Backup Strategy for Small Business: 3-2-1 Rule and Restore Tests
A practical backup strategy for small business: apply the 3-2-1 rule, protect copies from ransomware, cover SaaS and n8n, and test restores every quarter.
Most small companies have backups. Far fewer have a backup strategy, and almost none have tested whether they can actually restore. This guide walks you through a backup strategy for small business that covers files, servers, cloud apps like Microsoft 365 and your CRM, and self-hosted tools like n8n, and it ends with a quarterly restore test and a one-page runbook you can fill in this week.
Backup strategy for small business: start with what to back up
When owners think about backups, they usually think about the shared drive. In a modern small company, the shared drive is often the least critical piece. Your business now lives in a dozen places, and each one fails differently.
Start with an inventory. Open a spreadsheet and list every system that holds data you would miss:
| Category | Typical examples | What is easy to forget |
|---|---|---|
| Files | File server, NAS, shared folders, laptops | Files that only exist on one person's laptop |
| Email and collaboration | Microsoft 365, Google Workspace | Shared mailboxes, Teams or Chat history, calendars |
| Business apps (SaaS) | CRM, accounting, project management, shop | Custom fields, attachments, automation rules |
| Databases | PostgreSQL or MySQL behind your apps, Directus, internal tools | The database is the app; the server is replaceable |
| Automation | n8n workflows, credentials, encryption key | The encryption key that makes stored credentials readable |
| Configuration | Docker Compose files, .env files, DNS records, firewall rules | The settings nobody wrote down |
| Secrets | Password manager vault, API keys, recovery codes | Admin recovery codes for your main accounts |
For each row, write down three things: who owns it, how much data you could afford to lose (one hour, one day, one week), and how long you could live without it. Those two numbers are your recovery point objective (RPO) and recovery time objective (RTO). You don't need the jargon, but you need the answers, because they decide how often you back up and how much effort you put into fast restores.
A common result: the CRM and the accounting system can lose at most a few hours of data, the file share can lose a day, and the marketing website can be down for two days without real harm. That ranking tells you where to spend your time first.
The 3-2-1 backup rule, translated for a small company
The 3-2-1 backup rule is the classic starting point:
- 3 copies of your data: the live data plus two backups.
- 2 different types of storage, so one failure mode cannot take out everything.
- 1 copy off-site, away from your office and your main server.
For a small company today, "different storage" and "off-site" mean something slightly different from when the rule was coined. Here is a practical translation:
- Copy 1 is your live system: the server, the SaaS app, the laptop.
- Copy 2 is a fast, local or same-provider backup for everyday mistakes. Think snapshots of your virtual server, a NAS in the office, or the version history in your file sync tool. It is fast to restore from, but it shares some risk with the original.
- Copy 3 is an independent backup with a different provider, different credentials and ideally a different region. If your hosting account is compromised or closed, this copy must still exist.
Many teams now add two more rules on top, sometimes written as 3-2-1-1-0: one copy that is offline or immutable, and zero errors after a verified restore. The extra digits are worth taking seriously, because they address the two ways backups fail in practice: attackers delete them, or nobody notices they stopped working.
One more translation: "different credentials" matters as much as "different location." If the same admin account that runs your server can also delete your off-site backups, you have one point of failure, wherever the data physically sits.
Ransomware backup protection: keeping copies safe from attacks and accidents
Ransomware backup protection comes down to one idea: an attacker who gets into your network, or an employee who makes a mistake, must not be able to delete or encrypt every copy.
Separate the credentials
The backup system should push data to a destination that the production server cannot clean up. Practical ways to do that:
- Use a dedicated backup account at your storage provider, with permission to write new backups but not to delete old ones.
- Keep the admin login for the backup storage out of your normal password vault folder that the whole team shares. Protect it with two-factor authentication.
- Never store backup storage keys in the same
.envfile as everything else on the server, if you can avoid it. If the server is compromised, assume everything on it is known to the attacker.
Use immutable backups
Immutable backups for small business are no longer an enterprise-only feature. Several object storage providers offer a write-once mode (often called object lock) where a file cannot be changed or deleted until a retention period expires, not even by the account owner. Some backup appliances and NAS systems offer similar locked snapshots.
Set the retention lock to cover the time you might take to notice a problem. If an attacker sits in your systems for two weeks before encrypting anything, a seven-day lock does not help you.
Keep versions, not just copies
A backup that is overwritten every night with the latest state will happily copy encrypted or corrupted files. Keep a version history: for example, daily backups for 14 days, weekly for 8 weeks and monthly for 12 months. Tools like restic or Borg handle this with deduplication, so older versions cost little extra storage.
Encrypt before upload
Encrypt backups on your side before they leave your server, and store the encryption key somewhere separate (a printed copy in a safe is not old-fashioned, it is sensible). An encrypted backup without its key is worthless to you as well.
Backing up cloud apps and self-hosted tools like n8n
This is where most guides stop too early.
SaaS apps: read your contract
Many owners assume Microsoft, Google or their CRM vendor "has backups." They do run redundant infrastructure, but that protects the provider against hardware failure, not you against deleted data, a compromised account or a sync gone wrong. Most providers describe this as shared responsibility: they keep the service running, you are responsible for your data.
The exact terms vary by provider and by plan, and they change over time. Check your own contract and admin documentation for three things:
- How long deleted items stay recoverable, and whether an admin can restore them.
- Whether there is a built-in export or backup API you can use on a schedule.
