Note to the reader:
This part of the field guide comes from our 2019 version of the UX Research Field Guide. Updated content for this chapter is coming soon!
Want to know when it's released?
Subscribe to our newsletter!
You can’t build buildings without a solid foundation. Likewise, you can’t build products without a solid, research-based information architecture—and one of the most popular methods for testing and validating your information architecture is card sorting.
Card sorting is a cost-effective way to learn how participants understand, group, and categorize specific information. With this insight, you can build easy-to-navigate structures for your products and websites—by putting the “contact” button wherever users would intuitively search for the “contact” button, for example.
Here, you’ll find everything you need to know to put this research method into action.
Card sorting is a research method in which participants group topics together in a way that makes sense to them. Typically, UX researchers conduct card sorts by writing topics on individual cards, asking users to sort related topics into groups, and then asking users to name or categorize each group.
Card sorting provides visibility into a user’s mental model, their thought process about how something works, or—more to the point—how they think it should work. Using these mental models helps UX and product designers build more intuitive, easy-to-navigate information architecture.
Here’s an example of what a physical card sort might look like, from Product Designer Andrej Dragisic:
As you can see, the cards are grouped together into like categories: Dental, hair care, lotion, and soap. Conducted with only notecards, sticky notes, and a marker, card sorts like this one are relatively quick, easy, and inexpensive.
Although card sorting is most commonly used to explore and inform the IA of websites and apps, it does have other uses. IBM, for instance, uses card sorting to prompt study participants to tell a story, establish a multi-level hierarchy, and define priorities.
Card sorting is most often used as an initial step to help determine the most user-friendly structure for a website or product. Tree testing, on the other hand, is sometimes called ‘backwards card sorting’ because it involves showing the participant the existing information architecture and then observing the click path they take to accomplish a specific goal.
Card sorting helps you create the trees you test in tree testing. Read the Tree Testing Chapter to learn more.
No user research method is perfect. Here are some of the pros and cons of card sorting to take into consideration when planning your study.
Benefits of card sorting:
Potential drawbacks of card sorting:
Card sorting is a generative (rather than an evaluative) research method. It cannot tell you what is wrong with an existing site or product. Instead, card sorting is most useful at the beginning stages of a project, such as designing a wholly new product or redesigning an existing website.
Card sorting can help you understand your target users’ mental models so you can:
Insights gained from card sorting help you design better user experiences by giving you the information you need to tailor the UI to meet their expectations.
The early steps in the user research process are similar no matter which method you choose. It always starts with creating a user research plan, outlining your research goals, related company goals, baseline assumptions and hypotheses, and more.
Check out Chapter 2 of the Field Guide to read in-depth on planning user research.
There are three steps you need to take before you can use card sorting effectively:
The type of card sort you choose will depend on your research goals and the kind of information you’re hoping to learn. There are a variety of options to consider, including:
Let’s explore the right use cases for each.
Open card sorts, in which users sort cards into groups and then categorize the groups, are the most common type of card sorting study. In closed card sorts, users organize the cards into predefined categories.
Many researchers choose to use a combination of both open and closed card sorts, to first identify category labels and then determine whether or not those labels are effective.
Card sorts can be completed by teams or by individuals. In Card Sorting — Designing Usable Categories, Donna Spencer highlights the benefits of team card sorts as a way to capture a wealth of layered insight:
“During the card sort, the participants talk about what they are doing, argue about where various cards go, discuss different ways to group the cards, query what the content means, and talk about how they might use the content. This discussion is incredibly valuable—in many cases, the discussion is more valuable than the outcome of the card sort.”
This discussion helps illuminate the participants’ mental models, because they have to communicate their thought processes and defend their choices. Spencer finds individual card sorts to be a great way to capture a large number of responses quickly and efficiently, but notes that this approach—even when researchers ask participants to “think aloud” as they work—doesn’t deliver the same depth of insight as a team sort.
Of course, in any group dynamic, you run the risk of one dominant personality taking over or groupthink delivering a less-than-clear outcome. (Groupthink is a commonly cited pitfull of focus groups, which we cover in this chapter.)
Whether you choose team, individual, or a combination of the two methods, it’s important to observe the process directly so you can see for yourself how the process flows.
In a moderated card sort, researchers are able to ask follow-up questions to dig deeper into participants’ choices. Unmoderated card sorts are typically faster and less expensive to run, but don’t allow for this in-the-moment clarification.
