Open Card Sort
Synonyms: open card sorting, open sort, generative card sort, exploratory card sort
Definition
Use cases
- Blank sitemap: The team argues over menu labels for 3 weeks. Nobody asks a user, so the loudest stakeholder wins.
- Jargon labels: A SaaS app files reports under "Insights." Users call them reports, so nobody clicks.
- Card overload: A study has 120 cards. Participants tire halfway and dump the rest into "Misc."
How it's used in practice
- Pick 30 to 60 cards: Pull them from a content audit, written in plain words.
- Recruit at least 15 people per user group: NN/g data shows 15 participants reach a 0.90 correlation with results from 168.
- Mix unmoderated and moderated: Run most online, then moderate 3 to 5 to hear why people group things.
- Standardize labels first: Merge "Billing," "Payments," and "Money stuff" into one category before you read the similarity matrix.
- Validate next: Turn the strongest groups into a draft menu and check it with tree testing.
Challenges & limitations
- Slow analysis: 20 participants can invent 60 category names. Merging them takes hours.
- No task context: It shows how people group content. It can't tell you if they'll find a page inside a real menu.
- Label bias: One vague card name skews every result tied to it.
Free resources
- Card Sorting: Uncover Users' Mental Models (NN/g) — The core definition, when to run a sort, and how open, closed, and hybrid differ
- Open vs. Closed Card Sorting (NN/g video) — A short video on what each type measures about users' mental models
- How To Create An Information Architecture That Is Easy To Use (Smashing Magazine) — Shows how to draft an IA with an open card sort, then test it
- 10 Things to Know about Card Sorting (MeasuringU) — Jeff Sauro's stats-backed notes, including margin of error by sample size
