A Card Sort typically gets performed when a website is undergoing a design or re-design. It is incredibly helpful to have users give your feedback on how to present our Information Architecture using 1) meaningful groups of information, 2) a functional sequence between groups of information and 3) a clear and concise naming convention for labeling categories.
To conduct this a card sort study, an inventory needs to be first taken of the entire website. These items then end up being the “Cards” to be “Sorted” and analyzed for your Information Architecture
According to Usability.gov, to prepare your cards, you should create your list of content topics. Topics can be phrases or words, very specific or more general. As a suggestion, limit yourself to 50-60 topics or less. This means there might not be a card to sort for every page on the site.
For a website, your will need to audit the existing architecture of our website and list out our destination pages (objects) and the topics and tasks (definitions) that they cover:
According to Tullis & Wood, a card sort with only 20 to 30 participants can yield an average correlation coefficient well over 0.9. For best results, you should choose to run your study with participants that fall within your demographics. According to Courage & Baxter, running one or two group sessions with 10 to 12 participants of the same user type should be sufficient.
Once you have selected your participants, you can simply invite them via email to take part in your study!
First, introduce the participants and explain the study that you are conducting:
Then, have your participants group the cards and label their groupings as they see fit:
Here comes the hard part… This is where we will need to make sense out of our data.
To analyze the data from the study, you should start by analyzing the card groupings by user group for similarities and differences between user groups. According to Courage & Baxter, it is advised to analyze user groups separately since each group might have very different needs. Analyzing all the user groups together without first inspecting the differences between them could lead to an “averaged” user experience between each group that isn’t actually useful to any of the groups.
Here is where we can see what labels the participants used for categories:
We can then look at the Tree Graphs for grouping trends across each demographic group:
You can even go old school with the Tree Graphs by printing them out (for each demographic) and grabbing your sharpie:
You can also look for groupings using the Similarity Matrix:
Lastly, you will need to to conclude your study and make your recommendations. This is where you will use the data to propose a new information architecture.
We should also remember that when conducting a card sort study, it is important to accept that groupings across all user groups should be used as a guide, and adjustments should be made as needed, based on judgement.
This is where you should present the data to the stakeholder and finalize your recommended information architecture:
Your website should be much more usable once you implement this new information architecture. ;-)