Moderated card sorts
Unmoderated card sorts
When choosing between moderated, unmoderated, or a combination of the two, you’ll need to balance the depth of insight you’re looking for with your available time and resources.
In-person card sorts use traditional, paper cards while remote card sorts use digital, online tools. There are pros and cons to both methods:
Paper card sorts
Digital card sorts
Because digital card sorts are faster and easier to implement, they’re ideal for small, resource-constrained teams—or remote teams, which at this point is most of us.
Getting relevant, in-depth results from a card sort depends, in large part, on picking the right cards for the exercise.
How many cards should you include in a card sort?
With card sorting, more is not always better. In fact, too many cards can overwhelm participants, causing them to burn out toward the end of the exercise. Most experts recommend using between 30 and 60 cards, with 40 cards being the sweet spot.
Tips for creating the most effective set of cards include:
It’s important to note that just because you’re aiming to ultimately end up with 30 to 60 cards doesn’t mean you can’t start with a whole lot more options. In fact, it’s best to curate as robust a list as possible when you’re getting started, and then cull the list down through careful evaluation.
Information architect and digital experience designer, Andy Fitzgerald, has a helpful post that outlines a collaborative method for choosing topics. He recommends this approach because it’s a great way to access multiple viewpoints, gain additional insights, and generate buy-in from both project teammates and clients. It can also be a really fun activity!
It’s surprisingly easy to insert unintended bias into your cards. You can avoid this by removing cues that suggest patterns or hierarchy.
Stick to a consistent level of granularity rather than including multiple layers of parent and child categories. Mixing headings with the card set can be confusing for participants and skew results in unhelpful ways.
Make it easy for participants to quickly grasp the meaning of a card without any guesswork by making labels brief but specific. Use terminology that is familiar to participants and that reflects their perception of the content rather than internal labels and concepts.
The success of any UX research study depends on the quality of the participants you recruit—but finding and recruiting the right participants isn’t easy.
For best results, you need to recruit people who accurately represent the real-life users who will eventually be using your website or product.
Depending on the nature of what you’re trying to learn, it may be helpful to include both “experienced” and “inexperienced” users in the mix. This could mean recruiting and managing a panel of your own customers (more likely for closed or hybrid card sorts, where you already have an existing IA) or recruiting outside participants who’ve never used your product before (more likely for open card sorts).
You can find research participants by putting out calls for participants on social media, reaching out to your email list, or using UX research recruitment tools like User Interviews. Be sure to design an effective screener survey to weed out unqualified participants and offer fair incentives to thank participants and discourage no-shows.
If you’re not sure what a fair incentive looks like for a particular study, use our handy UX Research Incentives Calculator to get a data-backed recommendation.
Nielsen Norman Group recommends recruiting at least 15 people for the typical card sorting study, based on research published in 2004 by Dr. Thomas S. Tullis, senior VP of human interface design at Fidelity Investments, and co-author Larry Wood. For well-funded, high-stakes research, you could recruit up to 30 participants, but any more than that will likely have diminishing returns.
Still have questions about recruitment? Check out Chapter 3 of the Field Guide, Recruiting for UX Research.
If you are doing a moderated card sort, your goal is to maintain momentum and observe without accidentally leading the participant. A moderator should take care to give each participant equal opportunity to provide input. This is especially important in team card sort exercises in which a dominant personality can have undue influence. Take notes throughout the session to capture insightful comments, questions, and your own observations.
It’s also always helpful to record each session so that you can look back later to make sure you didn’t miss any details or patterns in the process. This can help you in analyzing data as well.
Analyzing the data from a card sort can be challenging, especially if your study includes a large number of participants working with a substantial number of cards. Finding meaningful patterns in the data can be quite time-consuming and usually involves advanced techniques like cluster analysis and dendrograms.
Even when researchers are adept with the relevant types of analysis, the analytical output can be difficult for clients and other non-experts to grasp. To overcome this hurdle, Shanshan Ma, a principal UX designer, developed a quick-and-dirty analysis method that uses spreadsheets to visually represent card-sorting data.
Here’s an example from the article:
Her process includes six steps:
Ready to get started? Thanks to a wide range of great card sorting tools and templates, you don’t have to reinvent the wheel.
🧙✨ Discover more tools in the 2022 UX Research Tools Map, a fantastical guide to the UXR software landscape.
As with any other kind of UX research, combining card sorting with other methods is the best way to get the most complete and accurate picture. Because card sorts are generative, they are typically done early in the process, and are often followed by evaluative tree tests. You can also uncover additional valuable data by adding user interviews or surveys to your research plan.
