Everyone believes in backups the day after losing something. The decade of family photos on a single ageing laptop, the thesis on one USB stick, the business records on a drive that clicked twice and went silent... data loss is rarely a surprise in hindsight, because single copies of important data are not "backed up by luck"; they are pre-disasters on a schedule nobody knows. The fix is a structure, not a product, and the structure fits in three numbers.
3-2-1, unpacked
Keep three copies of anything that matters: the original plus two backups. Keep them on two different types of storage, because a batch of identical drives bought together has a habit of failing together. Keep one copy off-site, because fire, theft and floods do not distinguish between your computer and the backup drive in the same drawer.
In practice, for most households, this translates to: the working copy on your computer or phone, an automatic backup to an external drive at home, and a cloud copy. Each leg covers the others' blind spot. The external drive restores terabytes in hours when the laptop dies; the cloud copy survives the burglary that takes both machines; and three copies means no single click, spill or failure is ever final.
What actually needs backing up
Not everything... and trying to back up everything is how backup projects die. Sort by replaceability. Irreplaceable: photos and videos, personal documents, tax and business records, creative work, password manager exports, scans of identity papers. Replaceable with effort: purchased media, software settings. Replaceable trivially: the operating system, applications, downloads. Aim the structure at the first category, be relaxed about the second, ignore the third. For most people the irreplaceable core is under a couple of hundred gigabytes... a far smaller and cheaper problem than "my whole disk".
Automation or it does not exist
The most common backup failure is not technical: it is the manual routine that lasted three weeks. A backup that requires remembering is a backup that stops the month before you need it. Every layer should run unattended... the operating system's built-in tool to the external drive (Time Machine, File History or equivalents, set once), the cloud layer syncing continuously in the background. Your role shrinks to glancing occasionally at a status icon, which is a job humans can actually hold.
The two failure modes people discover too late
First, the unverified backup. A backup nobody has restored from is a hope, not a plan. Twice a year, pick a random file and actually restore it... the five-minute drill that catches the drive that silently stopped in March, the sync that has been erroring behind a closed laptop lid, the cloud folder that was never actually selected. Schedule it with the clock changes if you need an anchor.
Second, the connected backup. Ransomware encrypts every drive it can reach, including the backup drive that lives permanently plugged in, and sync services will faithfully replicate the encrypted versions over your good copies. The defences: keep the local backup drive disconnected between runs, or use a backup tool with versioning so last week's intact copies survive this week's disaster, and prefer cloud services that retain file history. One generation of versions has rescued more people from both malware and their own accidental deletions than any other single feature.
The weekend version
Saturday morning: list your irreplaceable data and find out where it actually lives (the honest answer usually includes "scattered"). Saturday afternoon: buy one external drive sized at twice that volume, plug it in, switch on the system backup tool. Sunday: enable cloud sync or backup for the irreplaceable core, then verify both copies exist by restoring one file from each. Put a reminder in your calendar for a restore drill in six months. That is the whole project... a few hours and the cost of one drive, against losses people routinely describe as worse than the hardware that caused them.