This is a long weeknotes post, focusing on something that I’ve been working on in the background for a while: the information architecture (IA) for my website. I’ve posted about this on LinkedIn a few times throughout the process, and was surprised how interested people seemed to be. So I thought it might be worth a deeper dive into how I approach IA, the mistakes I made, and everything I learnt from the process.
The problem
I’m in the middle of the seemingly never-ending process of overhauling my website. I’m doing it because:
- The services I offer to clients have changed
- The platform I use to sell digital products provides a poor user experience
- The website structure doesn’t work for the kind of content I’m creating now.
That last one is the big one when it comes to IA. The IA for my website has evolved organically over time, based on whims, vibes, and experience. It was fine, but it’s no longer fit for purpose. I currently post most of my editorial content in a category called ‘Thinking’, which is a blog in all but name. But I don’t think this works, as most of this content is evergreen rather than ephemeral. (More on that later.) Plus, as I offer this as a service to clients, I really need to do it properly.
So what I’m looking for from my new IA is something that’s easy for people to navigate, whether they want to:
- see what services I offer and whether I’m someone they want to work with
- find information, advice, or templates covering a content strategy or content design topic
- sign up for my email newsletter
The hypothesis
My hypothesis was that the website IA should look something like this:
- Services/what I do/a similar label:
- Consultancy
- Mentoring and coaching
- Speaking
- Resources/library/a similar label:
- A big, flat section made up of advice/information and editorial content (articles, guides, interviews, tools, courses, templates). People would then be able to filter by content type, topic, and whether the content is free or paid for.
- About
- Weeknotes
- Contact
- Subscribe
However, I had some reservations about the ‘Resources’ section and had been weighing up a second option, where I would separate out free content that people read (advice and information, reports, interviews) from content that people interact with and sometimes pay for (tools and templates, courses). So I wanted to do a card sort and then a tree test to prove (or disprove) the hypothesis and help me refine my approach.
The card sort
My first step was to run a card sort test. In a card sort, you show people a set of cards that reflect the topics and tasks that need to be on the website to meet user needs. They then organise the cards into categories, and name them.
Video description: a screen capture showing someone sorting cards in groups and naming a category.
An aside: running this kind of thing publicly is nerve-wracking. Some of my users are my peers – fellow content designers, content strategists and digital design professionals who know how to do this stuff – so it feels like an extra level of scrutiny.
There were some interesting trends in the results:
- Services: Almost 9 out of 10 participants created a group with this name (or something very similar). And there was 100% agreement among them that the category should include these 3 core cards: Consultancy, Speaking, Mentoring and coaching.
- Tools and/or templates and/or courses: Just over 7 in 10 participants created a category they called either ‘Tools’, ‘Templates’, ‘Courses’ or some combination of all 3. There was a high level of agreement on a core set of cards for this category, too. People were marginally more likely to use the word tool.
- Blog: Almost 6 out of 10 participants created a category called some variation of ‘blog’. But the content that they put in that category varied a lot. Some people included guides, interviews, reports, and others did not.
- Other content type-based groupings: People created a pretty wide range of other content-type based formats too, including:
- Interviews – 4 in 10
- Courses – 4 in 10
- Resources – 3 in 10Articles – 2 in 10
- Topic-based groupings: There was also a light smattering of topic-based groupings like:
- Content strategy and operations
- Content design advice
The card sort confirmed some of my hypothesis about the ‘Services’ category, so I felt confident taking that idea forward.
But I still felt pretty unsure about how to approach the advice/information, editorial, and tool-based content. The card sort showed that people were likely to group the interactive content, like tools, templates and courses together. But it didn’t give me any good leads on how to categorise the evergreen advice/info and editorial content. The most common category people created for this kind of content was ‘Blog’. But this is one of those times where I’m not going to follow what the research tells me – at least not directly anyway – because using a blog is one of the big issues with my current site.
I’m a stickler about what a blog is for. I believe that for something to be a blog, it must:
- be regularly updated with new posts
- have a date on posts
- have a named author for posts
- focus on ephemeral editorial, opinion-based, or personal content
The content that most people put into their ‘Blog’ categories was evergreen as well as editorial. It has longevity, and – to me – it’s not suitable for a blog. Plus one of my objectives is to use the IA to help people discover this content in the long-term. If I was to use a blog, this wouldn’t work.
The tree test
Based on this, I decided to tree test the flat ‘Resources’ section approach. But I decided that I’d include content type groupings and content topics as part of the tree (even though they would be filters rather than true categories).
