All 141 tools / Research and usability testing
Turn open card-sort results into a similarity matrix and suggested categories from participant agreement
Card sort analyserFreeFor researchers, product teams and designers
45 min
saved on a typical run. Our estimate, against tallying co-occurrence and clustering by hand in a spreadsheet.
Before: what you give it
The example input, as typed in. You give it structured JSON.
participants: [{"Pay":["Invoices","Refunds","Plans"],"Get help":["Contact us","FAQ","Status"]},{"Billing":["Invoices","Plans","Refunds"],"Support":["FAQ","Contact us","Status"]},{"Money":["Invoices","Refunds"],"Help":["Contact us","FAQ","Status","Plans"]},{"Account":["Plans","Invoices"],"Help centre":["FAQ","Contact us"],"Other":["Refunds","Status"]}]
threshold: 50
After: what comes back
The first lines of a real run on that input. Nothing here is edited by hand.
# Card sort analysis: 4 participants, 6 cards Similarity is the share of participants who put two cards in the same group. Clusters come from average-linkage hierarchical clustering, cut where the average agreement between merged clusters falls below 50%. Method after Donna Spencer, Card Sorting; the cut-off and the low-agreement rule are conventions of this tool, not hard thresholds. Note: 4 participants is below the roughly 15 often suggested for card sorts (convention, after Spencer's guidance; unverified against the source). Treat clusters as leads, not findings. ## Suggested categories (2) - Contact us, FAQ, Status (3 cards, 83.3% average pair agreement). Names people used: Get help / Support / Help. - Invoices, Plans, Refunds (3 cards, 66.7% average pair agreement). Names people used: Pay / Billing / Money. ## Strongest pairs - Contact us + FAQ: 100% (4 of 4) - Contact us + Status: 75% (3 of 4) - FAQ + Status: 75% (3 of 4) - Invoices + Plans: 75% (3 of 4) ... full output on the tool page
What you get
Pair agreement percentages, clustered category suggestions, low-agreement cards and the group labels used
What it cannot do
Takes 3 to 60 participants and 5 to 100 cards, warns under about 15 people, and cannot say why people grouped cards.
Who it is for
- Researchers
- Product teams
- Designers
Free and Pro
| Free | 25 runs a day per IP address, up to 100,000 characters a run. |
|---|---|
| Pro | No daily limit, 300 requests a minute, up to 200,000 characters a run. |
Use it from your assistant
The tool is called analyse_card_sort. Connect the server once, then ask for it by name or just describe the job. Install steps for each app.
claude mcp add --transport http designers https://designers.allthepossibles.com/mcpcurl -X POST https://designers.allthepossibles.com/api/card-sort-analyser -H 'content-type: application/json' -d '{"participants":[{"Pay":["Invoices","Refunds","Plans"],"Get help":["Contact us","FAQ","Status"]},{"Billing":["Invoices","Plans","Refunds"],"Support":["FAQ","Contact us","Status"]},{"Money":["Invoices","Refunds"],"Help":["Contact us","FAQ","Status","Plans"]},{"Account":["Plans","Invoices"],"Help centre":["FAQ","Contact us"],"Other":["Refunds","Status"]}],"threshold":50}'Try it here
The form starts with the same example input, so you can run it first and then swap in your own.
More in research and usability testing
- Usability numbers
Score SUS results and work out sample sizes for usability tests, A/B tests and surveys.
- Heuristic review
Get a Nielsen heuristic worksheet for a screen or flow, with signals showing where to look first.
- Critique plan
Plan a design critique for the stage the work is in, with timed rounds, roles and feedback to hold back.
- Research questions
Find leading, double-barrelled and hypothetical questions in an interview or survey script and rewrite them.
- Trunk test
Check whether a page answers Krug's six trunk-test questions: site, page, sections, options, location, search
- Form check
Check a form's markup against Wroblewski's Web Form Design rules and get a fix for each finding