Card sorting checklist: before, during, and after the study

This checklist covers the full card sorting process from preparation through analysis and follow-up validation. Use it to track progress and ensure nothing is skipped — especially the pilot test and the follow-up tree test, which beginners most often omit.

Before the study

Define the research question: are you generating a new architecture (open sort) or validating an existing one (closed sort)?
Select 30-50 cards representing the content, features, or topics to be sorted
Write card labels in plain user language — no jargon, no internal terms, no duplicate lead words
For closed sorts: define the predefined categories and verify they are mutually exclusive and collectively exhaustive
Set up the study in OptimalSort, UXtweak, Maze, or prepare physical materials for in-person sessions
Write clear instructions for participants (what to do, how long it takes, that there are no wrong answers)
Run a pilot with 3-5 people and adjust cards, labels, and instructions based on feedback
Recruit 15-30 participants who represent the target user population

During the study

For remote sorts: monitor completion rates and quality (flag sorts completed in under 2 minutes)
For in-person sorts: ask participants to think aloud and explain their groupings
Record the reasoning behind groupings, not just the groupings themselves
Note cards that participants hesitate on or ask questions about — these signal labeling problems

After the study

Remove incomplete and low-quality sessions from the dataset
Standardize group labels (map synonyms to canonical terms)
Generate similarity matrix and dendrogram
Identify strong clusters (60-70%+ agreement)
Document orphan items (cards with no consistent group placement)
Draft a proposed category structure based on the analysis
Follow up with a tree test to validate findability in the proposed structure
Present findings with the dendrogram, recommended structure, and items needing cross-linking
Archive raw data for future comparison if the sort is repeated after a redesign