Using Visual Card Sorting for Digital Asset Management Taxonomy


Card sorting is a popular technique for getting at users' understanding of content structure, relationships and terminology. I often will break out a card sorting exercise early on in a taxonomy project using terms extracted from the client's content or pre-existing taxonomies to get a sense of what kind of organizing principles are important to the users. I also find it to be a great way of socializing the project and educating people in taxonomy and information architecture.

Recently, I've been doing a lot of digital asset management (DAM) projects, where our taxonomy is meant to help organize and give access to a collection of images or other visual assets, usually through faceted search. While the typical card sorting activity is still relevant, given that we'll be using words/labels to create the taxonomy for the assets, I've found it to lack a certain something in this context.

So during my current project with a global pharmaceutical company, I decided to try a different approach: a visual card sort.

The same rules apply here as they do with traditional card sorting: you can do an open card sort (no pre-defined categories given), or a closed card sort (established set of primary categories). The only real difference being you do it with pictures, not words. And it tends to be a bit easier and fun, as many people are inherently visual.

The Test Run

I tried it out during an educational working session on taxonomy that was part of a larger gathering of marketing managers from around the globe. It was an excellent opportunity to not only teach taxonomy and DAM concepts, but also gather input on cross-cultural issues and differing mental models. To reinforce some of the concepts, I split the room into 6 groups and had half of them do a traditional text card sort, and the other half the visual version (I'll refer to this as the text/image combo) so that I could use their feedback to compare and contrast organizing text vs. images.

What you'll need:

  • Post-its (or other blank cards) for creating group labels
  • Markers
  • Color-print outs of images, either specific to the context or totally different 

If you decide to do the text/image combo, you'll need a set of terms printed out and cut up into individual labels that address the same subject matter as the images.

Important note on selecting content: choose some images that are challenging – ambiguous or strange – as well as more straightforward selections. If you are going to do a closed or thematic sort (see below), make sure you select enough images that fit the options you’ll be suggesting. You need enough similarity for groups to form, while challenging users with oddball images that are hard to slot and will make them think.

For my exercise, I wanted to avoid people getting caught up in the minutiae of their own content, so I decided to bring everyone out of the pharmaceutical domain and into something simpler and common to all: shopping. More specifically, shopping at IKEA. I printed out color images of products from IKEA, and labels from their site navigation (including some good red herrings).

The directions:

You’ve got a couple of options as to how you want to direct the visual sorting, depending on your purpose.

Open sorting (info gathering): Let participants know that they have free reign as to how they want to organize and group the images – no holds barred. If they want to do color, fine. If they want to do style, fine. Whatever they agree to as a group. Use this approach when you are at the outset of a project and you want to gather user data for your taxonomy framework.visual card sort exercise open sorting

Thematic sorting (education or testing): Assign each group a different theme around which to organize the images. Optionally, you can assign the themes so that each group is unaware of what theme other groups are following. An interesting post-sort activity is then to ask groups to walk around the room and guess the overarching themes of the other groups. Use this approach when you want to educate users on faceted classification of images, subjectivity of tagging, or when you want to test an existing taxonomy framework to see if it holds up with sample images. (You can additionally provide top-level categories within a theme if you are doing a closed sort for testing.)

If you are doing the text/image combo, divide your groups so that half of them are doing the text card sort and the other half do the images. You can use the open or thematic sorting options for text as well. In the IKEA example, I opted for the thematic sorting, as it was more of an educational activity. I assigned each group a different theme (some were specific to images): product type, room of the house, style, mood, etc.

The Results:

Once folks are done sorting, invite each group to talk about their results.

If you opted for the open sort, have the group explain why they chose their grouping principle, and what kind of use case it supports. Ask them what was easy or difficult about the exercise.
If you opted for the thematic sort, ask them to describe the process they followed to define the theme (if it was a bit vague, like “style”) and what was easy or difficult.

If you opted for the text/image combo, do the same thing with the text groups but take some time to compare and contrast the two approaches. For example, with our multinational group doing the IKEA exercise, the visual groups described how it was very challenging to agree on where the images should be classified for subjective themes like “style” or “mood” – “what’s modern for me is not necessarily modern for visual card sort resultsyou.” The group names were a great source of laughs too: "frugal", "boring", "wild", "sexy"...  The textual groups described linguistic challenges they faced with some of the labels: e.g. “hutch” means a credenza where you keep dishes in the US, but it means dog kennel in the UK.

I could easily make parallels to their own content and the difficulties inherent in describing and defining pharmaceutical marketing assets, like a picture of an elderly woman smiling and hugging her daughter – what does this image signify? Health? Happiness? Family? Cardiovascular disease?

All and all, I’ve found this exercise to be an excellent way to introduce the importance of taxonomy in digital asset management, namely the ambiguity of language and subjectivity of images, and the challenges of tagging. It’s simple and fun for participants, and can be an invaluable tool for socializing concepts or gathering initial data for your taxonomy project.