Everyone has backups until they need one. Then it turns out the backup was files-only, or lived on the crashed server, or had been silently failing since March. A working backup setup answers four questions: what, where, how often, and — the one everyone skips — does it restore?
1. Back up the whole site, not half
A WordPress site is files (themes, plugins, uploads, wp-config) plus the database (every post, order, user and setting). Some tools default to database-only. Verify yours captures both — a restore from half a backup is a rebuild with extra steps.
2. Store it off the server — the 3-2-1 shape
The most common backup failure is location: copies stored on the same server that just died, or in the same hosting account that was just compromised. Send backups automatically to independent storage (S3, Google Drive, a backup service’s cloud). The classic 3-2-1 rule — three copies, two media, one off-site — translates for WordPress as: host snapshots plus plugin/service backups to external storage. Two independent systems, so no single failure takes both.
3. Match frequency to what you can afford to lose
The question is: how much work vanishes if you restore yesterday’s copy? A weekly-updated brochure site: daily is plenty. A daily blog: daily at minimum. A WooCommerce store: an order placed after the last backup is an order lost — use real-time/incremental backups or your host’s continuous option for the database. Retention matters too: keep dailies for weeks and monthlies for a year, because some disasters (slow hacks, silent corruption) are only discovered long after they started.
4. Drill the restore
Quarterly, restore your latest backup to a staging environment and boot it. Does it complete? Is the data current? How long does it take? The first time you do this should not be during an outage. A backup you have restored is an asset; one you have not is a hope with a progress bar.
5. Automate, then verify the automation
Manual backups stop happening within a month — automation is mandatory. But automated backups fail silently (full storage, expired API keys), so the weekly maintenance habit includes glancing at the log: did it run, did it upload, what size? That thirty-second check is the difference between a backup system and a false sense of one.