DAM fundamentals
How to design a DAM metadata model and taxonomy
Quick answer
A DAM metadata model defines which fields describe every asset, which values are allowed and who fills them. Design it in four layers (administrative, descriptive, rights and technical), keep mandatory upload fields to 6 to 10, use controlled vocabularies for anything people filter on, and map fields to IPTC and XMP for portability.
Key takeaways
- Split the model into four mutually exclusive layers: administrative, descriptive, rights and technical metadata.
- Keep mandatory upload fields to 6 to 10; every extra required field lowers completion quality (practitioner range).
- Use flat facets with at most three hierarchy levels; deep folder-style trees slow search and confuse uploaders.
- Map fields to the IPTC Photo Metadata Standard (version 2025.1) and XMP (ISO 16684-1) so metadata survives export and migration.
- Store controlled values as language-neutral IDs with translated labels to support multilingual European teams.
A good DAM metadata model is small, layered and controlled. Define four layers of metadata, make only the fields that drive search and governance mandatory, and govern their values with faceted vocabularies mapped to IPTC and XMP standards. This design determines whether users find assets in seconds and whether your data survives the next migration.
The four layers of a DAM metadata model
Every DAM field belongs to exactly one of four layers, and each layer has a different owner. Separating them keeps upload forms short, because users only fill the descriptive and rights fields that systems cannot derive.
| Layer | Purpose | Example fields | Filled by |
|---|---|---|---|
| Administrative | Identify and track the asset | Asset ID, status, owner, upload date, version | System and workflow |
| Descriptive | Make the asset findable | Asset type, product SKU, brand, campaign, market, keywords | Uploader, AI tagging, librarian |
| Rights | Control legal use | Copyright owner, licence type, expiry, territories, model release | Uploader, legal |
| Technical | Describe the file | Format, dimensions, colour profile, duration, codec | System extraction |
The so what: when a requirement document mixes these layers, upload forms grow to 30 fields and completion quality drops. As a practitioner range, keep mandatory upload fields to 6 to 10 (editorial estimate).
Taxonomy design: use facets and shallow hierarchies
Faceted taxonomies outperform deep hierarchies because users filter on several independent dimensions at once. A manufacturer might define six facets: product line, asset type, usage channel, market, language and lifecycle status. Combining three filters across six facets narrows a 50,000 asset library faster than any folder tree.
Keep hierarchies within a facet to three levels at most, for example "Packaging > Primary packaging > Front of pack". Add synonyms in the vocabulary so that "packshot", "product shot" and "cut-out" return the same results. Platforms differ in schema flexibility: products such as Picturepark and Hyland Nuxeo let administrators model custom content types and relations, which matters when assets need richer structure than flat fields.
Map metadata to IPTC, XMP and Dublin Core
Standards make metadata portable. The IPTC Photo Metadata Standard, at version 2025.1 since November 2025, covers descriptive, rights and AI-related properties, and since version 2023.1 it includes a Data Mining property for text and data mining permissions. XMP, standardised as ISO 16684-1, is the container that embeds these properties in JPEG, TIFF, PDF and many video formats. Dublin Core (ISO 15836-1:2017) offers 15 generic elements that most DAMs map to.
The implication is practical. Embedded metadata travels with the file to agencies and CMS platforms and survives a vendor change. Map every field to a standard property or a custom XMP namespace before go-live, and test export round trips as part of your DAM migration plan.
Multilingual metadata for European teams
Multilingual libraries need controlled values stored as language-neutral IDs with translated labels. A brand operating in Germany, France, the Netherlands and Poland can then show "Verpackung", "Emballage", "Verpakking" and "Opakowanie" for the same value, and a search in any language returns the same assets. Free-text fields such as captions need a separate policy: translate on demand, or keep them in one working language.
Product assets need a shared key with the PIM, usually the SKU or GTIN. Agree that key before modelling, as described in the PIM and DAM integration patterns.
Govern metadata quality with measurable KPIs
Metadata quality decays without an owner and a metric. Track three KPIs monthly: completeness of mandatory fields (target above 95 percent), share of searches returning zero results (target below 5 percent) and vocabulary terms unused for 12 months (candidates for removal). These targets are practitioner ranges.
Treat AI auto-tagging output as suggestions for descriptive keywords, and keep rights fields under human control. Assign a named metadata owner within your DAM governance model and review the vocabulary each quarter.
What to do next
- List the top 20 searches users run today and derive your facets from them.
- Sort every proposed field into the four layers and cut mandatory upload fields to 10 or fewer.
- Map each field to an IPTC, XMP or Dublin Core property, or to a custom namespace.
- Define language-neutral IDs and translated labels for every controlled vocabulary.
- Set completeness and zero-result targets and name the metadata owner before go-live.
Frequently asked questions
What is the difference between metadata and taxonomy in DAM?
Metadata is the set of fields attached to each asset, such as title, product SKU, licence expiry or campaign. A taxonomy is the controlled, structured list of allowed values for a field, such as the official list of product categories or markets. The metadata model defines the fields; the taxonomy governs their values.
How many metadata fields should a DAM have?
Most mid-market DAMs work well with 20 to 40 fields in total, of which only 6 to 10 are mandatory at upload (editorial estimate). System fields and AI tagging can fill many of the rest automatically. Fields nobody searches or filters on should be removed at the next review.
Should DAM metadata follow IPTC standards?
Yes, for photographic and editorial assets. Mapping your fields to IPTC and XMP lets metadata travel embedded in the file to agencies, CMS platforms and future DAMs. Custom business fields such as SKU or campaign code can sit in a custom XMP namespace.