Card Sorting (Wikipedia)
Wikipedia’s article on card sorting, a knowledge-elicitation method for understanding how people mentally categorize information, used to design information architecture, navigation, and menu structure.
License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/
Key points
- Basic method. A representative participant receives index cards bearing item names, groups them in whatever way seems logical, and names each group; this repeats across several participants, and the results are analyzed afterward for patterns (a category tree, or “dendrogram”).
- Open card sorting (generative): participants invent their own category names — reveals both how people classify items and what words they use for the categories, useful for discovering a new organizational scheme.
- Closed card sorting (evaluative): participants sort into a fixed, predetermined set of category names — reveals how much agreement there is on what belongs where, useful for judging an existing or proposed scheme rather than generating one.
- Reverse card sorting, more commonly called tree testing: tests an existing category hierarchy in isolation. Participants are given a task and have to find the right place for it by navigating the card-based hierarchy from the top down, with none of a real site’s navigational aids (search, breadcrumbs) — isolates whether the structure itself, not the UI around it, supports the task.
- Modified-Delphi card sorting (created by Celeste Paul): only the first participant does a full independent sort; each subsequent participant iterates on the previous participant’s result rather than starting over. Builds consensus with fewer participants than independent sorting.
- Analysis falls into two families: semantic methods examine the labels/descriptions participants gave (e.g. gist analysis, grouping differently-worded labels that mean the same thing — “Difficult to understand” and “Confusing”); syntactic methods examine the structure of the sorts themselves — card-based analysis builds similarity/dissimilarity matrices of which cards landed together, feeding cluster analysis and dendrograms; sort-based analysis measures the “edit distance” (card moves needed to turn one participant’s sort into another’s, calculable via the Hungarian algorithm) between whole sorts.
- Online/remote card sorting: web-based tools reach more participants at lower cost and can automate analysis, but lose the moderator’s ability to ask a participant why they grouped something a certain way.