failure β€” drive death, house fire, theft, user error, malicious encryption. Losing one copy should never mean losing the only copy.

Different media is the part people skip. A backup on a second partition of the same drive dies with the drive, and a backup on a synced cloud folder dies when the sync client deletes it alongside the original. Buy the difference: a local external drive for speed of restore, a cloud target for distance, and a rotating drive kept somewhere else for the offline leg.

The cost of the third and fourth copies is smaller than the arithmetic suggests. A 2 TB external drive runs a modest one-time purchase, and most cloud tiers include versioning that holds older copies for a set retention window β€” often 30 days on entry plans, longer on paid ones. Confirm the retention window on your provider's plan page before you rely on it for ransomware recovery.

Match storage to recovery time: cloud vs. local vs. rotating drive

The choice between cloud and local isn't about which is better. It's about which failure you're recovering from and how long you can wait. A cloud restore of 500 GB over a typical 100 Mbps home upload path takes hours or days, while an external drive restore of the same data takes tens of minutes. Local wins on speed; cloud wins on distance.

Storage target Copy position Typical restore speed Defends against Main limitation
External USB/NVMe drive Local copy Fastest β€” limited by USB or drive speed Drive failure, accidental deletion Dies with the house; can be encrypted by ransomware if left connected
Cloud backup service Offsite copy Slow β€” limited by upload/download bandwidth Fire, flood, theft, hardware failure Subscription cost; retention window varies by plan
Synced folder (Drive, Dropbox, OneDrive) Not a backup on its own Fast if cached locally Device loss only Deletions and encryption sync immediately; versioning window is the safety net
Rotating offsite drive Offline copy Fast when retrieved Ransomware, fire, theft, wide-area outages Manual discipline required; can be stale

The synced-folder row deserves emphasis. Cloud sync is a convenience layer, and it's excellent at what it does, but a file deleted at 9:00 a.m. is deleted in the cloud at 9:00 a.m. Version history is the only thing standing between you and data loss, and that window closes. Treat sync as a distribution method, never as your only backup.

Pick a combination, not a side. A reasonable home setup is a nightly local image, a continuous cloud backup of tier-one and tier-two folders, and a rotating drive swapped monthly. A small business adds a second offsite target or a second cloud account so a billing dispute or compromised account can't take both copies at once.

Build order: a numbered walkthrough for a working first backup

Follow these steps in order. The sequence matters because testing what you build at each stage prevents a broken backup from aging into a false sense of security.

1: Decide the tier-one list and write it down. Use the inventory from the first section. Name folders, not drives β€” "C:\Users\you\Pictures" and "D:\Clients" are clearer than "my computer." 2: Connect and format the local backup drive. Use a drive at least twice the size of your tier-one data so versioned and image backups have room to accumulate. Label it something you'll recognize, like "BACKUP-LOCAL." 3: Install backup software and create a local job. On Windows, the built-in File History (Settings β†’ Update & Security β†’ Backup, or Control Panel β†’ File History) covers user folders, or use a third-party tool for full-disk images. On macOS, Time Machine (System Settings β†’ General β†’ Time Machine) does the equivalent. Point the job at the tier-one list, not the whole drive. 4: Set the schedule and enable versioning. Daily is the sensible default for documents; hourly for active project work. Keep at least 30 days of versions where the software offers it, because versioning is your defense against silent corruption and ransomware. 5: Create the offsite job. Choose a cloud backup service, install its client, and select the same tier-one list. Start the first upload and let it run β€” full uploads can take days, so don't judge the plan until the initial seed finishes. 6: Add the offline copy. Copy tier-one data to a rotating external drive, then physically disconnect it. Store it somewhere out of the building β€” a locked drawer at work, a relative's house, a safe deposit box. 7: Set a rotation reminder. Monthly, swap the offline drive and refresh its contents; quarterly, re-walk the inventory. 8: Run a restore test before you trust any of it. Restore a folder from the local job, a folder from the cloud job, and a file from the offline drive to a scratch location. Verify the files open and the contents match.

If a step fails, stop and fix it before adding the next layer. A local job that errors on locked files silently skips them; check the job log rather than the completion message. If the cloud upload hasn't finished the first seed after a week on a home connection, your bandwidth may simply be too small for the total β€” shrink the tier-one list or budget for a longer first run.

