*This article has been updated since it was originally published in 2009*


This is number 3 in a 3-part series about card sorting, written by Akanksha Hirokawa and Tania Lang. You can find the other 2 articles here:

Part 1 - An introduction to card sorting for IA design

Part 2 - Running a card sorting session

Writing cards for a card sorting activity

Deciding which words to include on each card is probably the most important part of the whole process and is often where UX researchers go wrong. Here are a few card writing tips:

  • Try to avoid the use of 'Information about'. These are superfluous words and you will no doubt end up with a group titled ‘Information about…’. (In fact, we suggest advising users they shoudn’t create generic group names such as ‘information’ or ‘miscellaneous’)
  • Try not to include multiple topics on one card e.g. "Car insurance policy details and quoting". These are 2 different topics and participants may want to put them into different groups:
    1. Car insurance policy details
    2. Get a quote for car insurance.
  • Having said that, do try to include different possible schema on the cards so you don't drive users to a certain type of schema e.g. task based, audience based, topic based. Mix up tasks, user groups and topics in the content objects, and change the order so you don't lead users. For example:
    Get a quote for car insurance – QLD Residents
    NSW Residents – home & contents insurance claim.
  • Ensure one word doesn’t appear on many cards otherwise we can guarantee participants will group these together and use that word to label the group e.g. HR policies, HR forms, HR recruitment procedures, HR contacts etc. Another common example that has caught us out in the past is "How to…".  Quick tip: copy all your card wording into a word cloud generator to check if you have overused particular words. Try to use alternate words to describe the same thing such as "Guidelines for how to…." and "Process for …"
  • Keep cards short - no more than a brief sentence.
  • Don't provide too many cards as it overwhelms users - if you have a large number of content items, limit to around 50-60 tasks and give to users in two batches:
    Batch 1: about 30 representative cards from across the whole site
    Batch 2: the remainder of the cards which they can just add to their existing groups or create new groups.
  • If doing remote online card sorting, you can choose randomly give participants a portion of the cards to sort e.g. 35 of 50 possible cards.

Facilitating the session (in-person)

Demonstrate the process using a fun activity

The best way to describe card sorting to somebody is to get them to do it. Have a small set of cards, no more than 10 -15, and use something fun. Pictures are best because they are easy to recognise and sort. We use pictures of various cartoon family members: Family guy, Simpsons and Family guy; and ask users to sort them.

We discuss the results and explain this is what they are going to do for the session but with content from the client's internet site. This is a fun ice breaker activity that only takes a couple of minutes and also helps communicate what you want them to do.

Set up participants and give instructions

Each participant should get their own piece of paper and set of cards (unless you're doing a group card sort, in which case each group needs these items). Ask participants to spread all their cards out and encourage them to group similar items in vertical columns – this makes it easier for you to stick their cards down at the end.

Don't give all the instructions about what they have to do at the start. It is too much cognitive load and people won't remember. Monitor individual's progress and give them the next instruction/step when they are ready. This is where a co-facilitator can help.

Give them space – but offer help too!

  • Walk around the room the whole session and check how people are going and that they understand all the cards. We give participants highlighter pens to highlight any words or phrases they don't understand.
  • Ask participants to avoid using generic terms, such as 'miscellaneous'.
  • If they have time, we might ask them to break up some of the categories into sub-categories and give those names too.
  • Before they leave, check your understanding of their work. Make sure that you can read each participant's writing and get them to clarify if you're not sure what something means.
  • Once participants are sure they have finalised their groupings, stick everything down with sticky tape. Before you go to pick up their sheet, run your fingers over the page to make sure that all the cards are secure. Participants hate to see all their hard work end up in a pile on the floor!

Analysing the results

Now that you have this wonderful data from your users, it's time to make it work for you.

There doesn't appear to be any consensus regarding analysis techniques. We have heard of practitioners taking 2-3 days of effort to enter and analyse all of the data but we generally only spend a couple of hours. Our preference is generally to spend less time analysing card sorting results and developing the first iteration of the information architecture (IA) and to spend more time testing and iterating the IA with users until we get it right.

We find that simply entering all the first and second level categories into an excel spreadsheet and spending a bit of time colour-coding similar headings is a great way to start if you don’t have access to tools such as OptimalSort. This gives you a feel for the common themes that came out throughout the session and provide a good starting point for designing the first iteration of your IA.

However, the best way to analyse data is to enter the groupings into tools such as OptimalSort and let them do the hard work. You can then enjoy the benefits of obtaining better results from moderated card sorting alongside the benefit of digital collation and presentation of results.

When finalising the IA and specifying where content belongs, most of the time it is fairly obvious but we often refer back to the original sheets created by the participants to see exactly where each participant put a particular card.

In most cases, the top-level IA will be clear by the time you've analysed all your results. However, that's not always the case. For one of our clients, participants consistently grouped information in 2 ways. To resolve this issue, we actually developed 2 versions of the IA, and tested them both to determine which one was going to work better on the actual site.

Try to use the terminology that was supplied by your participants in your final IA – these are the terms that your actual end-users came up with, so they're likely to make sense on your site as well.

If you're looking for a tool to help you with your analysis, Donna Spencer has a great spreadsheet you can use or you can use online card sorting tools like OptimalSort.


This is number 3 in a 3-part series about card sorting, written by David Humphreys and Tania Lang. You can find the other 2 articles here:

Part 1 - An introduction to card sorting for IA design

Part 2 - Running a card sorting session