Card Sorting Template for User Research

Thirty to sixty cards works best. Fewer than 20 cards produces groupings that are too obvious to be useful. More than 80 cards causes fatigue; participants start rushing through the final cards and producing unreliable groupings. If you have more than 60 items, consider running two separate studies focused on different parts of your product.

Use an open sort when you have no existing structure or want to challenge your assumptions about how users think. Open sorts reveal unexpected groupings. Use a closed sort when you have a proposed structure and want to validate it. Hybrid sorts work well for refinement: you have a draft structure but suspect some categories need adjustment. The [MECE Framework](/frameworks) can help ensure your proposed categories are mutually exclusive before running a closed sort.

Cards placed in an "Unsure" pile or distributed across many categories signal a labeling or scoping problem. First, check if the card label is clear. "Advanced Settings" might mean different things to different users. Second, check if the item genuinely spans multiple categories. If so, consider making it accessible from multiple navigation paths rather than forcing it into one location.

You can, but treat internal results as directional, not definitive. Internal teams have expert knowledge that biases their mental models. They will group items by how the product is built, not how users think about it. Run an internal sort first to identify your assumptions, then validate with actual users. Compare the two results to see where internal and external mental models diverge.

Translate your findings into a proposed information architecture, then validate it with a [tree test](/templates). A card sort tells you how users group items. A tree test tells you whether users can find items in your proposed structure. Running both gives you high confidence in your final IA before investing development time. Use the [RICE Calculator](/tools/rice-calculator) to prioritize which navigation changes to implement first. ---