Essential Research for Information Architecture Design

17th November 2009

To classify and categorise is to be human … a bold claim perhaps. But people are fantastic and natural meaning finders. True, only someone somewhere on the obsessive-compulsive scale might have the patience to classify every variety of butterfly (there are over 50 in Britain alone). But clump together any random collection of words and people will try and make meaning of it.

So found George Lakoff in his seminal “Fire, Women, and Dangerous Things” opens in a new window . Women congratulated him on his insight in selecting his title: sure, women – they’re fiery; they’re a bit dangerous too. Actually, he’d meant it as only a semi-random collection of words, but his female friends couldn’t resist making associations. And who’d be brave enough to contradict them?

Categorising: A Natural Tendency

This natural tendency is also seen with images like Rorschach test ink blots and Tarot cards, and is fantastically useful in the design of Information Architectures (IA). Rorschach test

So often, IAs are heroic attempts to relate the unrelated. If you have ever tried to design an Intranet which combines the corporate tribes of financiers, techies, marketers, HRers, business development and the operational staff, you’ll know what divides them is often larger than what unites them.

Despite this, present a collection of topics to someone, and they will find it irresistible to start forming links. This is the basis of the Card Sort, an exercise often used to in make sense of disparate items.
For this year’s World Usability Day’s on sustainable design, we at User Vision asked our visitors to have their own go at bringing order to a semi-random collection of waste waiting to be recycled.

Waste: A Classic IA Problem

We all need to recycle more; actually, quite a lot more. But waste really only has one thing in common – it’s waste. We don’t need it anymore. Before it was waste it was a tin of Quality Street, a tub of coleslaw, a deodorant spray, a jar of beetroot, the wrap around some black pudding, some baby wipes … well I don’t have to reveal all of User Vision’s shopping habits, you get the idea.

Card Sorting is a great way of trapping what sense people make of this hotchpotch. If you haven’t done one before, it’s really simple to do. Make up a list of topics, on cards, and ask users to sort the cards into piles. We could have chosen to give users some already labelled groups, such as “paper”, “plastic” and so on. This is a Closed Card Sort and useful if you have some sort of constraint on how people can group things.

We chose to do an Open Card Sort, which gives users full freedom to categorise as they find meaningful or helpful. It allows the users to give the groups they create their own labels. The advantage of this is that it gives full creative play to how people categorise things. The disadvantage is that people’s creativity is huge. For each user, they are likely to come up with five unique categories that other people haven’t. So, ten users creates 50 categories, 100 users gives us 500 categories, and so on.

Card Sorting Analysis

Fortunately, several analysis tools have been evolved to help make sense of people’s unrestrained creativity. The cluster matrix, below, shows you what people put with what (and by implication, what they didn’t put together). So 100% indicates that two items were always put together (shaded dark blue) and 0% (shaded white with no number) that they were never put together.

Card sorting table of recycling items

This is an excellent mine of information. An IA doesn’t just fall out of an analysis like this, but it helps to see the obvious associations and disassociations. Of course, there are always the lonely friendless items that no one wants to be with. Here, even cotton buds and baby wipes don’t care to hang around with disposable nappies.

An alternative is to see best fit by Dendrogram (Tree Diagram). This also uses the number of times items were placed together to create a set of links (see below). The further items are grouped together to the left, the higher the number of times they were grouped together.

In this example, the items are split into seven groups. The groups show quite a variety of sizes, from single items (Johnny-no-mates nappies) to 13 items (a third of all the items). This isn’t ideal when building an IA, but can sometimes reflect the reality that some items are just outside the main orbit of other topics.

Card sort dendogram showing relationship between items

The Art of IA

The art of the Information Architect is to try and make these outliners seem not to arbitrary. Nor will anybody thank you for categories labelled “other” or “miscellaneous”. It’s import to try and pick a theme and try and run with it. So in recycling, do you categorise by:

  • What it was made of: Glass, Plastic, etc. 
  • What it was originally used for: Packaging, Childcare, etc. 
  • How it will be dealt with: Recycled, Bio-degraded, etc.
  • Or construct a faceted IA that reflects all three of the above ways of grouping waste together

There is no perfect answer, there never is. But Card Sorting is a powerful way of bringing in people’s powerful and idiosyncratic worldview to the categorisation mix. Without it, the IA designer is only left with their own capacity to organise things into groups. This maybe formidable, but it leaves out the richness of people’s creativity to think differently, utterly differently, from you, and find completely different meanings.

Want this article on your website?

If you liked this article, feel free to republish it on your own website. All that we ask is that you include the citation below, including links, at the end of the article.

This article was written by User Vision, a usability and accessibility consultancy that helps clients gain a competitive advantage through improved ease of use.

81% of shoppers said convenience was an important factor of e-commerce. 59% only shop at sites that they are used to and comfortable with.

Gartner Group.