Stellar IA: the required inputs

If the output is to be a stellar information architecture, what are the inputs we need?

Card sorting is a technique to gain insights and understanding into how users would organize the content we provide in our digital products, and what labels they would give to that content.  It uncovers users’ mental models related to the content on our site, app, software, etc. We, as IA’s, should use that insight to inform a better IA for our product bearing in mind that card sort results are a great input, but not the only one we need to consider.

In our last post on Information Architecture, we talked about card sorting and about the fact that as a precursor to doing a card sort, we need to understand a few things so as not to introduce bias:

• What’s the content we have to work with (we’ve done a content audit or content inventory)
• Who is using it (we researched and developed our user personas)
• How, when, where, how often do these personas use it (we’ve researched their scenarios of use)
• How do those users categorize that content (we have insight into their mental models)

Then we conduct the card sort and gather the data. So what now?

There are a few good online tools for card sorting that don’t require a lot of deep knowledge about the statistical analysis behind the results. Knowing that the kind of stats that are running in the background of these tools is a kind of cluster stats is helpful conceptually, and knowing that the output is a hierarchical dendrogram sounds impressive; but, the answers to how to structure our IA are only suggested by the impressive and complicated looking dendrograms -� not prescribed.

So, recall one of the definitions of information architecture:

1. Information architecture  is the art and science of organizing  and labeling web sites,intranets, online communities  and software to support usability and findability

At this point in the process of creating a stellar IA, (we’ve done our user research, have our user personas, conducted our card sort), we have to start combining the art with the science. The kinds of questions that the science part, (the card sort results), will answer for us are:

• Which terms did users always gather together into groups
• Are there any legitimate “mistakes” in those groups -� that we or content SME’s (subject matter experts) would catch
• What did users call those groupings (did they use domain-specific jargon, or “lay” terms)
• Are there any similarities in these grouping names
• How many distinct groupings did users create with the content
• Can we combine any of these groupings to make broader-narrower relationships
• Were there any terms that some users consistently put into one group and others consistently put in a different group (suggesting we create a polyhierarchy)
• How well will the groupings “fit” into the device for which we’re designing this IA

Now we start to inject the “art” part -� these are also essential inputs we need to consider. With what we know from the science and what we (should) know about our users and their scenarios of use:

• should we take a persona or role-based approach to the IA
• should we take a content-based approach to the IA
• should we take a format approach to the IA
• should we take a task-based approach to the IA
• should we take a subject or topic based approach to the IA
• should we take some combination of these approaches and if so, which ones make the most sense

And, finally, what does the business need? Are there requirements they insist upon in terms of the voice or tone that’s conveyed through the IA? Does an organisational approach to the IA make sense in their context? The impact of a good IA to a business can’t be understated…but that’s another post for another time.

By now, only a couple of these approaches will likely strike you as the “best” to use, so start out trusting that feeling and start to play with the top level groupings. Don’t labour over this too much -� just start creating the IA, let it sit for a bit, come back to it and tweak it. Validate that the groupings make sense from the point of view of a SME, but exercise your good judgement about the balance that’s needed between a SME’s point of view and your users -� if your users aren’t all also SMEs. Then start to test it -� again -� with real users by conducting a reverse card sorting activity … which we’ll discuss in an upcoming post.