Write a recovery runbook before the emergency

A backup plan your family or colleagues can't execute is a personal plan, not a shared one. Write one page: where the local drive lives, which cloud account holds the backup, how to sign in, where the recovery keys and the offline drive are stored, and who to call if the machine is gone. Print it and keep a copy somewhere outside the home.

Record the recovery key for encrypted backups in that runbook and in a password manager, and keep the two separate. An encrypted backup with a lost key is a very organized way to delete your own data. If your backup software supports a recovery-media USB for full-system restores, create one and refresh it after major OS upgrades.

Name a second person and walk them through one restore. The test is whether they can find the drive, sign in, and restore a folder using only the runbook. Most people discover gaps in the runbook during that walkthrough β€” an old password, a missing account recovery method, a drive stored at a relative's house nobody mentioned.

Date the runbook and note the last restore-test result at the bottom. A document that says "restore tested March 2026, files verified" tells you more about your safety than any storage brand logo, and it turns a vague intention into a record you can audit.

Test restores on a schedule and log the results

An untested backup is a hypothesis. Test it on a fixed cadence β€” quarterly for home users, monthly for anything with client or financial data β€” and restore to a scratch folder, never over live files. The rule is simple: one folder from each copy position, every cycle.

Verification means opening the restored files and comparing them to the originals. File counts and timestamps catch truncation and skipped-file errors; a spot check on a document, a photo, and a database export catches corruption. If your backup tool offers a checksum or verify option, enable it β€” reading the backup back and comparing hashes is the only way to catch silent bit rot on aging drives.

Log the result with a date and a note: pass, partial, or fail, plus which folders failed. A partial pass usually means locked files, excluded file types, or a path with a permission problem. Investigate within the same week; the longer a failure sits, the more likely you forget the detail that explains it.

The test itself is cheap. A 5-minute folder restore plus a 15-minute verification covers the common failure modes, and the log becomes the single strongest signal that your plan would actually work on a bad day.

Ransomware and accidental deletion reshape what counts as a backup

Ransomware is the scenario that breaks connected backups, and latest reports from major security vendors describe double-extortion groups that encrypt primary systems while copying the backups first. The defensive posture is the immutable fourth copy β€” a backup that the account doing the encryption can't write to or delete, whether that's a disconnected drive, an append-only cloud repository, or a backup service with immutability locks.

Accidental deletion is far more common than ransomware and deserves the same treatment. A file deleted today needs a version from yesterday, so your backup retention has to outlast the time it takes you to notice the loss. If your cloud provider keeps versions for 30 days and you discover the missing folder on day 45, the backup was there and now isn't. Check the retention window on your plan's page.

A useful edge test is a "what if" exercise. Suppose the main computer is stolen at 2 a.m. and the connected drive goes with it. Suppose you delete the wrong folder and empty the trash by habit. Suppose a synced folder's recycle bin is emptied at the same time. For each scenario, name which copy survives and how long recovery takes. Gaps in your answers are the exact edits your plan needs.

Offsite storage, encryption, and retention fine print

Offsite can mean a cloud account, a drive at a relative's house, or a bank safe deposit box. In a regional emergency β€” a flood, a wildfire evacuation, a city-wide power failure β€” offsite copies in the same metro area can share the disaster with you. A drive at a location beyond the affected area, or a cloud account hosted in another region, is the stronger choice for the failures that matter most.

Encryption belongs on every copy that leaves your control. Use the backup software's built-in encryption rather than zipping a folder with a password, and keep the recovery key out of the same bag as the drive. Cloud services may encrypt on their side by default, but server-side encryption protects the provider's disk, not your account β€” the account password and two-factor authentication are what protect your data from an attacker who gains your credentials.

Retention is where the cost hides. Cloud backup services often price by storage volume, and deleted-file retention varies by plan, sometimes by days, sometimes indefinitely. Read the plan details and note the window in your runbook. Also confirm how long a cancelled account keeps your data; cancelling the subscription can delete the backup, which is a design feature, not a bug.

