Card Sorting

Card sorting is a powerful technique for assessing how users group related concepts together. In its simplest form, a researcher would write concepts - usually menu items for interaction design - on cards and ask users to group related items together. In a closed card sort, the number of groups and their names are fixed. In open card sorts, the number and names of groups are determined by the participants, although the researcher may specify limits (3 to 5 groups, for example).

In a fully manual sort, the researcher would then have to go through the cards and produce a tally of how many times concepts were grouped together. The resulting proximity data can then be used establish overall grouping, usually through a statistical procedure called cluster analysis.

Naturally, processing card sorts by hand can be tedious: it would not be unusual to have 15 participants x (50 cards + 7 groups) = 855 cards. But computers are good at repetitive tasks, so several card sorting software packages appeared, most notably IBM's free EZSort software consisting of USort and EZCalc. Card data is entered by the researcher and participants manipulate the groups on-screen with USort. The resulting data is analysed with EZCalc, using cluster analysis techniques.

This makes the researcher's job much easier but raises serious concerns about the quality of the results since manipulating the cards on-screen can be confusing and intimidating for users. Fully automated card-sorting also usually have no provisions for participants to question or change the terms being used. (Via Design for Usability)