Designing Discovery for 100 Years of Creative Work
Redesigning a 100-year creative archive to lift task completion from 27% to 73%.
The Challenge
Emily Carr University holds over 100 years of creative work: graduate theses, paintings, films, installations, critical writing. Students came looking for inspiration and most left empty-handed. The system had been built for academic publishing, and creative research doesn't fit that mold. A migration to Drupal opened a rare window to rebuild how people actually find things instead of just where those things get stored.
How do you make a hundred years of wildly different creative work discoverable to someone who doesn't know what they're looking for yet?
Company
Emily Carr University of Art & Design
Timeline
2024
—
2025
Role
Product Designer · Information Architect · Discovery & Navigation · Taxonomy & Metadata · Complex Systems Design
Discovery Research
Before designing anything, I needed two things: a sense of how other institutions had tackled this, and a clear picture of how researchers actually behaved inside the existing system.
User Research
20
Graduate students & faculty interviewed
Vancouver-based researchers who actively use the collection.
Environmental Scan
Built for the library, not the researcher.
Built around the library's needs rather than the researcher's. I benchmarked digital collections across Canadian universities and the pattern held everywhere.
The way researchers described their process (following a feeling more than a keyword) pointed toward systems outside academia.
Alongside the collection, participants were using Spotify, Pinterest, and Notion every day, systems solving a similar discovery problem at a much bigger scale. Spotify picks up on taste you never stated out loud. Pinterest connects a vague hunch to actual content. Notion holds all kinds of knowledge together without cramming it into rigid categories.
Across all three, the pattern was the same: the best discovery systems meet users inside their own mental models instead of forcing them into the system's. That became the lens for everything that followed.
System
Spotify
Are.na
JSTOR
Notion
ECU Redesign
Discovery mode
Passive + Active
Browse first
Collection based
Search only
Structure as needed
Browse + Search Seperated
Handles Ambiguity?
Yes, taste graph replaces rigid genre labels entirely
Yes, designed for users who can't articulate what they want
Yes, user-generated taxonomy, no fixed categories
No, assumes you know discipline,
journal, date
Yes, user applies structure, system doesn't impose it
Yes, 8 category structure limited to old conventions of discipline
Relationship Mapping
Tracks connect through listening behaviour, not metadata
Pins link to boards, boards link to people and themes
Channels connect any content across media
None, works exist in isolation
Blocks link across pages and databases
Tracks connect through metadata
User Mental Model
Mood, Listening, History, Vibe
Visual instinct & accumulated interest
Ideas that accumulate, not document types
Known-item seeking only
Heterogeneous knowledge that resists fixed shapes
Themes, materials, methods, people
Design Lesson for ECU
When categories fail, relationships takes over. Add a Related works module.
Browse and search must be separate. A researcher exploring is in Pinterest mode. A researcher with a name is in Google mode.
Creative researchers think in ideas across media. Materials + Methods fields move in this direction.
The academic standard that graduate students actively avoid exploration.
Rigid metadata fields that force creative works into predefined shapes fail the same way MODS failed. Schema flexibility in Drupal is the version of Notion's block system.
Synthesizes the best pattern from every single system above. Meets researchers in their own language, not the systems.
Discovery patterns across six systems, and what they mean for the ECU redesign.
Three failure patterns surfaced in every conversation
Organizational Systems
Topic was supposed to be the main entry point, but users searched by author instead. The design taxonomy was incomplete, so subject-based exploration rarely worked.
I always search by name, but half the time nothing comes up.
Navigational Systems
The schema had no relationship fields at all. Users would find one relevant work and hit a wall, with no way to follow a thread to related people or projects.
There was no way to see what else that person had done.
Labeling Systems
The same concept showed up under different labels depending on the record, so search terms returned inconsistent results.
I searched 'textile' then 'textiles' and got completely different results.
Core Insight
The system was built around the wrong question.
Librarians asked: "What is this work?" Researchers were asking: "What work feels like mine?"
Researchers weren't searching by classification. They were searching by resonance, by what a work touches, who made it, what ideas sit next to it. The old system had no way to answer that. It described objects, while researchers were after connections.
That became the principle behind every decision that followed: Stop describing works. Start connecting them.
Design Principles
Before touching the architecture, I turned that insight into four constraints, not values. Each one came from something broken in the old system, and each became a test: does this decision honor the principle, or compromise it?
