How to organize design assets so your team can actually find them

Ask a marketing lead to find the webinar banner from last quarter. Watch what happens. They open the shared drive, scan a folder named after a campaign that has since been renamed, open Figma, scroll through twelve files called “Marketing 2026,” then ask the designer — who left the company in March.

The failure isn't laziness. It's that most teams organize assets the way the creator thinks, at the moment of creation: by project, by quarter, by whoever owned the work. But retrieval happens along different dimensions — someone searches by product (“the pricing page visuals”), by format (“a 1200×630 social card”), or by style (“that dark illustration set”). A folder structure can encode exactly one of those dimensions. Every other query fails.

So the fix isn't better folders. It's a structure where one asset can answer many different questions at once.

A library that holds everything holds nothing. Before structure, draw the line: the library stores finished, approved, reusable work — not working files, not explorations, not the 47 rejected logo variants.

A reasonable starting inventory for a B2B SaaS team:

• Brand assets — logos, lockups, color and type references
• Campaign deliverables — banners, social cards, email headers, landing-page heroes
• Product UI — approved screens, feature visuals, screenshots of shipped work
• Templates — deck layouts, one-pager skeletons, ad-size sets
• Illustrations and icons in their final exported form

Working files stay in Figma. The library holds the artifact plus a link back to its source file, so anyone can trace a final asset to the frame it came from. That separation is what keeps the library scannable — every item in it is usable as-is.

The counterintuitive move: delete the folder tree. One flat library per brand, where every asset carries metadata instead of a location — an asset type, a few tags, a date, a source link.

Hierarchy fails because assets are multi-dimensional. A hero illustration for the Q1 launch of your analytics product belongs simultaneously to “illustrations,” “Q1-2026 launch,” and “analytics.” A folder forces you to pick one and orphan the rest. Tags let the same asset surface in all three queries.

Flat also removes the most expensive decision in any filing system: where does this go? When the answer is always “the library, with three tags,” importing takes thirty seconds and people actually do it. Filing friction is the silent killer of every asset system — the moment adding an asset requires a judgment call, assets stop being added.

Tags fail in one of two ways: too few (nothing is findable) or too many (every filter returns noise). The discipline that prevents both is a small set of categories with a capped vocabulary.

Choose 3–4 categories

Categories mirror how your team actually asks for things. For most B2B SaaS teams that's product (“Pricing page,” “Mobile app”), campaign (“Q1-2026 launch,” “Black Friday”), brand (“Enterprise,” “SMB”), and a custom category for whatever doesn't fit. Asset type — banner, deck, illustration — should be a first-class field, not a tag.

Cap the vocabulary at 5–15 tags per category

When a category grows past fifteen tags, the filter dropdown becomes a second search problem. Merge near-duplicates quarterly (“Q1 launch” + “Q1-2026 launch” → one tag), and datestamp campaign tags from day one so they sort and expire naturally.

Tag at import, never “later”

“Later” is where tagging goes to die. Make tagging part of the import gesture — three to six tags, one per relevant category — and the cost rounds to zero. Tools that auto-tag on import remove even that step; review the suggestions instead of typing from scratch.

Tags handle filtering; titles handle the last ten feet — the moment someone scans a grid of results and picks one. A title that works reads naturally at a glance:

Pricing page — Hero illustration — Dark

The pattern is what it's for, what it is, what makes it different. Compare that to final_v2_export_FINAL(3).png, which encodes a revision history nobody needs and answers none of the questions a searcher has. Don't write a twelve-rule naming convention document — write the pattern once, rename the ten most-used assets to match, and the example propagates better than any policy.

Tags describe what an asset is. Collections express what a group of assets is for. The distinction matters because the two decay differently: descriptions stay true forever, purposes expire.

Collections that earn their place:

• A campaign kit — every approved deliverable for the Q1 launch, shared with the whole go-to-market team
• A sales enablement set — the twelve visuals the sales team is allowed to drop into decks
• A brand-refresh moodboard — references collected during an exploration, archived when it ships

Keep the count low — five to fifteen live collections, not fifty. If a collection's membership is rule-based (“everything tagged Q1-2026 launch”), use a smart collection — a saved query that updates itself — instead of curating by hand. Manual curation is for judgment; queries are for criteria.

Here's the test of everything above: a teammate who has never seen your library types two words and finds the asset in under thirty seconds. If people are browsing folders or scrolling grids to find things, the structure has already failed — browsing is what retrieval looks like when search has nothing to work with.

Search quality is downstream of the habits in sections 04 and 05: descriptive titles and consistent tags are exactly what full-text search indexes. A bilingual team should also check that search handles both languages — a French marketer searching “bannière” should find what the designer titled in English, which is where semantic search starts paying for itself over plain keyword matching.

Then train the entry point. When someone asks for an asset in Slack, answer with the library search that finds it, not with the file. It feels pedantic for two weeks; after that, the Slack questions stop.

Every asset system decays toward entropy at the speed of your busiest week. The countermeasure is small and boring: one named owner — usually a senior designer or whoever plays design ops — and a recurring 15-minute monthly pass:

• Merge duplicate tags and retire ones with one or two assets
• Archive collections for campaigns that have shipped
• Spot-check the ten newest imports for titles and tags

And make the system part of onboarding: a new designer or marketer learns to import and tag in week one. The library compounds — every well-tagged asset makes the next search better — but only if everyone deposits into it.