DAM metadata design means deciding which fields describe each asset, which values are allowed and who maintains them. A good model has four layers: technical, descriptive, rights and product context. It embeds standard fields in IPTC and XMP, uses controlled vocabularies, and links every product asset to a GTIN or SKU.
DAM metadata and taxonomy design decides whether assets can be found, trusted and published automatically. The most reliable approach separates metadata into four layers, uses international standards for the fields that already have one, restricts free text with controlled vocabularies, and links every product asset to a product identifier. Design it before migration, because retrofitting metadata on 100,000 files is slow and expensive.
Which metadata layers does a product DAM need?
A practical model groups fields by purpose and by who maintains them. This keeps ownership clear and prevents one team from editing another team's fields.
| Layer | Example fields | Source | Owner |
|---|---|---|---|
| Technical | Format, pixel dimensions, colour profile, file size, duration | Extracted automatically | DAM system |
| Descriptive | Title, description, keywords, image type, view angle, language | Creator, AI tagging, librarian | Content team |
| Rights | Creator, copyright notice, licence type, expiry date, territories, model release | Contract, photographer | Legal or brand |
| Product context | GTIN, SKU, product family, variant, primary or secondary flag | PIM or ERP | Product data team |
| Lifecycle | Status, approval date, campaign, archive date | Workflow | DAM administrator |
The product context layer is where most product-led DAMs fail. Without a GTIN or SKU on the asset, the DAM and PIM integration depends on file names or manual linking, both of which break at scale.
Which metadata standards should a DAM support?
Use existing standards for any field they already define. Three are central for product imagery.
- IPTC Photo Metadata Standard. The current release is 2025.1, published in November 2025. It defines IPTC Core and IPTC Extension schemas covering creator, credit line, copyright notice, licensor, Web Statement of Rights, people, locations and products shown. Release 2025.1 added four properties for AI prompts and AI systems used.
- XMP (Extensible Metadata Platform). XMP is the serialisation format that embeds IPTC and custom fields inside image, PDF and video files. Part 1 is published as ISO 16684-1:2019 and was confirmed in 2024.
- GS1 Product Image Standard. Version 5.0 (July 2026) defines a file naming convention that starts with the 14-digit GTIN followed by codes for image type, facing, orientation and packaging state. These codes map directly to descriptive metadata fields.
Embedded standard fields have a direct commercial benefit. Google Images reads Creator, Credit Line, Copyright Notice, Web Statement of Rights and Licensor URL from IPTC metadata and shows them alongside the image.
How should DAM taxonomies be designed?
A taxonomy is a controlled list of permitted values. In a product DAM, most taxonomies should be reused from elsewhere in the organisation instead of being invented inside the DAM.
- Reuse the product classification from the PIM for product categories, so a search in the DAM matches the product taxonomy.
- Define a closed list for image type (packshot, lifestyle, detail, packaging, 360 spin, 3D model, video, document).
- Use the GS1 facing codes for view angle (front, left, top, back, right, bottom).
- Keep campaign and season lists time-bound, with an owner who archives old values.
- Allow free-text keywords only as a supplement to controlled fields.
Flat lists of 10 to 50 values are easier to govern than deep hierarchies. Use hierarchy only where users genuinely browse by level, such as category trees.
Who owns metadata quality?
Metadata decays without ownership. Assign each layer to a role, set completeness rules per asset type, and block publication when mandatory fields are empty. A packshot without a GTIN or licence expiry should never reach a channel.
AI auto-tagging can populate descriptive keywords quickly, but it should not write rights or product context fields without review. The trade-offs are covered in AI auto-tagging and visual search in DAM. Rights fields deserve the strictest controls, as explained in managing usage rights in DAM.
Review field usage twice a year. Retire fields with low completion and no search use, and merge synonyms in controlled lists.
Key takeaways
- Separate metadata into technical, descriptive, rights, product context and lifecycle layers, each with a named owner.
- Use IPTC Photo Metadata 2025.1 fields and embed them with XMP (ISO 16684-1) so information travels with the file.
- Link every product asset to a GTIN or SKU; GS1 Product Image Standard 5.0 naming makes this machine-readable.
- Reuse the PIM classification and closed lists for image type and view angle instead of free text.
- Keep mandatory fields minimal and block publication of assets missing rights or product links.
Frequently asked questions
What is the difference between metadata and taxonomy in a DAM?
Metadata is the set of fields attached to each asset, such as creator, GTIN, image type or licence expiry. Taxonomy is the controlled, often hierarchical list of values that some of those fields may take, such as product categories, campaign names or usage types. Metadata gives an asset its description; taxonomy keeps that description consistent across thousands of assets.
Should DAM metadata be embedded in the file or stored in the database?
Both. The DAM database holds the authoritative record and supports search and workflow. Core descriptive and rights fields should also be written into the file as IPTC fields in XMP, so creator, copyright and licence information travels with the asset after download. Product-specific fields such as GTIN can be embedded as custom XMP properties if downstream systems read them.
How many metadata fields should a DAM asset have?
Keep mandatory fields to the minimum needed for findability, rights and product linkage, typically eight to twelve. Add optional fields only when a named team or channel uses them. Every extra mandatory field slows ingest and lowers completion rates, so measure field usage after six months and retire fields nobody searches or filters on.
For a DAM vendor directory and DAM knowledge base, see the independent resource DigitalAssetManagementSystems.com.