UX Research & User Testing

Expanding Your Information Architecture

Paul Boag

Paul Boag

UX Research & User Testing

Check out a free preview of the full UX Research & User Testing course

The "Expanding Your Information Architecture" Lesson is part of the full, UX Research & User Testing course featured in this preview video. Here's what you'd learn in this lesson:

Paul discusses the process of expanding the information architecture by integrating more tasks and accommodating subsections. He explains how to conduct a closed card sort to gather user input on task categorization and how to analyze the results using an agreement score. He also provides guidance on organizing tasks into subsections and suggests using ChatGPT to help with the process.


Transcript from the "Expanding Your Information Architecture" Lesson

>> So yeah, you've done your initial information architecture, your top-level information architecture, but we probably want to, well, we do need to expand that out, right? So after we've done our top-level information architecture focused on our primary tasks, well, it's not all about our primary tasks. There was all those other tasks that got voted on, but just didn't get votes that were as high, so we need to accommodate those as well, right?

So to do that, we need a two step process, right? Step one is we need to integrate more tasks, those tiny tasks, let's call them that. But step two is, we also need to accommodate subsections, right? So not everything's gonna be in our top-level navigation. So, how do we do those two things?

Let's talk about our tiny task first. So we did an open card sort with about 30 cards, right? Now, if time is tight and that's all you could do, all power to you. Fair enough, great. Something is better than nothing. Let's move on. If you get the opportunity to do a little bit more, I would now run a closed card sort.

Now in a closed card sort, notice how the top-level sections are predefined, right? So this is where I've gone in and I've said based on the results of the open card sort, this is what I think the top-level sections should be, right. And then, but I've added one more, if you notice, I'm not sure, right.

So you've got all the sections that I created, and then we've got one extra that I added in called I'm not sure. So now because all users have got to do is basically take a card and drop it in whichever box makes sense to them, now I can test with a lot more cards.

I can add all of my cards in there and they'll go bang, bang, bang, bang, bang, and put them in different sections and we're good to go. And if they're not sure where to put it they just put it in, I'm not sure, right. So this is a great way of making sure our information architecture works for every task, not just our top tasks, okay?

So to create this is really, really simple. In UX metrics, again basically, instead of selecting open card sort at the beginning you select closed card sort instead, so you create a new closed card sort. Then what you do is you add all the top-level sections you made up based on what you did on the open card sort from looking at the results and etc.

So now we we add those all in as new groups. With that additional I'm not sure option. And then we go through and add all of our cards in, right, for all of our top tasks, the whole seven list of 70. It's very annoying. They need a bulk import, which doesn't exist in the app, but anyway.

And then you just publish that online and share it with whoever and get them to fill in the results, right? Yes.
>> Can you do this live? If you were doing it on Zoom or something, can you do it with a live?
>> You could do it live, yeah.

You could do it that way if you wanted to talk to people and understand what they're doing, cuz at the end it's just giving you a URL and whether you send that out via email, let people do it whenever they want, or you do it facilitated. That's up to you.

I don't see a huge amount of value if I'm honest in doing it live. You get the interaction and the conversation, which is always good, but it's probably a bit of a luxury, right? For me, I think I would prefer larger numbers of people completing it than a smaller number facilitated.

Does that make sense?
>> Cuz we've done it with Figjam.
>> Yeah.
>> It's much clunkier.
>> Yeah, it's not great. I mean, you could do it with Figjam, or what's the other one, Miro. Yeah, but this tool just makes it so easy cuz you can just send it out and they get on with it.

And again, we got to analyze the results, which is pretty much same as before. You go to the results page, but this time we're gonna click on Cards, right? So what we're gonna do is we're gonna look at all of the cards, and we're gonna be paying attention to the agreement score, okay?

So what the agreement score is, is it's saying how many people or what percentage of people agreed that that particular task, card, whatever you want to call it, should be in that section, right? So if it's got a very high agreement score, you can be confident that that task belongs in that section.

