Card Sorting: Better Information Architecture by Aligning with Users' Mental Models

Businesses typically have a more nuanced understanding of their products than customers do. When a group of people spends 40 hours a week, week after week, poring over a product’s details and fixing its problems, they learn things about it that normal humans never would.

It’s actually pretty cool. They become experts.

But they also develop blind spots, especially when it’s time to market their product. Industry-specific knowledge, so useful in meetings and on the production floor, ends up confusing customers.

We see this a lot on the web.

Say you want to buy a durable, waterproof watch. You have a brand in mind, go to their site, and find a dropdown menu full of obscure model numbers. What good is that?

When a company structures its website like an internal document, relying on product classifications and insider terminology that customers don’t comprehend (or care about), it creates UX friction. Bounce rates balloon, conversion shrivels, and users abandon all hope of finding what they’re looking for.

Luckily, businesses don’t have to settle for websites with poorly organized content.

What Is Card Sorting?

There’s a way to learn what customers know, what they think they know, and how they prioritize information (a.k.a. mental models).

Card sorting is a generative UX research method that reveals users’ mental models by having them arrange topics into groups that make sense to them.

Designers use the data from a card sort to improve an app or website’s information architecture, a design factor that profoundly affects people’s ability to find the content they’re looking for and carry out the tasks they want to complete.

There’s immense benefit here. Card sorting allows businesses to create intuitive user experiences by organizing and categorizing content the way customers do in their minds.

How to Perform a Card Sort

Most UX research methods can be conducted in a variety of ways, and card sorting is no different. There’s…

• Open vs. Closed: Do users create their own category names (open), or are category names determined beforehand (closed)?
• Moderated vs. Unmoderated: Is the card sorting session led by a facilitator (moderated), or do users work through the session on their own (unmoderated)?
• Paper vs. Digital: Are topics written on tangible index cards (paper), or are they typed onto cards in a simulated environment (digital)?

There’s also a reverse card sort (a.k.a. tree test) where users are given a collection of cards that are pre-organized into categories and subcategories and asked to complete tasks by navigating from the top down.

In general, the step-by-step process of a card sort is straightforward. To get the most accurate results, plan to test 15-30 people. After 30 people, card sorting fails to yield a substantially clearer picture of users’ mental models.

Step 1: Select Topics

• Create a set of 30-70 topics that relate to an app or website’s content.
• Each topic should be written on its own card.
• Topics can be general or specific, but refrain from using topics with shared words. Users will likely put those cards into the same pile even if they aren’t truly related.

Step 2: Think Aloud

• Have users think aloud while they work through the card sort.
• This can provide even more context into why users associate certain topics the way they do.

Step 3: Create Groups

• Mix the cards, and ask the user to put topics that go together into piles.
• Piles can be large or small, and users can create as many as they want.
• If a user doesn’t know what to do with a particular card, it can be placed in a pile of “uncertain” cards.
• Make sure that users don’t place cards into piles arbitrarily, and let them know that it’s ok to change their minds as they go.
• Cards can switch piles, piles can be combined, and so on.

Step 4: Name Groups

• When, and only when, the user is finished creating groups, have them name each pile.
• Doing so will show how users classify content topics in their minds, but it won’t necessarily provide clear labels to use on an app or website.

Step 5: Ask Questions

• Have users describe the underlying logic of the groups they made. Ask them if there were topics that they found difficult to categorize and whether or not they noticed any topics that could fit in multiple groups.
• Also, if they created an “uncertain” pile, ask them to explain why these topics were confusing.

Step 6: Combine Groups (Optional)

• Once users have finished sorting, naming, and explaining the rationale behind their groups, ask them to divide big groups into subgroups or combine related small groups into a larger one.

How to Use Data from a Card Sort

Once the card sort is complete, the real fun begins. All the user data collected needs to be analyzed and placed into a report that can be shared with designers and project stakeholders.

The report will reveal the mental associations and assumptions users make about the test content. It will also highlight words or topics that are confusing or potentially advantageous. All of this can be used to refine an app or website’s information architecture.

Step 1: Organize the Data

• Each card sorting session creates data that needs to be documented and carefully organized.
• In addition to the categories created by users, there will likely be notes and recordings for every participant.
• Make sure that nothing is lost and that all artifacts are easily accessible and transferred to digital.

Step 2: Evaluate Qualitative vs. Quantitative Information

There are two types of information that can be extracted from a card sort:

• Qualitative information, which reveals why users think the way they do, is found in the comments made by users (another advantage of having users “think aloud”).
• Quantitative information is numerical in nature and looks at things like how often certain cards are grouped together or how many users created the same category names.

Step 3: Review Notes and Recordings

• There’s no reason to write notes or make recordings if they aren’t mined for insights.
• A single user might not reveal much, but when comparing the notes and recordings of multiple users, unexpected themes may emerge.

Step 4: Digitally Analyze and Visualize

• It can be immensely helpful to input card sorting data into a digital program for analysis (Excel and Sheets are popular choices).
• To go a step further, online card sorting tools like OptimalSort and UserZoom provide data visualization options like standardization grids, similarity matrices, and dendrograms.
• When viewed in graphic format, important relationships and patterns in data become evident.
• Be careful, though. It’s possible to misuse data visualizations to draw conclusions that aren’t really there.

Step 5: Create a Report

• Once everything’s been organized, analyzed, and summarized, combine all findings into a report.
• Strive to make the report easy to understand, and where possible, provide simple context notes that explain difficult concepts.
• Remember that people that aren’t familiar with the ins and outs of UX research should be able to read the report and quickly come away with a solid understanding of what it says.

Card Sorting Is the King of Information Architecture

When people come to a website, they don’t want to search, decipher, or interpret. They want to find what they’re looking for—fast. No other UX research method matches card sorting’s ability to illuminate the inner workings of user mental models. Practically speaking, it’s affordable, simple to conduct, and relatively intuitive for users to participate in.

A seamless user experience is built on big questions. Questions like:

• When they see the content on our site, how will our customers interpret it?
• Do we have content that makes sense internally but confuses our customers?
• What exactly are our customers looking for when they visit our site, and what routes do they expect will lead them to it?

Card sorting unearths all these UX treasures and equips designers to engage the nuances of information architecture with confidence rather than conjecture.