Card Sorting

How Card Sorting Works

Card sorting is used during the information architecture design phase to understand how users naturally categorize content. In an open card sort, participants receive cards labeled with content topics and freely create and name their own groups — this reveals users' mental models and preferred category names. In a closed card sort, participants sort cards into predefined categories — this validates whether a proposed navigation structure matches user expectations. Hybrid card sorts allow participants to use predefined groups but also create new ones. Card sorts are conducted with 15-30 participants for quantitative analysis; tools like Optimal Workshop's OptimalSort or Maze Card Sort collect responses and automatically generate dendrogram visualizations and similarity matrices showing which items cluster together most consistently.

Why Card Sorting Matters for B2B Marketing

For B2B website redesigns, card sorting is particularly valuable when client stakeholders disagree about navigation structure, or when existing analytics show high use of site search (a signal that users can't find things in the current navigation). A 20-participant card sort can resolve navigation debates with user data instead of internal opinion, and the resulting taxonomy is demonstrably user-centered.

Card Sorting: Best Practices & Strategic Application

Best practices include writing card labels in plain language that users will recognize (not internal product code names), keeping the card set to 30-40 items (more causes participant fatigue and noisy data), analyzing results using standardization rates (how frequently each item was grouped with the same items across all participants) rather than just majority-vote grouping, and following open card sorts with a closed sort to validate the derived taxonomy.

Agency Perspective: Card Sorting in Practice

MV3 uses Optimal Workshop for remote card sorts as part of IA redesign engagements. We typically run an open sort with 20 participants to generate a taxonomy hypothesis, then validate it with a tree test (rather than a closed sort) to measure task completion in a navigation-like structure.