- What happens to your data if you stop paying or the account is suspended.
If the built-in retention does not match your RPO, use a third-party backup service for Microsoft 365 or Google Workspace, or a scheduled export. For smaller SaaS tools like a CRM or project management app, a weekly automated export to your own storage is often enough. An n8n workflow can pull data through the app's API and write it as JSON or CSV to an object storage bucket, then alert you when the export fails.
Self-hosted databases
For PostgreSQL, a nightly pg_dump gives you a consistent, portable backup. Copying the raw database files of a running database does not, unless you stop it or use a snapshot method designed for that.
# Nightly logical backup of a Postgres database running in Docker
docker exec postgres pg_dump -U app -Fc appdb > /backup/appdb-$(date +%F).dump
Then let your backup tool pick up /backup and ship it off-site. If you can lose at most an hour of data, look at continuous archiving (write-ahead log shipping) instead of nightly dumps.
n8n
A self-hosted n8n instance has three parts you must back up together:
- The database (SQLite by default, PostgreSQL in most production setups). It holds workflows, credentials and execution history.
- The encryption key. n8n encrypts stored credentials with a key. If you lose it, a restored database gives you workflows but unusable credentials, and you will re-enter every API key by hand.
- Your configuration: the Docker Compose file, environment variables and any custom nodes.
On top of database backups, export your workflows as JSON and keep them in Git. That gives you a readable history of every change and a second, independent copy. We describe that setup in n8n as code with GitOps. If you run local AI models next to n8n, as in our guide to self-hosted AI automation with n8n and Ollama, include the model configuration and any vector database in the inventory too; the models themselves can usually be downloaded again.
Backups are also part of hardening. The n8n security hardening guide covers the access control side; backups cover what happens when access control fails.
Testing restores: a quarterly routine that takes an hour
A backup you have never restored is a hope, not a plan. Restores fail for boring reasons: a missing encryption key, a database dump that was empty for three months, a password nobody remembers, a backup job that silently stopped when a disk filled up.
Here is a test backup restore routine that fits into one hour per quarter.
Before the test (10 minutes)
- Pick one system from your inventory. Rotate, so every critical system is tested at least once a year and the most critical ones every quarter.
- Pick a random file, record or point in time, not the one you know works.
- Prepare an isolated place to restore into: a temporary virtual server, a separate folder or a test account. Never restore over production for a test.
During the test (40 minutes)
- Start a timer.
- Follow only the written runbook. If you need knowledge that is not written down, note it.
- Restore the data and the configuration.
- Check that it works, not just that files appear: open the restored document, log into the restored n8n and run a workflow with test data, query the restored database for yesterday's records.
- Stop the timer and compare the time with your RTO.
After the test (10 minutes)
- Update the runbook with everything you had to look up.
- Record the date, system, result and time taken in a simple log.
- Delete the test environment, including any copies of personal data in it.
Daily monitoring belongs here too. Every backup job should report success or failure somewhere a person looks. An alert when a job fails is good; an alert when a job has not reported at all for 26 hours is better, because it catches the job that never started.
A one-page backup runbook template
Copy this into your wiki or a shared document and fill it in per system. Keep a printed copy, because the wiki might be one of the things that is down.
SYSTEM: ______________________ OWNER: ________________ DEPUTY: ______________
What is backed up: ____________________________________________
Backup method/tool: ____________________________________________
Schedule: ____________________________________________
Retention: daily ___ / weekly ___ / monthly ___
Copy 2 (fast restore): ____________________________________________
Copy 3 (independent): ____________________________________________
Immutable/offline? yes / no Lock period: __________
Encryption key stored: ____________________________________________
Max data loss (RPO): __________ Max downtime (RTO): __________
Restore steps:
1. ______________________________________________________________
2. ______________________________________________________________
3. ______________________________________________________________
How to verify it works: _____________________________________________
Access needed (accounts, where credentials live): ____________________
Monitoring/alerts go to: ____________________________________________
Last restore test: ________ Result: ________ Time taken: ________
The "deputy" line matters. If the only person who can restore your server is on holiday, your real RTO is the length of the holiday.
Frequently asked questions
How often should a small business back up?
Match the schedule to how much data you could afford to lose. Business-critical databases often need hourly or continuous backups, file shares usually daily, and slow-changing systems weekly. Decide per system, not once for everything.
Is cloud sync like OneDrive or Google Drive a backup?
Not on its own. Sync copies deletions and encrypted files to every device. Version history and recycle bins help with small mistakes, but they are limited in time and controlled by the same account. Treat sync as copy 2 at best.
Do we need immutable backups if we are a small company?
If you would struggle to recover from losing all your data at once, yes. Automated attacks do not check company size, and an immutable copy is often the cheapest insurance against a compromised admin account.
How long should we keep backups?
Long enough to notice problems, usually several months for operational data. Legal retention duties for invoices or personnel records are a separate question; check them with your tax advisor or lawyer. Keeping personal data longer than needed can conflict with GDPR or PDPA.
Who should do the restore tests?
Ideally someone other than the person who set up the backups. If they can restore using only the runbook, the runbook is good enough.
Get a second pair of eyes on your backups
Writing the inventory and running the first restore test is the hardest part, and it is worth doing yourself so you understand your own systems. If you want help, our IT Security service includes backup and incident readiness: we review what you back up, set up immutable off-site copies for servers and self-hosted n8n, and write the runbook with you so a broken server stays an incident and not a crisis.