Speak the
researcher's
language
Connections
over
containers
Built to be
maintained
Flexible enough
to hold what
doesn't fit
Site Map
With the principles set, the next step was mapping the full structure they'd need to support, meaning both what content existed and how every piece connected to every other piece. The sitemap shows the shift. The old structure was flat: 60-plus standalone subject tags, and no Creator, Material, or Method fields tying records together.
The redesigned structure adds those fields as first-class connectors. Every thesis record now sits inside a web of related people, materials, and ideas instead of dead-ending in a category.
Content Model
I built the content model diagram to map relationships between content types. The full taxonomy vocabulary and field-level audit came out of collaboration with the content strategist.
With the taxonomy restructured, the next question was how content types related to each other, not just what they were. The navigational failure pattern, users hitting a dead end at every record, turned out to be a schema problem more than a UI problem. The content model was built to fix that at the data level.
Information Architecture
To restructure the subject taxonomy, I ran card sorting sessions with graduate students, asking them to group 40 subject terms into categories that felt natural. The guiding question for each term: what does this mean in the context of your research?
One example: figuring out whether "Textiles" belonged under Materials & Making, Identity & Culture, or Social & Political. The clusters that emerged from these sessions became the eight parent categories in the redesigned taxonomy.
Scroll Inside the box to view complete design.
Annotated Wireframes
With the architecture defined, I built wireframes to test how the browse and search flows would actually work, and to test the core hypothesis behind the whole project: that restructuring entry points around mental models instead of metadata categories would cut down on abandonment.
Browse Flow
The browse flow was built for the researcher who shows up without a known destination, someone working from a theme, a feeling, a material, rather than a title or an author. Most users entered through the Topic facet but left without clicking a single result.
The research behind this: users faced with a flat list of 60-plus terms spent an average of 40 seconds scanning before giving up.
Grid clusters reduced cognitive load and gave users a recognizable entry point.
Search Flow
The search flow was built for precision, for an advisor looking up a specific student, or a researcher who already has a name in mind.
The research behind this: author name was the most common search query, and there was no Creator field to support it.
The search experience was redesigned around that reality. Creator now surfaces as the first filter instead of getting buried in an advanced options panel.
Record View
The two flows converged at the record view, the individual thesis page. This turned into the most important screen in the system, the place where the core insight had to show up loudest: stop describing works, start connecting them.
The research behind this: users who reached a record either got what they needed right away or gave up completely. There was no next step in between. Two design decisions responded directly to that dead end.
The Related Works module turns every record into a new entry point through shared tags. Instead of ending the journey, each thesis page becomes a continuation, a way to keep moving through the collection without going back to search. This was the single most important structural change in the whole system, and it solved the navigational failure at the UI level.
The clickable Author field pulls up every thesis by that creator from a single name. Find one work by someone and their whole body of work opens up.
High Fidelity Prototype
With the interaction patterns validated in wireframe, the high-fidelity prototype brought the whole visual system to life. Typography, imagery, and spacing were tuned to make dense archival metadata feel browsable instead of clinical.
Built across desktop and mobile with working filters and a fully navigable grid, it turned every wireframe decision, breadcrumbs, filter chips, the Related Works module, into a real, clickable interface. This was the version tested with 20 participants in the next phase.
Testing & Outcomes
I ran user testing with 20 participants, a mix of graduate students and faculty. Each session ran 90 minutes, and participants worked through three tasks: find a thesis by a specific creator, explore works related to a given theme, and get from a single record to a connected work without going back to search.
Usability Testing Results
73
%
task completion
vs 27% on the old system
New
Old
73%
27%
Participants who completed the browse task
40s
20s
average time spent scanning the subject taxonomy before selecting
a category with the new grid format
of the participants used the Related Works Modules without being prompted
"
I didn't feel like I hit a wall this time - I just kept finding more things
⸻ Testing Participant
Reflection
This project started as a migration. It ended as a rethinking of what discovery means in a creative archive.
The old system asked what a work is. The new one asks what it connects to. That shift, from description to connection, reshaped the schema, the taxonomy, the navigation, and ultimately the experience of everyone who comes into the collection looking for something they can't yet name.
What I would do differently?
The taxonomy was rebuilt collaboratively through card sorting, but the maintenance model, how library staff keep it up over time, was documented and never actually tested with the people who'll use it. The next phase of this work should be a staff-facing audit workflow that keeps the taxonomy from drifting back into the mess we started with.
What remains open?
Personalization. The system surfaces connections between works now, but it doesn't learn from individual researchers yet. That's the next layer of this problem.
