Every few years another platform closes, takes its archive with it, and a number of artists discover that the only copy of work from a particular period was the one they uploaded. This is entirely avoidable and the practices that avoid it are boring, cheap and well established. The main obstacle is that nothing bad happens for years, and then it all happens at once.
The goal is not merely to keep files. It is to keep files that are findable, openable and identifiable a decade later, which requires slightly more thought than copying a folder somewhere and hoping.
Keep the masters, not the exports
A flattened export is a photograph of a decision. A layered master is the decision itself, and it is what lets you fix something, produce a different crop, re-colour for print, or extract an element years later. Storage is cheap enough that discarding masters is almost never the right trade.
Where a master is in a proprietary format, keep a flattened archival copy alongside it in something durable, because file formats age and applications disappear. A lossless format with wide support is the usual choice, and the extra file costs almost nothing against the risk it removes.
Three copies, two media, one elsewhere
The old backup rule remains correct: three copies of anything, on at least two different kinds of storage, with at least one physically somewhere else. A single external drive is not a backup, because the most common causes of loss are theft, fire, flood and dropping the bag, and all of them take the drive too.
Automate it. Backups that require a person to remember are backups that stop happening in the third week of a busy month. Whatever system you use, test a restore once a year, because an untested backup is a belief rather than a copy.
Name files so they sort
Dates first, in year-month-day order, so folders sort chronologically without any tooling. Then a short project code, then a description, then a version. Avoid spaces if you ever move files between systems, and never rely on a folder's position in a hierarchy to tell you what something is, because hierarchies get reorganised.
Include the character or series name in the filename rather than only in a folder. Files escape their folders constantly, and a file that carries its own identity survives that. This is the same principle behind embedding a signature in an image rather than relying on a caption, as described in the piece on artist credit.
Metadata is worth the two minutes
Filling in author, title, date and a short description inside the file takes very little time and makes an archive searchable by tools that have not been invented yet. It also travels with the file, which captions and spreadsheets do not, and it is the only attribution that has any chance of surviving a copy.
A simple index alongside the archive helps too: a plain text or spreadsheet list of works, dates, clients and where each was published. It is not glamorous, and it is what makes it possible to answer questions about your own back catalogue years later without opening several hundred files.
Platforms are distribution, not storage
Treat every upload as a copy you may lose, because you may. Platforms close, change policy, and delete categories, sometimes with notice and sometimes without, as described in the piece on why publishability varies. Nothing that exists only on a platform is safe.
The corollary is to keep a canonical home you control, plus a mailing list, so an audience can find you when a channel disappears. That is a resilience measure rather than a marketing one, though it happens to be good marketing as well.
Keeping an archive that outlives the platforms
In practice: layered masters plus an archival flat copy, three copies across two media with one off-site, dated and self-describing filenames, embedded metadata, a simple index, and no reliance on any platform for storage. An afternoon to set up and a few minutes a week thereafter.
The commercial reasons this matters are covered in the piece on where the money comes from, since an archive is an asset and licensing requires being able to produce the original file.