Card Sorting

Users organize cards or topics into groups and name categories, revealing their mental model and preferred information structure for navigation design.

Process

Key Fields

Question it answers	How do users naturally group and categorize information? What naming and organization structure makes intuitive sense?
Participants & timing	20-40 participants · 20-30 min per sort · 2-3 weeks analysis
AI compatibility	AI analyzes sort patterns, builds dendrograms, and generates IA recommendations; human interpretation of ambiguous clusters is required.
Output	Dendrogram visualization, cluster analysis, naming conventions, recommended IA structure, usability test targets

Use when

Redesigning information architecture
Creating a new categorization system (catalog, help content, feature organization)
Validating a proposed IA structure before design
Quantitative data on grouping preferences is needed

Skip when

Categories are already well established and consensus is clear
Fewer than 20 cards or more than 100 cards

Common Mistakes

Not analyzing the "why"

Knowing 60% of users sorted item X into category A is useful, but understanding why informs better labeling. Always ask "Why did you put these together?" and code the reasoning.

Over-interpreting consensus

A dendrogram showing two potential groupings may indicate genuinely ambiguous information, not a single wrong answer. Both groupings might be valid; choose based on mental model data or user expertise level.

Using long or technical card labels

"Process feature for configurable workflow automation" vs. "Workflow settings" - natural language wins. Test label wording in the sort itself.