A calm information architecture begins before the menu exists.
Concept of the week · Let the Structure Surface
Card sorting makes hidden mental models visible before information architecture (IA) becomes expensive to change. At the edge of attention, internal labels add extra steps, manual lookups, and decision overhead because users must translate the system before they can act.
Decision rule: Organize content only after users have shown how they expect it to belong.
Years ago, during the early design phase of a patient monitor user interface (UI), a new UX designer introduced card sorting to help us shape the IA before screens became too convincing.
We wrote clinical tasks, terms, and functions on cards, then watched users group them into piles that made sense for their daily work. The value was immediate: their hands made hidden mental models visible, and the emerging structure no longer reflected our internal component logic.
The drawback was real as well. At the medium scale, preparing the cards, running multiple rounds, interpreting outliers, and aligning stakeholders took time. But the shift was worth it—from a structure the team had to explain to one users could recognize.
Every change to the information architecture changes how people search, recognize, decide, and recover when they take the wrong path.
Ship this by versioning the card set: keep each card label, source, participant segment, grouping decision, and unresolved conflict in one decision record before changing navigation.
Card sorting turns scattered content into visible structure.
AI and Card Sorting
AI can reduce the time required for card sorting, especially when preparing labels, identifying duplicates, clustering results, and summarizing minority patterns. However, it should support synthesis rather than replace the user’s mental model.
A 2025 study on large language model (LLM)-based card sorting found useful agreement with real participant results, while also showing limits: mental-model differences, card count, and label complexity affected accuracy. For calm design, treat AI-generated categories as hypotheses. Keep outliers visible, then validate the information architecture with users through tree testing, usability testing, or task walkthroughs.
Asset & resource
Card Sort Decision Record. Use one page with five fields:
Card label: the exact wording shown to participants.
Source + segment: where the card came from and which user group sorted it.
Dominant grouping: where most users placed the card.
Outliers / unresolved conflicts: meaningful alternatives, disagreements, or split patterns.
Chosen IA label + rationale: the final category name and why the team accepted, adapted, or rejected the pattern.
Card Sorting: Designing Usable Categories by Donna Spencer
Use the book as a practical companion for planning, running, and interpreting card sorts. It supports the main idea of the episode: card sorting does not let “the user design the product” directly; it gives the team evidence about how people expect information to belong.
Light wisdom & reflection
“You can’t use it if you can’t find it.”
Which part of your product still requires people to translate your structure before they can act?
Mindful practice
For the next two minutes, look at one small area in your home where things keep getting misplaced.
What groupings do your hands already create when you are tired?
Which label, shelf, drawer, or container would reduce the tiny search before the next use?
Next time you reach for something there, notice whether the place matches the way you think in that moment.
Best,
Andreas Walden
Share this with someone designing for clarity.