In a tree test, you give users a set of tasks, for example ‘You want to learn how to do a content audit. Where would you expect to find information to help you with this?’. They then click through a basic representation of the IA and indicate where they would expect to find it.
Video description: A screen capture of someone taking a tree test and looking through a list of nested labels to find one that matches the task.
Across the whole tree test, the success rate was 70%. So in 7 out of 10 cases, people could find the content in the information architecture. I would usually want to see an average success rate of 80%.
I tested 7 tasks in total, covering some core user needs and a few edge cases that I wanted to probe a bit:
| Task | Success score | Directness* |
| 1. You want to sign up for the 10 Things email newsletter. Where would you expect to be able to do this? | 92% | 90% |
| 2. You want to learn how to do a content audit. Where would you expect to find information to help you with this? | 92% | 89% |
| 3. You work for an organisation and you’re looking for an agency or freelancer to support you on a project. Where would you go to find information to help you with this? | 89% | 77% |
| 4. You’ve just heard the term ‘content model’ for the first time and want to find out what it means. Where would you expect to find that information? | 79% | 50% |
| 5. You’ve heard that there’s a really good Q&A with a content leader from the charity Scope on the site. Where would you expect to find it? | 67% | 79% |
| 6. You want to find out about booking Lauren to speak at your conference or event. Where would you expect to find that information? | 62% | 91% |
| 7. You work for a charity, and you’re looking for ideas and inspiration to help you with a redesign of your homepage. Where would you expect to find content to help you with this? | 8% | 74% |
*Directness score reflects whether people went right to the correct location, or had to go back and forth looking for an answer.
For task 5, the question about finding the Q&A, the only real pattern in the incorrect paths was a small group choosing ‘Weeknotes’ – perhaps because they were looking for an editorial-ish category? I’m not going to make any changes to the IA to address this, because in reality, I think this is ‘push’ content more than it is ‘pull’.
For task 6, the question about booking me to speak at a conference or event, around 1 in 3 participants chose the Contact page. Based on this, again, I’m not going to make any changes, as people would be able to achieve their goal by taking this path.
And last but not least, question 7. The question that dragged down the overall score with a spectacularly poor success rate. For this question – about finding ideas and inspiration to help with with a redesign of a charity homepage – I was hoping people would take one of the following paths:
- Resources > Insights and interviews > Charity homepage content report
- Resources > Tools and courses > Charity homepage toolkit
- Resources > Content topics > Content design and creation > Charity homepage toolkit
However, almost 8 in 10 people chose the path:
- Resources > Guides and advice > Website redesign guide
In retrospect, I think the question was flawed. Asking people to imagine that they worked in a different sector (for the non-charity participants) wasn’t a great choice. And ‘ideas and inspiration’ is really vague. But I wanted to avoid using leading phrases like ‘a resource to help you redesign your charity homepage’ or ‘a report about charity homepages’. Finally, the list of content in the tree was quite long, and the ‘Website redesign guide’ card was third in that long list, which might also have influenced people’s answer.
But despite the flawed question, the data also showed me that the ‘Insights and interviews’ category name probably isn’t ideal. I was concerned about this category name, as ‘insights’ felt vague to me. I want this category to include research (like the charity homepage and blog reports I’ve done in the past), interviews, and the occasional opinion-based things I write. Instead, I’m going to use a broader set of more specific content-based categories.
Another interesting learning was that people used the topics categories to navigate to content. I had half expected to see more of this in the card sort as, in my experience, people prefer to find content by topic than by content format. This didn’t happen in the card sort – perhaps because it’s hard to come up with these categories. So it was helpful to see this behaviour in the tree test and validate the choice to add the topics in.
This leaves me with this as my IA:
- Services:
- Consultancy
- Mentoring and coaching
- Speaking
- Resources
- Content format filter: Guide, Tool, Course, Interview, Report, Opinion
- Content topic filter: Content strategy, Content design and creation, Operations and governance, Workflow and process, Discovery and research, Collaboration and influence, Content leadership, Technology and tools
- About
- Weeknotes
- Contact
- Subscribe
Next steps
The next steps are to:
- Map my existing content to the new IA and taxonomy
- Do some user testing on my staging site to see how people use the navigation and filtering
I’m coming into a very busy period with client work, so this will probably have to wait for a while.
Thanks so much to everyone who participated in the card sort and the tree test. As I hope you can see, it was really useful to me.
P.S. I used UX Metrics for the card sort and tree test. It’s fine, but Optimal Workshop is much better and worth the extra expense – especially when it comes to the reports.
