The navigation of a digital product is usually designed in one of two ways.
- Based on the experience of the project team, enriched with guesses, beliefs and knowledge about the latest UI / UX / AI solutions,
- Based on the results of qualitative research.
The second route offers, among other things, card sorting. For a long time, this type of research has been conducted offline, based on physical cards. Hence the name, which today also applies to the more modern and popular online form.
Card sorting is about checking how users categorize and prioritize information, what criteria are most important to them and what structure their knowledge has. This is where the user’s mental model comes from, i.e. a set of subjective beliefs about the product and its use. The navigation and information architecture need to be adjusted to this model.
Different approaches to card sorting
The sorting is based on labeled tabs that contain the names of the navigation items and categories. The test can be performed in one of several ways:
- Open card sorting – users assign any names to groups that they have created based on their own feeling. We will see how users would name the category for the information that the interface is supposed to convey. This is the most popular variant.
Question: I have different kinds of functions in my job management system. What groups will I combine them into and how will I name them? For example, “Task list” with “Most important tasks”, “Tasks with ending time”. Then the users could point to the name “Task List”.
- Closed card sorting – information, names and categories are prepared by the research team and users have to combine them. We check how users assign information to predefined groups.
Question: I have a category in the system “Administration” and “Users”, to which category I will match “Add users”.
- Hybrid card sorting – ready cards to be sorted, plus additional empty ones for replacement, if there are alternative ideas. We give the respondents more opportunities and leave room for their ideas.\
- Tree testing – in this version, instead of cards, the category tree and digital product are shown to users. The task is to reach selected subpages / products / content. The results of the research are the search paths that the respondents choose. It is often a form of confirmation of results from previous studies
Our researchers recommend combining card sorting with tree testing.
How to conduct a card sorting study?
While moderated tests with paper cards are still organized, online tests are becoming more and more popular. Carried out via dedicated drag-drop software – Optimal Workshop, UserBit, UserZoom, Maze.co, UXTweak, Userlytics. As it is a question of finding a realistic and consistent sorting pattern, 15-40 subjects are generally recruited.
Depending on the choice – moderated or not moderated survey; online or offline; variant (list above) – the organization and course of card sorting will differ. The following process can be a starting point for creating your own study.
- Establish a range of topics that represent the most important content of the interface. There should be around 40-80 of them. Each assigned to an individual card. You should avoid repeating words on cards to avoid automatic for some people assigning cards based on the same words.
- Assign a deck to each test person, with the task of dividing it into piles as they see fit, dealing only with one card at a time. It is important that you can:
– create any number of stacks,
– assign any number of cards to each pile,
– change your mind during the test,
– freely combine and divide piles at any time,
– specify precisely if any card cannot be clearly assigned,
- Asking to give each stack a name. Although the ideas of the respondents will most often not be suitable for implementation into the interface, the invented names will indicate the range of categories that can be found in the mental model of the statistical user.
- If the research is moderated, it is worth supplementing the results with interviews with participants in order to learn about:
– the motives behind the elections,
– differences in the level of difficulty with assigning individual cards or naming piles,
– a proposal for the division of stacks into substacks,
– details of doubts about cards put aside, uncategorized.
Card Sorting – summary
The end of the research is analysis. In the material, you need to find duplicate names of the stacks; the content and topics that were grouped together most often; as well as the cards most often put away without assigning them.
These formulas, together with the answers of the respondents after the sorting phase, will tell you where the problems are and why. Perhaps some of the topics are not associated with each other due to insufficient knowledge of users or specific names have not been understood because users do not use them, while the card deck lacks a matching synonym.
Reverse card sorting, on the other hand, compares search paths and looks for patterns to match the tree in the future with the thinking and action sequence of users.