Cloudinary: How to Organize Your Assets

One of the most common questions I hear from customers is deceptively simple: How should we organize our assets in Cloudinary?

Should we rely on folders or metadata? How should we handle our legacy metadata?

There’s rarely a single correct structure. Every organization has different workflows, teams, and systems interacting with its media library. But after working with many implementations, certain patterns consistently lead to libraries that scale well, remain searchable, and stay manageable over time.

When thinking about asset organization in Cloudinary, I usually look at a few key tools that serve different purposes: folders, structured metadata, tags, collections, and portals. The key isn’t choosing one over the others; it’s understanding what each one is best suited for.

Folders are usually the first place teams start when organizing assets. They’re intuitive and familiar, and they provide a clear sense of structure within the library.

In practice, folders often reflect ownership or broad organizational areas, such as business units, brands, or operational areas like e-commerce, campaigns, and corporate content.

Folders are particularly useful when assets need to be separated by responsibility or access control, or when teams want a simple way to browse the library. However, folders work best when they stay relatively simple. Deeply nested folder hierarchies can quickly become difficult to navigate and maintain.

Folders provide a strict, high-level structure and ownership. Within those boundaries, structured metadata and tags introduce the flexibility needed for real-world classification and discovery.

While folders help people navigate the library, structured metadata is the operational backbone of asset management. It drives discoverability, supports lifecycle governance, enables audit trails, and connects the asset library to external systems like CMS, ecommerce platforms, or PIMs.

Structured metadata fields allow teams to define consistent attributes that describe their assets. These fields typically reflect the information users rely on most when searching or managing content, things like asset type, category, or product line. SMD goes beyond classification. Fields like approval state or publication stage make lifecycle management explicit and trackable. Back-references to connected systems, a product ID from your e-commerce platform, or content ID from your CMS turn the asset library into an integrated part of your broader stack rather than an isolated silo. And because field changes can be logged, SMD also provides an audit record of how assets evolve over time.

Because these fields are defined and controlled centrally, they provide consistency across the library. They can also be configured as required fields, helping ensure assets are properly classified and attributed at the point of upload.

Part of the implementation process is making deliberate decisions about that legacy data: which fields are worth formalizing into structured metadata, which descriptors can remain as tags, and which information can simply be retired. This doesn’t need to happen all at once, but getting a clear plan early will usually have the biggest impact on how searchable and manageable the asset library becomes over time.

Tags provide a flexible way to enrich assets with additional descriptors.

In many implementations, the primary use of tags is AI-assisted tagging. Automated services can analyze images and videos and generate tags that describe visual elements such as scenes, objects, environments, or moods. For example, a retail brand uploading product photography might automatically receive tags of identified objects, people, and descriptors that would be impractical to apply manually at scale.

These automatically generated tags significantly improve asset discoverability without requiring manual classification.

A secondary use for tags is grouping assets for frontend delivery scenarios: identifying assets that belong to a particular experience, layout, or dynamic grouping used by applications or websites.

While tags are powerful, they typically complement structured metadata rather than replace it. Structured metadata provides controlled classification; tags add flexible descriptors that improve search and automation.

Consider a mid-size retailer managing product photography across multiple brands and seasons. A practical organizational structure might look like this:

Folders reflect ownership and operational area, for example, brand-a/ecommerce/ws25 or brand-b/campaigns. High-level, stable, and team-aligned.

Structured metadata handles everything folders can’t:

• Asset type: Product shot, lifestyle, and campaign hero.
• SKU: Product identification and PIM back-reference.
• CMS and e-commerce ID: Back-references to the CMS or e-commerce platform.
• Color, material, and year: Classification attributes.
• Approval state: Lifecycle tracking from agency delivery to publish-ready.
• Channel: Web, mobile, print, and social.

Tags are applied automatically via AI at ingest; they identify objects, scenes, or environments in product and lifestyle imagery, making assets discoverable without manual effort.

Tags can also be derived from PIM attributes at upload, for example, a tag based on the SKU value to enable dynamic asset grouping and context-aware delivery by product.

In this setup, when an agency uploads a new product shoot, SMD fields populate automatically from PIM attributes, as described in the Cloudinary Architecture Blueprint. The folder tells you who owns it. The metadata tells you what it is, where it is in its lifecycle, and what system it belongs to. The tags make it findable.

Cloudinary also offers collections and portals, a second layer that builds on your organizational foundation and becomes increasingly valuable as libraries grow and teams expand.

Collections allow teams to group assets based on how they are used rather than how they are classified, e.g., a campaign launch kit, a set of homepage assets, or a curated seasonal bundle. The key advantage is that collections can bring together assets from multiple folders or metadata categories without changing their underlying organization. They’re often the easiest way for teams to curate sets of assets for specific workflows, reviews, or distribution.

Portals take this further by making curated content available to external audiences, like agencies, creative partners, franchisees, or regional teams, and without exposing the entire asset library. Organizations can present a clean, branded interface focused on exactly what external users need, while maintaining full governance internally.

Organizing assets in Cloudinary isn’t about choosing between folders, metadata, tags, or collections. Each of these tools solves a different problem.

Folders structure the library and reflect ownership. Structured metadata defines how assets are classified, governed, and connected to the rest of your stack. Tags enrich assets with AI-generated descriptors and flexible grouping options. Collections and portals help teams use and share those assets effectively.

If you’re running a small operation with limited assets and a single team, a simple folder structure combined with a handful of well-chosen SMD fields may be all you need. If your library spans multiple brands, business units, or external stakeholders, or if you’re migrating from a legacy system with years of accumulated metadata, the organizational decisions become significantly more complex. In those cases, getting expert eyes on it early, whether from your own solutions architect or Cloudinary’s Professional Services team, is worth the investment. Getting the structure right upfront is far easier than untangling it later.