A card sort is only as good as its stimulus cards:
Granularity: Aim for 30 to 50 cards. Fewer than 20 cards fails to surface organizational depth; more than 60 cards induces participant fatigue.
Avoid Category Leaks: Do not include category names in card titles (e.g., 'Financial Account Settings' vs 'Change Billing Password'). If card titles share keywords, participants will group them mechanically by word matching rather than semantic meaning.
Sample Size: For qualitative open card sorts, recruit 15 to 30 participants to detect meaningful statistical clusters on dendrograms.
Topics
Discovering Taxonomy: The Power of Card Sorting
Preparing Cards and Sample Sizing
Validating with Tree Testing: The Reverse Card Sort