Insurance and legal considerations apply here for a reason. For business data in regulated industries, your obligations to preserve records may exceed what a consumer backup plan provides, and consumer cloud terms may shift archive copies to a region you didn't expect. Consult your professional adviser and your provider's terms before treating a personal-tier backup as compliance-grade retention.

A household routine that survives sick days and moves

Plans fail on the operator, not the storage, so make the routine small. One device per person is the practical limit for home plans; put the automatic jobs on the main machine and back up the phone separately through its built-in cloud service. Two people should know the schedule and the password, and one of them should be capable of running a restore without you.

Build around triggers rather than memory. Refill the offline drive on a recurring date tied to something you already do β€” the first weekend of the month, same as the calendar reminder. Re-walk the tier-one list at the same time. After a laptop migration or a new phone, treat the move as a restore test: pull the old folder from the backup, open a few files, and confirm the new machine has them.

Set a maintenance window once a year for the bigger checks: confirm the subscription is still active and the payment method current, replace drives older than five years, refresh the recovery media, and re-run the full restore test. A backup plan that lapses because a card expired is one of the most common avoidable failures, and it's silent until the day you need it.

Expect gaps in coverage and plan around them. OneDrive and Google Drive hold the categories they're good at β€” documents and photos β€” and nothing for apps that store data only on the local disk. The audit is repetition on purpose: every new app, account, or folder is a chance for a tier-one item to fall outside the backup.

FAQ

Is one backup enough?

No. A single backup is one failure away from data loss, and the specific danger is that a connected drive, a synced folder, or a cloud account can be encrypted or deleted alongside the original. The 2-2-1-1 structure exists because each copy defends against a different failure mode, and losing any single copy should never mean losing the only copy.

How much does a basic backup plan cost?

Expect a one-time storage purchase plus a recurring cloud fee. A multi-terabyte external drive costs a modest one-time amount at retail, and consumer cloud backup subscriptions generally run a monthly or annual fee that scales with the storage volume you need; check the provider's current plan page for exact pricing. Consumer cloud rates change frequently, so treat any price you remember as a starting point only.

How often should I test a restore?

Quarterly is the reasonable floor for a home user, and monthly for anyone holding client or financial data. Restore one folder from each copy position β€” local, cloud, offline β€” and verify the files open. A 20-minute test covering three copies catches the failures that matter most, and the log gives you a dated record that the plan worked.

Do I need a backup if everything is in the cloud?

Yes. Cloud sync is not a backup. Deletions and ransomware encrypt files in the cloud as fast as they do on the local drive, and version history expires on a schedule set by your provider. The cloud is an excellent offsite copy position; it's not the whole plan, and it shouldn't be your only copy of irreplaceable data.

Why do my backups keep failing on some files?

Files that are open and locked by an application β€” email databases, browser profiles, cloud clients β€” are commonly skipped because the backup software can't read them mid-write. The fix is usually to use the application's own export for those categories or to enable a volume-level image backup that reads the disk below the file lock. Check your job log for the skipped-file list.

What happens to my backup after I cancel a subscription?

Many services delete stored data after a grace period, often 30 to 90 days, and some keep certain version types on a different schedule. Read the cancellation terms before you reach that point, and if you're switching providers, restore everything you need to a local drive first. The cancellation cliff is a silent failure mode that catches people mid-migration.

Does this protect me from ransomware?

The offline or immutable copy is the part that does, because ransomware running on your connected machine cannot reach a drive that is unplugged or a repository that refuses writes and deletions. The local and cloud copies can still be encrypted if the attacker reaches them. Rotate the offline copy, keep it disconnected except during refresh, and verify the cloud service's versioning window before you count on it.

What if the backup drive is stolen along with the laptop?

That's the scenario the offsite copy is for. If both devices are in the same bag, the local backup is gone with the laptop, and recovery depends on the cloud copy or the rotating drive stored elsewhere. Keep the offline drive away from the machine between refreshes, ideally out of the building, so a single theft can't remove both your original and every copy of it.

One boundary before you finish: this article is general educational information about backup practice, not a guarantee of recoverability, a compliance opinion, or a substitute for a professional who knows your industry's record-keeping obligations. Test your own plan against your own data, and consult a qualified local adviser for regulatory or record-retention questions.