Every editor I know has a folder called _OLD_PROJECTS or _2024_ARCHIVE that is silently eating their drive. Inside it is a mix of stuff that pays rent forever, stuff a client might genuinely come back for, and stuff that should have been wiped two years ago.
The right answer for any given project is not "always archive" or "always delete." It's a function of three things: how likely the client is to come back, what your contract actually requires, and how much your time is worth if you had to remake the work from scratch.
Here is the decision tree, by job type, plus the math for when archive bills exceed the cost of just redoing the job.
The Three Variables
Like a rental house deciding whether to keep an obscure lens in inventory or sell it, the question is always: does the carrying cost beat the cost of getting it back when someone wants it?
For a finished video project, the three inputs are:
- Client revisit likelihood. What percent chance per year that someone calls and asks for a re-cut, a re-export, or a license pull?
- Contractual obligation. Does your contract or your industry require you to hold the source material for a fixed period?
- Time-to-recreate value. If you had to remake this from scratch (because you tossed it and the client is now offering money), what would your hours be worth?
If you multiply revisit-likelihood by recreate-cost, you get the expected value of keeping the project. If that beats the annual storage bill, archive. If it doesn't, delete.
By Job Type, Practically
Most editors don't run that math job by job. They use a default by category. Here are the defaults that actually make sense in 2026.
Weddings: archive 5 years, lean. Couples come back. Anniversary edits are real. The five-year mark usually catches the "send us our wedding for our anniversary" request and the "we got divorced, please delete it" request both. Strip to ceremony footage, toast cameras (even unused angles), reception highlights, and the final cuts. About 8 percent of wedding clients will ask for something within three years.
Commercial work: archive 2 years, per typical client contracts. Most commercial deals require the production company to retain source for one to two years post-delivery for compliance, re-cut, or talent rights. After that, contractually you can usually delete. Hold longer only if the agency relationship is worth it.
YouTube and social: delete after publish. This will sound aggressive. The platform is the archive. Your final cut lives on the channel forever. The source files are not coming back into a re-edit. Keep the project file and final export, dump the source media. Save yourself $30 a month per series.
Indie film and personal documentary work: archive forever, lean. This is your portfolio. It's also your future feature. Strip aggressively but never delete the camera originals.
Corporate work: archive 1 year, then dump. Re-cut requests happen in the first six months or never. After a year, the brand has moved on, the spokesperson left, the product changed.
| Job type | Archive duration | Strip aggressively? | Revisit rate (annual) |
|---|---|---|---|
| Wedding | 5 years | Yes, but keep angles | 8% |
| Commercial | 2 years | Yes | 12% in year 1, 3% after |
| YouTube/social | Delete after publish | N/A | <1% |
| Indie film | Forever | Yes | Hard to predict |
| Corporate | 1 year | Yes | 15% in year 1, 1% after |
When the Archive Bill Beats the Recreate Cost
Here is the test that actually settles the argument.
Take a 5 TB corporate project. Storing it on B2 at $6.95/TB is about $35 a month, or $420 a year. That's a year's archive cost.
If a client ever came back asking you to recreate that project from scratch (assuming the source no longer exists), you'd quote them at least 20 hours of edit time. At $100/hr that's $2,000.
Multiply that recreate cost by the annual revisit probability. If that number is less than $420, delete it. Corporate work in year 3 sits at maybe a 1 percent revisit rate. 1 percent of $2,000 is $20. You're paying $420 a year to insure against a $20 expected loss.
Delete it. The math is not even close.
For wedding work, flip the numbers. A 2 TB wedding archive on B2 is about $14 a year. Recreate cost from raw camera media is maybe 8 hours, so $800. At 8 percent annual revisit, the expected value of keeping it is $64. You're spending $14 to insure against a $64 expected pull. Keep it.
The calculator can help you run the actual numbers for your storage volume and provider.
The Lean Archive: Cut Before You Keep
Whatever you decide to archive, archive lean.
A typical "wrapped" project on a freelance editor's drive is 30 to 40 percent useful and 60 to 70 percent garbage you can throw away without losing anything important. Cache files, proxies, render files, and the unused 60 percent of imported media you never put on a timeline.
A lean archive has:
- Camera originals from clips that actually touched a timeline
- The final project file
- One master mezzanine export per deliverable
- One web-spec export per deliverable
- A README with title, client, date, and notes
Everything else is either regenerable or unused. Regenerable means cache and proxies (Premiere will rebuild them). Unused means the 200 clips you imported "just in case" and never dropped on a sequence.
Skip those two categories and your archive shrinks to roughly a third of the active project size. That's the difference between paying $30 a month forever and $10 a month forever, on the same project.
How to Find What You Can Cut
Stripping cache and render files is a Premiere project cleanup. Identifying unused source media is harder. You can scrub through every bin and try to remember what made it onto a sequence, or you can let a tool follow the actual edit references.
This is the workflow Clip Sweeper was built around. It reads the .prproj, walks the SubClip references back to the source media, and produces a definitive list of what's used and what's not. Before you commit to an archive decision, knowing which 60 percent is safe to leave behind makes the keep-or-delete question a lot smaller.
Decide once. Strip thoroughly. Archive what's left, lean.