Learn how to run card sorting sessions to design intuitive navigation and content structures. Open vs closed methods, remote tools, analysis techniques, and common mistakes.
Card sorting is a research technique where participants organize content into categories that make sense to them. It is one of the oldest and most reliable methods for designing information architecture: the structure and labeling of websites, applications, and other information systems. The technique works because it reveals how real users think about your content, rather than how your organization thinks about it.
Most navigation problems stem from the same root cause: the people who designed the structure organized it around internal logic (departments, product lines, technical categories) rather than user logic (tasks, mental models, goals). Card sorting closes this gap by letting users create the structure themselves.
A university website organized its content by administrative department: Financial Aid, Registrar, Student Affairs, Academic Services. Students searching for "how to drop a class" did not know (or care) which department handled that task. Card sorting with students revealed that they thought about the website in terms of lifecycle stages: Applying, Enrolled, Graduating, Alumni. Reorganizing around these mental models reduced support ticket volume for navigation-related questions by over 30%.
Card sorting is particularly valuable during the Define stage of design thinking, when you are translating user research into structural decisions, and during the Prototype stage, when you need to validate that your proposed structure makes sense to users before building it.
In an open card sort, participants receive a set of content cards (each card represents a page, feature, or piece of content) and create their own categories. They group the cards however they see fit and label the groups themselves.
Use open card sorting when you are starting from scratch and need to discover how users naturally categorize your content. The output is a set of user-generated categories and groupings that inform your initial information architecture.
The trade-off: open card sorting produces the richest insights but is harder to analyze because every participant may create different categories with different labels. With 20 participants, you might get 15 different organizational schemes. The patterns within this variety are the insights.
In a closed card sort, you provide pre-defined categories and ask participants to place content cards into those categories. The categories are fixed; participants only decide which card goes where.
Use closed card sorting when you already have a proposed structure and want to validate whether users can find content within it. The output tells you which categories are intuitive and which cause confusion.
The trade-off: closed card sorting is easier to analyze (you can calculate agreement percentages per card) but may miss cases where your categories themselves are the problem. If your categories do not match users' mental models, a closed sort will show confusion but will not tell you what the categories should be instead.
A hybrid sort provides pre-defined categories but allows participants to create new ones if none of the existing categories feel right. This approach captures the best of both methods: you learn whether your proposed structure works and you discover where it does not.
Select 30 to 60 content items for the sort. Fewer than 30 does not provide enough complexity to surface meaningful patterns. More than 60 creates fatigue. Write each item on a card using the language users would recognize, not internal jargon.
Avoid biasing the sort with your card labels. "Employee Benefits Portal" nudges participants toward an HR category. "Health insurance, retirement, paid time off" describes the same content without suggesting an organizational home.
Use physical index cards on a large table. This works best with 5 to 8 participants per session (though they sort individually, not as a group). Allow 30 to 45 minutes per participant. After sorting, ask participants to explain their groupings. The rationale is often more valuable than the groupings themselves.
Ask follow-up questions: "Was there any card you were not sure about?" and "Were there any groups that felt like they did not quite fit?" These questions surface the edge cases that reveal structural weaknesses.
Tools like Optimal Workshop, UserZoom, and Maze offer digital card sorting interfaces. Remote sessions scale better (you can run 50+ sorts) and participants can complete them at their own pace. The trade-off is that you lose the opportunity for follow-up questions unless you add a post-sort survey.
For remote sorts, aim for 30+ participants to generate statistically meaningful patterns. With in-person sorts, 15 participants is typically sufficient because you gain qualitative depth from the conversations.
A similarity matrix shows how often each pair of cards was placed in the same group. If 85% of participants put "reset password" and "change email" in the same category, those items belong together in your navigation. If "billing history" is split evenly between "Account" and "Payments," you have identified a structural decision that needs additional research to resolve.
A dendrogram is a tree diagram that shows how cards cluster together based on how frequently participants grouped them. Cards that cluster tightly should be near each other in your navigation. Cards that only cluster at a high level might belong in different sections.
For open sorts, look at the category labels participants created. Group similar labels together. If participants call a category "Settings," "My Account," "Profile," and "Preferences," they are describing the same concept with different words. The most commonly used label is usually the best candidate for your navigation.
Cards that participants consistently struggled to categorize (placed in many different groups across participants) indicate content that does not fit cleanly into any single category. These items may need to appear in multiple places (cross-linking) or may indicate that your content itself needs restructuring.
Card sorting results do not automatically produce a navigation structure. They provide evidence that informs your structural decisions. The translation process involves:
Tree testing (also called reverse card sorting) validates a proposed navigation structure. You give participants a text-only version of your navigation hierarchy and ask them to find specific items. "Where would you go to reset your password?" If most participants navigate to the right place, your structure works. If they consistently go to the wrong category first, you have a labeling or placement problem.
The ideal workflow is: open card sort (discover user mental models) followed by tree test (validate your proposed structure). This two-step process produces navigation that is both user-informed and empirically validated.
Card sorting fits naturally into the design thinking workflow:
Card sorting reveals how users naturally organize information, but the real work begins when you translate those patterns into a navigation structure and test whether it actually works. User testing validates that your information architecture holds up under real task pressure, while affinity diagramming offers a complementary technique for organizing qualitative research data using similar clustering principles. Once your IA is defined, rapid prototyping lets you test navigation flows before committing to full implementation, and grounding the entire structure in accessibility-first principles ensures that your information architecture works for everyone, not just the participants in your card sort.
Related guides: wireframing techniques · ab testing design thinking · rapid prototyping
Design Thinker Labs Home · All Guides · How It Works · Pricing