However, if it's got a low agreement score, say below 50%, then it means people are unsure about where it should go. And maybe there's an issue with the information architecture then, all right? You also obviously need to pay attention to cards that are assigned to the not sure group, because those are potential problem areas as well that people don't know where to put them.

But you will often find that it's not as bad as it first looks when you look at it. IA is not an exact science. It's not gonna churn out an information architecture at the end of it. It's just providing you evidence to make good decisions. So it will need to be a judgment call, right.

You'll need to decide what agreement score level is acceptable to you and how to handle uncertain items, right? And that will be largely dependent on your organization's culture and their attitude to risk, your audience, your content type. You have to go with your gut to a degree on this which is a bit rubbish for workshop, but there is no other way of saying it.

However, as a general rule of thumb, this is what I kind of think, right? If something has got an agreement score of less than 50%, in other words, if only 50% of people, or less than 50% of people agreed on what section it should be in, then I pay attention to it.

I also pay attention to items that are marked not sure, right? In most cases, the problem can easily be solved by assigning it to the second highest rated place, right? So let's say we had a task of dental care, right. And there was a healthcare group which had 48% of people assigned it to health care, right.

And then the next one down was benefits where I know 32% of people or designed it to benefits, all right. In most cases, if people aren't agreeing on the top-level section and there is a relatively close second one, you just assign it to one, doesn't matter which one and then crosslink it from the secondary one.

Does that make sense? So that would be a link in benefits as well as a link in in healthcare. That way you're covering your bases. If people don't agree, we'll just go for the second. Also in the not sure category. There'll be a lot of things in the not sure category, and you'll freak out because they go, these things are not sure.

But if you look at it, they'll oftentimes the second one was pretty high in the number of votes it had, right? Not all might be coming in at 51% but benefits was coming in at 49, well, we'll just put it in benefits then, right? So oftentimes just look at that second option and that'll help you kind of work it out a little bit.

So that's your belt and braces approach for doing top-level navigation. People are gonna have trouble arguing with you if you use all that, but it's quite a lot I know. So if you want to skip steps, go for your life, whatever you can practically do, because we do need to think about about lower level sections as well, cuz at the moment all we've done is create top-level navigation.

If you've got a large number of tasks assigned to a single top level section, then we're going to need to organize those tasks into subcategories, subsections. So typically I try and avoid having more than about seven elements in a sub section just as a general rule of thumb, right?

So if you've got 10 or 15, then you're probably gonna need two or three sub sections in that section, right? So there's two approaches to the sub section problem, right? One, and the grown-up proper, you're gonna be taught to do it the right way approach is more card sorting, right?

You would do card sorting for each section. You can see how this is gonna quickly get out of hand and beyond what most people are able to do, because you could have six top-level sections, and then you'd have to do card sorting for each of those. It's never gonna happen, is it?

Let's be honest, right? So, option two, make an educated guess, right? And this is where I get in trouble with all the real UX research experts that tell me off for having such a cavalier attitude. But sometimes existing knowledge and research is enough to identify appropriate subsections. And anyway, we can use the tree testing, which I'm coming on to in a minute, to double-check everything.

Guess what I'm gonna say now? We can use ChatGPT to help us do that, right? So, ChatGPT is really good at organizing conceptually similar concepts and ideas together, right? So we can say, act as an information architecture. You've been asked to organize the content I will give you into subsections within, in this case, the university's HR website's benefits section, right?

Avoid having more than seven subsections, cuz we don't want hundreds of subsections. Keep the label of each section to one or two words. Avoid ambiguity, cuz I said that earlier, in labeling, or two labels becoming too similar. Assign content into subsections, but based on conceptual similarity, the goal is to ensure users can easily find the content within the hierarchy.

Are you ready for the content, that list of section that we've got within that benefit section? Or do you have any questions? If it says it's ready, you copy a listing of stuff that was in the benefits section and it'll organize it for you. It's not the grown up way of doing it, right?

But it works most of the time. It gives you reasonably good results and it saves time.

Learn Straight from the Experts Who Shape the Modern Web

  • In-depth Courses
  • Industry Leading Experts
  • Learning Paths
  • Live Interactive Workshops
Get Unlimited Access Now