Card sorting is a technique for exploring how people group items, so that you can develop structures that maximize the probability of users being able to find items.
Card sorting is appropriate when you have identified items that you need to categorize.
Card sorting is particularly useful for defining web site structures.
Card sorting can be conducted in a variety of circumstances using various means - one-on-one, during workshops, by mail, or electronically. The following is the basic process.
Names of items to be categorized are printed on individual cards. Cards should be large enough to accommodate the names in a font that participants can read easily when spread out on a desk or table-at least 14 point.
Participants are asked to group items in a way that makes sense to them.
Participants may also be asked to name the resulting groups.
Once all participants have completed the exercise, enter the data in a spreadsheet, and examine the groupings. There will be general agreement about many items, and these groupings will be fairly apparent. For example, all participants may group 'Technical Support' with 'Complaints' and 'Product Assistance'.
You can use cluster analysis to get a pictorial representation of the resultant groupings.
Pay special attention to items about which a consensus does not exist. Would re-naming the item improve the situation, or does it need to be included in more than one category?
Make sure that all participants are representative of the eventual users of the structure you are designing.
Try to get as many participants as convenient, and at least six. However, be aware that the more participants you have, the more data you will need to analyze.
People vary widely in the amount of time they take to sort cards. As a rule of thumb, allow half-an-hour for a participant to sort 50 items.
There are many variations on the basic technique.
By making the cards slightly larger, you can add some simple questions. For example, ask whether terms are understood (yes/no/unsure) and whether there are suggestions for alternative terms.
It is possible to have people collaborate in grouping the items. However, this runs the risk of reaching a consensus that is not an accurate reflection of any one individual's perceptions.