Evaluative Card-Sorting Methods

Where open and modified-Delphi card sorts are generative, used to discover a structure, closed card sorting and reverse card sorting are evaluative: used to validate a structure you've already proposed. In closed card sorting, participants sort content cards into categories you've already defined, rather than creating their own. It's especially useful when you're adding new content to an existing information architecture that already has good findability, since it tests whether the new content fits cleanly into the categories that already work.

Reverse card sorting, a variant of closed sorting, goes a step further: it quantitatively rates how well an existing or proposed top-level hierarchy supports findability. Participants either sort cards into top-level categories or respond to realistic task scenarios by saying where they'd expect to look. The data lets you calculate the percentage of participants who placed each card where you intended, a direct, numeric read on how well your hierarchy matches people's expectations.

Closed card sorting asks "does this new piece fit the puzzle?" Reverse card sorting asks "does the whole puzzle actually make sense to people?" They're close cousins, but they answer different questions.

Exercise

The scenario: A content team is adding a significant new section to an existing e-commerce site. The site's current information architecture is well-established and performs reasonably well, customers rarely struggle to find existing product categories. Two researchers disagree on which evaluative card-sorting method to use to check whether the new section will fit.

Given that the existing hierarchy is already known to work well, and the specific question is whether new content fits into it, which method is the better fit?
Try It With Your Data: Conflict Resolution

Fill in your own details below; the prompt updates as you type. When it's ready, copy it into Claude or whatever AI tool you use.

What this prompt does

Helps you resolve cards that participants sorted into conflicting groups, with a clear primary home and a rationale you can defend.

  • When card sort data shows significant disagreement (<40% consensus)
  • When a card is placed in 3+ different groups by participants
  • Before finalizing taxonomy decisions with ambiguous cards
  • To make defensible placement decisions
The card name, e.g. "OAuth Setup"
One group per line, e.g. "Authentication, 12 participants (48%)"
One or two sentences
Names of related cards, comma-separated
What users are trying to do with this content
Whole number
Assembled prompt

        
        
      
Try It With Your Data: Similarity Matrix Generator

Fill in your own details below; the prompt updates as you type. When it's ready, copy it into Claude or whatever AI tool you use.

What this prompt is made of (RICE breakdown)
R

Role

Casts the AI as: You're a research analyst creating similarity matrices from card sort data.

I

Instructions

Names 4 specific outputs to produce, so the response comes back structured rather than a general summary.

C

Context

Study type and Purpose tell the AI what your specific situation is, not a generic one.

E

Expected format

Matrix visualization + cluster insights

Whole number, e.g. 25
Comma-separated list of every card in the sort
One participant per line, which cards they grouped together
Which method you ran.
Why you're running this analysis
Assembled prompt

        
        
      
Try It With Your Data: Hierarchical Cluster Analysis

Fill in your own details below; the prompt updates as you type. When it's ready, copy it into Claude or whatever AI tool you use.

What this prompt does

Use this prompt when: Analyzing card sort results quantitatively.

  • After completing a card sort study with 20+ participants
  • When you need to determine optimal number of categories
  • To visualize how tightly content groups together
  • To identify natural "cut points" for hierarchy levels
Whole number, e.g. 25
Whole number, e.g. 30
Which method you ran.
e.g. "Redesigning the docs top-level taxonomy"
Paste as "Participant | Card | Group | Label" rows, or a pre-calculated matrix
Brief description, or "none"
Any requirements the taxonomy must satisfy
Assembled prompt