Digital Asset Management Taxonomy Template
Four documents and three sheets that trace every crop to its licensed original, so a rights exposure shows up by asset, not by expiry date.
Free download · No account needed
Asset Register · Marlowe Athletic · library reconciled 2 September 2026
41,208 files. 6,842 real assets. Two of them are live past their licence.
| Source asset | Origin | Derivative files | Live placements | Rights state |
|---|---|---|---|---|
| SRC-04560 | Owned | 214 | 61 | Owned |
| SRC-00418 | Stock licence | 31 | 4 | Expired |
| SRC-02904 | Stock licence | 9 | 1 | In term, outside scope |
What each row means
| Finding | Detail |
|---|---|
| Wordmark cluster | 214 files trace to one canonical lockup. 198 are unsanctioned re-exports nobody can find by searching, so they get remade instead |
| Editorial licence, commercial use | SRC-00418 is licensed for editorial use only. Its licence expired 3 May 2025 and it is still running in 4 paid placements |
| In term, wrong territory | SRC-02904 is licensed for North America only, inside its term until November 2026, and live on the EU storefront today |
Nothing about the third row is visible from an expiry date. The licence has fifteen months left. The exposure is the territory field, joined against where the asset is actually running.
Every digital asset management taxonomy guide online covers the same three fields: subject, campaign and file format. They make a library searchable, which is half the job description a rebrand or a licence review actually needs. The other half is provenance: which file was cut from which original. Without it, hash-based deduplication misses every crop and resize that is not byte-identical to its source, and a licence attached to an original never propagates to what a designer exported from it eighteen months later.
One field carries the whole mechanism: source_asset_id, the original a file was cut from. Group files by it and duplicate counts collapse to real assets rather than files; propagate a licence's scope and expiry through it and every crop inherits what its source was actually granted, not just when that grant expires. The IPTC Photo Metadata Standard already defines rights fields built to travel inside the file itself: a copyright owner, rights usage terms and a web statement of rights, so a crop exported next year carries them instead of arriving blank.
At Marlowe Athletic, a fictional apparel brand with eleven years of library, 41,208 files collapse to 6,842 source assets once provenance is resolved, 61 per cent of the export by file count. Two hundred seventy of those source assets are expired or expiring with no live use and retire outright; forty-one more are still live in a placement today. Pair it with the rebrand rollout pack for the surface-by-surface migration this library feeds, and the content inventory template for the same collapse-and-retire mechanism applied to a website instead of a library.
What you get
Taxonomy Standard
The full field list split into descriptive, structural and rights classes, with source_asset_id named as the one field that carries both the duplicate count and the rights exposure.
Naming Convention
A filename pattern built around the source asset id, so provenance survives an export even after the file leaves the DAM and its metadata behind.
Lifecycle Policy
Five computed rights states and four distinct retirement mechanisms, including why collapsing a duplicate is never the same operation as retiring an expired licence.
Rights Exposure Method
How a licence, its granted scope and a list of live placements join into one exposure report, catching an in-term asset running outside its own grant. A modernized brand guide checks the same kind of mismatch, between what an inherited PDF actually specifies and what a channel ships.
Asset Register
One row per source asset with its origin, scope, live placement count and computed rights state, expanding into every derivative file that traces back to it.
Duplicate Report
One row per provenance cluster rather than per file, naming the canonical survivor and every near duplicate a hash-only check would have missed.
Retirement Queue
Every file leaving the library, batched by mechanism with an owner, a due date and a verification step that checks the actual outcome instead of a closed ticket.
How it works
- 1
Send the export
A DAM export or folder tree, licence and invoice records if they live outside the DAM, and a list of where assets are currently running.
- 2
River resolves provenance
Files group by the original they were cut from, not by hash alone, so a resize or a retouch pass collapses to its real source asset.
- 3
Rights join against live use
Licence scope and expiry join against current placements, surfacing an in-term asset running outside its own grant, not only what has already expired.
- 4
The queue ships batched
Duplicates and expired assets batch by retirement mechanism, each with an owner and a verification step, so the library ends smaller and current.
Frequently asked questions
Is it free, and what format are the downloaded files?
Free to download with no account required. The documents are .docx and every sheet is a .csv, so the register, the duplicate report and the queue all open directly in whatever office suite or spreadsheet tool the team already uses.
What does Edit with AI actually do?
It turns this pack into a private Space and River works through your own library export inside it: resolving provenance, filling in the Asset Register, joining your licences against current live placements, and batching Retirement Queue. Everything shown here is the worked example; yours replaces it.
Our DAM already has a taxonomy. Why add this on top?
Most DAM taxonomies stop at subject, campaign and format. This adds source_asset_id, the structural field most platforms leave optional, plus rights fields split by media and territory rather than one cleared checkbox. It works alongside an existing taxonomy rather than replacing it. A crawl already run for a technical SEO audit doubles as one more discovery source for images actually live on the site.
How does this find duplicates a hash check misses?
A hash check only matches byte-identical files. Provenance groups files by visual and structural similarity to their original, which is what catches a crop or a resize: not identical to its source at the byte level, but clearly cut from it. On the worked example, provenance found 280 near duplicates across five clusters that hashing alone missed.
What if I don't have licence records for older assets?
Register rows without a confirmed licence get marked rights_state unknown rather than assumed fine, and are listed separately so they can be chased down against whatever invoice or contract records exist. An unknown state is still more useful than a checkbox that assumes coverage.
Can this catch an asset that's still within its licence term?
Yes, and that is the exposure a date-only review misses. Scope is tracked separately from expiry, so an asset with a year left on its licence can still be flagged for running in a media type or territory outside what was actually granted.
Find out how much of your library is a duplicate, and what's still running past its licence
Send your export. River resolves provenance, joins your licences against where assets are actually live, and hands back a retirement queue batched by mechanism.
Edit with AI