Card Sorting: A UX tactic for IA

Card Sorting: A UX tactic for IA

Often misunderstood or completely overlooked, Information Architecture is “a structural design of shared information environments” as stated by The Information Architecture Institute.  It’s what helps you understand your surroundings and navigate more efficiently – or find what you’re looking for. In the realm of user experience, there are 2 fundamental requirements for doing information architecture well:

  1. Know the content: understand the body of content you are going to structure – the more you know about this, the better – do an audit or an inventory
  2. Know the users and their scenarios: understand who uses your content, what they use it for, where they use it, when they use it, how they use it and why they use it – you can never know too much about this part – do user research

Card Sorting

There’s a third thing that will help you to do information architecture well. Calling this a requirement is probably overstating it, since from time to time it’s really not required. This third thing is card sorting – a form of user research that helps us understand how users think about the content and categories on our site or app.

Card sorting is a technique that is as old as printed stimuli. Originally physical cards with words (or pictures or even physical objects) were put on a big table and users were asked to group these cards into buckets or categories that made sense to them. Sometimes they were then asked to name those categories (open sorting); sometimes they were given the categories and asked to put the cards into them (closed sorting).

Although the steps in creating a card sort are simple, they’re not always easy. But first the steps:

Open card sorting:

  1. Determine which areas of your site are causing, or are likely to cause, users trouble
  2. Scrape the terms from these sections and put one term on each card
  3. Recruit real users of the site to sort those cards into groups that make sense to them
  4. Watch, if you can, how users sort the cards
  5. Note the labels they give each of their groupings
  6. Run (or have an online tool do this for you) some statistical analyses on the way users have sorted the terms to determine the groupings that occur most commonly and what the group headings used most often are
  7. Use all of this as input into your information architecture

Closed card sorting:

  1. Determine which areas of your site are causing, or are likely to cause, users trouble
  2. Scrape the terms from these sections and put one term on each card
  3. Recruit real users of the site to sort those cards
  4. Give users the headings you intend to use on your site – just the main headings
  5. Watch, if you can, how users sort the cards into your headings
  6. Run (or have an online tool do this for you) some statistical analyses on the way users have sorted the terms to determine how much overlap there is between users – does everyone put the same cards under the same headings; does there appear to be ambiguity or confusion in the main headings you’ve provided
  7. Use all of this as input into your information architecture

Among these 7 simple steps, the only easy part is watching users do the sort….and sometimes even that is painful if users are really struggling to understand the content that is going to be on your site.

Determining what content to sort, and then whether or not each term needs or should have an explanation is sometimes really tough.

Card Sorting Bias

Recruiting real users is not easy if you’ve not done any previous user research and sorting with those who know the site as administrators, designers or developers introduces bias and ultimately is a risky alternative to using real users.

Benefits of in-person Card Sorting

If you do your card sort remotely, by sending a link to the sort out, rather than in person, you lose the opportunity to observe where users struggle and talk to them about that. During that struggle we can ask users what’s happening and learn a lot about their thinking; a lot that is otherwise unknown to us. We will learn which terms users want to put in 2 places; helping us determine whether a polyhierarchical structure is warranted and if so, where. We will learn which terms users simply do not understand at all; enabling us to consider changing some of the labels we may have thought were known. We can see users grab a term, place in one group and then later move it to another group and we can ask why, again, helping us better understand the relationships and proximity issues we need to bear in mind when building out the IA.

Users aren’t perfect

Without real user input, we are going to be building an information architecture based largely on our own assumptions. Obviously in some circumstances users will ‘get it wrong’; they’ll make semantic connections between terms that in the context of our domain or our site don’t belong together. They’ll sometimes guess at the meaning of terms and then make erroneous assumptions about what to put together or occasionally, they’ll rush through without really thinking about anything other than finishing the sort, getting their incentive and getting on with their day.

The IA isn’t done yet

One last point to remember: card sorting doesn’t do the tough work of finalizing an IA for us. In the statistical analysis – simple to generate, but not easy to apply – we’ll be able to play around with the proximity of terms to each other. In the course of doing this, we’ll see that we can create an architecture with many main headings or only a few. We’ll need to draw on our understanding of users and their scenarios and context of use, as well as our knowledge of the entire body of content here, to make the judgement that will work best for most of our users most of the time.

Cindy Beggs, MLS, is Partner and Vice President at Akendi, a firm dedicated to creating intentional experiences through end-to-end experience design. To learn more about Akendi or user experience design, visit www.Akendi.com.

 

Leave a reply

Time limit is exhausted. Please reload CAPTCHA.