See How Users Organize Information
Navigation fails when it follows team logic instead of user logic. Labels feel clear in a sitemap but collapse when real users can’t find what they need. Menus get cluttered. Categories feel off. Users bounce.
Card Sorting cuts through those assumptions. It shows how people naturally group and label information, giving you evidence to structure navigation in a way that matches their mental models.
Why Teams Miss It
-
Copying competitors’ IA instead of testing their own
-
Relying on internal assumptions of “what makes sense”
-
Running the exercise but never quantifying the results
Card Sorting avoids those mistakes by turning subjective choices into measurable signals.
Why It Matters
Card Sorting connects directly to UX metrics in Glare:
-
Comprehension: do users understand labels and categories
-
Efficiency: how quickly they group items without confusion
-
Success Rate: whether groups align with expected navigation goals
-
Desirability: which labels feel most natural or appealing
These signals give you the clarity to build structures that match how users actually think.
How It Works
Raw card placements become UX metrics in Glare by analyzing consistency and accuracy. High agreement rates show Comprehension. Clear clusters signal Efficiency. Misplaced items expose navigation risks that lower Success Rate.
Preferred labels reveal Desirability. Small tests can show patterns in hours, while larger studies with 30–50 participants reveal stable signals.
In a Card Sort, participants group topics, features, or content into categories that make sense to them, often with the option to create their own labels.
**
Steps**
-
Define the scope: pick a set of items such as features, help topics, or product categories
-
Choose the type: open card sort (users create groups), closed card sort (predefined groups), or hybrid
-
Recruit participants: aim for 15–30 to see reliable patterns
-
Run the sort: online tools or moderated sessions work equally well
-
Analyze clusters: look for consistent groupings, disagreements, or outliers
-
Translate into structure: apply insights to menus, navigation, or IA and retest
The signal comes from comparing the user’s mental model with the one your design assumes.
Example in Action
An e-commerce site ran an open card sort on product categories. Users grouped “yoga mats” with “fitness gear” instead of “home accessories.” Realigning the navigation cut bounce rates by 18 percent and boosted conversions on workout products.
One signal reshaped the shopping flow.
Best Practices
-
Use open sorts early to explore mental models, and closed sorts later to validate IA
-
Segment participants by audience type to spot differences in logic
-
Pair with First Click Testing to confirm new navigation works in practice
-
Visualize results with dendrograms or similarity matrices to make insights clear for teams
When to Use
-
Early in design to build navigation from user logic
-
Before relaunching menus or site structure
-
To validate whether new labels or categories are clear
-
Alongside comprehension or first-click studies to ensure clarity
Card Sorting makes invisible mental models visible, so you can build navigation that works the way users think.
Brief History
Card Sorting originated in cognitive psychology and information science as a way to study categorization. By the 1980s, HCI researchers were using it to evaluate menu design and label clarity. With the rise of online tools in the 2000s, card sorting scaled from lab studies to remote unmoderated testing.
Today, it is widely used in IA, e-commerce, SaaS, and enterprise products to create structures that reflect user thinking rather than internal assumptions.
Run Card Sorting in 24 Hours
-
Pick 20–30 items that matter most
-
Choose open or closed sort based on your goal
-
Recruit at least 15 users
-
Run the sort with an online tool
-
Analyze clusters for agreement and confusion
-
Apply findings to your IA and retest
Try It Now
Choose one part of your navigation this week. Run a card sort with real users. If items scatter into inconsistent groups, you have proof the structure is unclear, and a signal to redesign.