A client came back four years after a job and asked for the original footage. I had it. It took me most of a day to find it, across two drives, in folders named things like "edit final" and "new folder 2".
Having a copy is not the same as having an archive, and the difference only becomes apparent when it is too late to fix cheaply.
The rule everyone quotes and half implements
Three copies, on two different kinds of media, one of them somewhere else. It is a good rule and most people implement the first part and skip the rest.
Two copies on two drives sitting next to each other on the same desk is one copy, functionally. The realistic threats — theft, fire, flood, a power event, dropping the bag they are both in — take out both.
The offsite copy is the one that does the work and it is the one people postpone. It can be a drive at a relative's house that you swap every few months. It does not have to be sophisticated.
Naming is the whole problem
The technical side of backup is largely solved. The retrieval side is not, and retrieval is what an archive is for.
What I use now, having tried several schemes.
Top level folder per project, named with the date first in year-month-day order, then the client or project name. Date first means the folders sort chronologically without any effort.
Inside, a fixed structure that never varies: footage, audio, graphics, documents, exports, project files. Same six folders every time, even when some are empty.
The consistency is the point. When you know the structure, you do not need to remember where anything was, and searching becomes unnecessary.
A text file in every project
The single highest value habit I have adopted and it takes two minutes.
A plain text file at the top of every project folder containing: what this was, who it was for, when it was shot, what camera and settings, who the contacts were, what was delivered and in what format, and anything unusual.
Four years later this is the difference between knowing what you are looking at and guessing. It is also searchable, which folder names are not.
I have started adding a line about where the other copies of this project are, which sounds paranoid and has already saved me once.
What to keep and what to discard
Keeping everything is expensive and makes finding things harder. Some judgement is required.
Always keep: original camera files for anything delivered, the project file, the final export, and any graphics or assets created for the job.
Usually discard: proxies, render caches, exports of intermediate versions, and downloaded stock you can download again.
The awkward middle is unused footage. My rule is to keep it for a year and then review, and in practice I keep it because storage is cheaper than regret. What I do is move it to slower, cheaper storage rather than deleting it.
Drives fail and the timing is not random
Worth knowing. Drive failure rates are higher in the first few months and rise again after several years, which means a new drive is not automatically safe and an old one is a known risk.
Practical consequences. Verify a new drive with a full read-write test before trusting it with anything. Replace archive drives on a schedule rather than waiting for failure. And do not buy two drives from the same batch for your two copies, because manufacturing faults cluster.
Unpowered drives on a shelf also degrade over long periods. Powering them up once a year and reading the contents is a reasonable precaution, and it doubles as a check that you can still read them.
Cloud storage as the offsite copy
Viable now in a way it was not a few years ago, with two caveats.
Upload time for video is substantial. A large project can take days on a domestic connection, and this needs to happen in the background rather than being something you do when you remember.
And retrieval costs vary enormously between services. Some archival tiers are cheap to store and expensive or slow to retrieve, which is fine for genuine disaster recovery and unhelpful if you need something back next week.
Read what retrieval costs before choosing, because the storage price is the number that gets advertised.
The test nobody runs
A backup you have never restored from is a hypothesis.
Once or twice a year, pick a project at random, and try to restore it fully from the backup alone. Open the project, relink the media, play the timeline.
The first time I did this I found that a project's media had been backed up and the project file had not, because it lived somewhere else on my machine and was outside the backup path.
That is exactly the kind of error that is invisible until you need the thing, and a single test found it.
The realistic version
All of this sounds like a lot and most of it is a one-time setup plus a two minute habit per project.
Consistent folder structure. A text file. Three copies with one elsewhere. An annual restore test.
That is enough to make a four year old project findable in ten minutes rather than a day, which is the actual measure of whether an archive works.