A Drill on a Quiet Morning: Proving the Backup Works Before Anybody Needs It

A completed backup job proves that files were copied somewhere. It does not prove that they can be read, that the right files were included, or that anybody in the household or the business knows how to get them back under pressure at seven in the morning. Those are three separate questions and one exercise answers all three. Book an hour on a quiet morning rather than waiting for the day you need it, because the entire value of the drill lies in finding the problems while they are still inexpensive to fix.

Writing Down What You Are Actually Protecting

Start with the things whose loss would genuinely hurt rather than with the drives they happen to sit on. For a household that is usually photographs, documents, financial records and the recovery data for the password manager. For a small business it is the accounting file, the customer records, current job files, contracts, and the credentials for everything the business subscribes to. Write six to ten items on paper, because building the list usually produces the first finding on its own when one item turns out to live somewhere nothing was backing up.

The Drill Itself, Step by Step

Pick a file from at least three months ago rather than yesterday, since old files test retention and retention is where most configurations quietly fail, a system keeping only the last seven days being a sync with extra steps. Restore it to a different location and never over the original, because you are testing recovery rather than replacing anything and restoring over live data is how a test becomes an incident. Then open the restored file, which is the step people skip and the only one that proves anything at all.

Time the whole exercise including the part where you work out how, since forty minutes spent remembering the process is your real recovery time on a bad day and it will be worse when you are stressed. Then restore something large, because one file proves the mechanism while a whole folder tests throughput, and throughput decides whether recovering everything takes two hours or two days. Finally, walk your list of six to ten items and confirm each one by finding it inside the backup rather than by trusting a configuration screen.

What the Drill Usually Finds

Something important is missing, which is by a wide margin the most common result: a folder that moved, a new application storing its data somewhere unexpected, an account whose files live only in a cloud service nobody counted. Retention turns out to be shorter than assumed, which is adequate against a hardware failure and useless against a problem nobody noticed for a month, such as a file corrupted in February and discovered in April when somebody finally opens it.

The restore requires something you do not have, meaning a license key, an account login, or an encryption password stored only on the machine that failed, and this is the finding that turns a two hour recovery into a two week one while being entirely preventable by writing those details down somewhere separate. Or it works but slowly, which is worth knowing in advance, because if restoring the accounting data takes six hours over a home connection you want a local copy as well, decided now rather than on the morning payroll is due.

The Data That Is Not on Any of Your Machines

The drill tests files you can see, and a growing share of what people would actually miss is not on a drive at all. Email held only on a provider's servers. Documents existing solely inside a web based office suite. Photographs living in a phone maker's cloud with no local copy anywhere. Accounting data inside a subscription product, website content, customer records in an online scheduling tool, and the message history in a business chat platform that nobody has ever thought of as a filing system.

Every one of those is protected against the provider losing a hard drive, which is the failure their engineering is designed around. None is protected against the account being closed, the subscription lapsing, a billing card expiring while you are away, or somebody with access deleting things, and those are the realistic failure modes. Most major services offer a full export buried in account settings, so run one for your two most important services and store the archive with everything else, which once a year is enough and frequently reveals that an account holds less than you assumed.

What Good Looks Like Afterward, and How Often

Three copies of anything you cannot lose, two of them on storage that fails in different ways so a single bad batch of drives cannot take both, and the third kept at another address entirely. That arrangement has been standard advice for decades because it survives a failed drive, a house fire and an attack reaching everything connected to the network, and it long predates the cybersecurity frameworks the National Institute of Standards and Technology now maintains for organizations trying to write the same idea down formally.

The copy that lives elsewhere has to be genuinely disconnected or genuinely immutable, since a drive permanently plugged into the computer it protects is a second copy of the same fate. Keep a one page recovery note alongside the backups covering what is protected, where it lives, which credentials are needed and the steps, written for whoever has to do this when you are unavailable, which may well be you at four in the morning. Twice a year for a household and quarterly for a business, and always after anything changes.