Defining, analyzing, and validating information architecture
“We define information architecture as the art and science of organizing and labeling websites, intranets, online communities and software to support usability an findability.”
– Information Architecture Institute (iainstitute.com)
Card sorting is a research technique used to help organize complex information into intuitive, user-friendly systems. I use card sorts on nearly every project to help bring structure to page content, define the navigation, and build complex taxonomy systems–the information architecture. The most common use of card sorting is to refine a complex navigation system — sometimes just one branch of the navigation.
Since defining the information architecture (IA) of a website is not done often, I tend to go all in on researching and validating the system. That means recruiting users to participate in card sorting and tree testing.
There are a few methods for conducting card sorting exercises: open, closed, and hybrid. In a typical open card sort, participants sort cards into any categories they want to create. This is the best method in my opinion because it reveals the language that real users use to describe concepts throughout a product. For example, formal/informal or general/specific language. In a closed card sort participants sort cards into existing categories provided for them. This method is ok if you already have deep knowledge of your user, or if the groups you are using are well established. In a hybrid card sort participants start with existing groups but can also create extra groups if needed. Some participants may not understand or have a need for certain cards and may choose to group those cards separately.
Like many UX research methods, card sorting can be carried out in-person using card stock, or remotely using tools like Optimal Sort and UserZoom.
To conduct a card sort you first need to have some knowledge about your user and the experience you are creating for them. Pull from personas, journey maps, and other user insights.
In a spreadsheet, define the chunks of information you want sorted: tasks, topics, features, products, services, activities, etc. Pick one area to sort at a time. Pull on specific user needs, writing statements and questions from the user’s perspective. In a task-oriented sort, for example, cards might include: “How much will it cost to microchip my new puppy?” and, “I need to make my annual vet appointment for my cat.”.
Chunks of information should be focused and unbiased. Don’t mix broad concepts with specific details because your participants won’t know how to form relationships between the cards.
Test your card sort with a colleague to prepare for the real thing — this will also ensure your cards can actually be sorted. Aim for 50 cards for the most effective results. If you have more than 100 cards, choose a representative subset of cards so participants are not overwhelmed or give up before they are finished.
Plan to run your card sort with as many participants as possible. I recommend 15–30 participants depending on the quality of your recruiting. Analyze your results throughout and remove or revise problematic cards.
In-Person Card Sorting
For an in-person card sort, write (or print) each chunk of information on a sturdy card stock and give your cards to a participant with instructions. Observe the participant and don’t influence their decisions. If they ask for help, encourage them to keep working as if you were not there. When the sort is complete, ask the participant to walk you through their decisions so you can understand their rationale. Photograph each sort and document the results in a spreadsheet.
After the first few card sorts are complete, you may be comfortable letting a small group of participants sort the cards together. These participants should discuss their decisions out loud giving you deeper insights into why they place cards in each group. I write my follow-up questions down in advance to make sure I remain neutral and avoid leading questions.
Remote Card Sorting
For remote card sorting, I highly recommend using a tool like Optimal Sort. Optimal Sort conducts the sorting in an intuitive, drag-and-drop user interface, and automatically compiles insights on the sort results! There are a few drawbacks to remote card sorting though. First, you have to ask follow-up questions in a survey format which usually means less relevant follow up questions and less insight into the participant’s rationale and confidence with each grouping. I recommend scheduling a follow-up call with participants if time allows.
At the end of your card sort, you will need to analyze the whole set of data, identify patterns, and construct your tree diagram before moving into the testing phase to validate your system.
If you are using software or an online platform to run the card sort, the analysis will be much easier. Optimal sort, for example, has a tool for standardizing group labels which use variations of the same term as well as a cluster analysis tool which highlights common grouping patterns.
Forming The Information Architecture
Converting your insights into a structured navigation system is not entirely straight-forward. Card sorting participants do not consider the context of your product or business restraints so the results should be considered guidelines, not rules.
Start by creating a diagram of the complete information hierarchy using a tool like Omnigraffle or Axure. Use the standardized group labels you created during analysis and decide if the language is effective in the context of a website navigation. Draw on mental models, design heuristics, and axioms to break apart your hierarchy into menu systems that will meet both user needs and expectations. You can do this several times to validate your assumptions.
Testing is critical to validating assumptions in the information hierarchy. Tree testing is a simple usability test to ensure users can complete tasks when given the information architecture you created in the card sort. This is also referred to as a reverse card sort. The more complex your data is, the more variations you will likely test.
In-person and remote tree testing are both options again. For in-person testing, write your working navigation structure on cards and lay down the top level navigation cards in front of your participant. Ask your participant to complete a task like: “You want to schedule an appointment, what is the office phone number?”. They should point to the card they think the phone number sits beneath. If they get the first card correct, add the next card, listing all the navigation items in that branch. If the participant navigates incorrectly, you can give them another chance or move on to the next task.
Enter each test into a spreadsheet, tallying the nav items where each participant attributed a task. Again, online tools can help facilitate the test and keep track of the data. I recommend Treejack for online tree testing.