Guide
I work on a lot of website redesigns, and the information architecture (or IA) is usually the aspect that excites and worries me the most. It excites me because IA is the kind of in-depth, meaning-making puzzle that I love working on. And it worries me because it’s a task that often gets glossed over, and thrown together by committee, based on a mix of what’s already there and what competitors/comparators are doing. These are both valid inputs, but they shouldn’t be the only ones. IA has a big role to play in influencing the overall success of your website, so if the IA isn’t designed with care, you’re taking a risk that could end up being expensive.
If you’re about to redesign a website, or you’re managing a website with an IA that you want to review, this guide is for you. It covers what IA is, why it matters, design methods, and heuristics to help you assess IA.
What is information architecture?
Information architecture is the way that we arrange the parts of something to make it understandable… We rely on Information architecture to help us make sense of the world around us.
Abby Covert, How To Make Sense of Any Mess
Information architecture (or IA) is how content is organised on a website, or in an app or digital product. It includes four main aspects:
- Categories: groups of information that belong together
- Labels: the names of the categories
- Relationships: how information connects across groups, and the order and hierarchy
- Metadata and taxonomy: the tags and attributes used to describe information
Information architecture vs navigation/How it’s expressed
IA is not the same as navigation. Navigation is a way that you express the IA and that users experience it. Your IA will also be expressed as and experienced via:
- Your sitemap
- URL folder structure
- Section landing pages and local navigation
- Filters and facets (a dimension used to classify something)
- Site search
- Search engine results (and increasingly in AI tools, more on that later)
Why information architecture matters
IA is a foundational part of usability that’s often overlooked or downplayed. It’s all about creating meaning and making sense for users, which has a huge impact on everything else that follows:
- Usability: the quality of your IA determines how easy it will be to find things
- Brand perception and goal performance: the choices you make about IA influence what’s most visible
- Efficiency and effort: a well-thought out IA makes content management easier, and reduces sprawl and duplication
- Equity and inclusivity: the way information is categorised affects if and how people understand and access it
When we organize information, we change it. The order in which it appears, the content that precedes or follows it, the ways we expand or condense it – everything we do to arrange information will alter its meaning. The key is to alter it in a way that enhances understanding
Lisa Maria Martin, Everyday Information Architecture
Information architecture and GEO/AEO
IA is also vital to how AI understands your content, which has an impact on how your content shows up in gen AI tools.
Search engines have relied on IA to understand what a website is about and how topics relate, and gen AI tools seem to be the same. This area of practice is still developing, but the theory behind how information architecture affects GEO/AEO is that:
- A good IA will give you a clear, consistent and logical approach to hierarchy, labels, URLs, and more.
- This makes it easier for AI to retrieve and understand the content…
- …which then makes it more likely that gen AI tools will summarise and surface your content in their answers. (Whether you think this is a good thing or a bad thing is a different question.)
How to design an information architecture
I’m not going to give you an exhaustive process – it’s complex and can vary a lot by scenario. And, ultimately, there’s no one ‘best’ way to design an IA. But what I can share with you are the methodologies I use, that you might want to consider as part of your process.
Research
A broad research phase will get your IA design work off to a strong start. Some of the research I’d want to carry out or gather include:
- User insight: Personas, user journeys, survey data, search keyword data, and so on. You’re looking for data that tells you what people need from the website, and the kind of language they use.
- Organisational information: Stakeholder interviews, strategy documents, service lists, org charts, impact reports. You’re looking to get a clear picture of what your organisation needs to communicate to the world.
- Content inventory/audit: A picture of all the content you have at the moment (and the gaps too). You’re looking to understand what existing content you need to find a home for.
Whenever I need to determine criteria for organizing digital content, I look to four key factors to shape my categorization:
The needs of the users
The goals of the business
The current state of the content
The strategic future of the content
Lisa Maria Martin, Everyday Information Architecture
Concept mapping
I use the outputs of the research to create a list of the different concepts that need to be in the IA. Concepts are tasks, topics, services, content types. Concepts are not pages – it’s hard to stop this association in the short term, but it will really help you to design a better IA in the long term.
For me, this generally takes the form of a spreadsheet with a list of concepts, with information on where it came from or the evidence for it. I also tend to start looking at broad categories and labels, and understanding what might be metadata/taxonomy (a tag or facet or filter, rather than a category).
Generative AI tools can be very useful for concept mapping, in my experience. They’re good at recognising patterns, can work through a lot of data quickly, and it’s easy to check the end results for accuracy. For example, on the last IA project I worked on (with permission from my client) I gave my tool keyword data and a content inventory, and it created a solid first draft which I then reviewed and refined.
Example concept mapping prompt
AI wrote this prompt and I edited it. It is a refinement of previous prompts that I’ve entered to create concept maps.
You are an expert in content strategy, content modelling and information architecture. Help me build a concept map for [organisation]’s [website or content area]. The goal is to understand what the content covers and how it could be organised, labelled and tagged.
Inputs I’ve attached (use whatever is here):
- [List of pages or content, such as a crawl, CMS export or sitemap]
- [Current taxonomy: categories, tags, content types]
- [Search or analytics data]
- [Audience research, personas or user research findings]
- [Project brief or hypothesis]
Before you start:
- Ask me a few clarifying questions about scope, priorities and output format. Wait for my answers before building anything.
- Check how complete and reliable each input is. Tell me what’s missing or skewed, and how that limits the findings.
How to approach it:
- Separate four kinds of concept:
- topics: what content is about
- facets: cross-cutting metadata and taxonomy labels that might be, such as content type or audience
- content model fields: information about the content, such as dates, scope or ownership
- anything that’s out of scope
- Test each proposed facet: would it actually help people filter or find things, and could it be applied consistently? If not, suggest a better place for it
- Group topics into a small number of clear top-level groups.
- Use the language that users and the organisation actually use. Say where each label came from, and flag any you’ve invented.
- Check the model against the main audiences: does each one’s main need have an obvious home?
- Keep the first version high level. Don’t produce a long list of detailed concepts unless I ask.
Outputs:
- A spreadsheet with:
- sources and caveats
- topics with definitions
- facets with values and tagging rules
- content model fields
- how the current taxonomy maps to the new model
- the evidence behind each concept
- open questions, each with a suggested starting position and a blank ‘Decision’ column
- A short summary in chat: key findings, the riskiest assumptions, and what to test with users.
Card sort
A card sort is a research method where you give participants a set of cards – each representing a topic, task, or piece of content – and ask them to group the cards in a way that makes sense to them, then name each group. It’s a good way to understand how people naturally categorise information, and the language they use.
Read the NN/G guide to card sorting
Tree test
In a tree test, you give participants a set of tasks and ask them to click through a text-only version of the proposed IA to find the answer. No visual design, no search. Just the structure. It will tell you if your proposed IA works or not, and how easily people can find information in it.
Read the NN/G guide to tree testing
Top tasks
Top tasks is a methodology designed by Gerry McGovern, to help you find the tasks that matter most to users and design a website structure that matches it. In brief the method is: gather a long list of tasks, consolidate them, then survey users to understand the 5-10 most important ones. It’s a good approach for service-based websites where people come with tasks in mind, or for products and apps. I used it to help a housing association design the IA for its customer account area, where people would come to manage their account, raise maintenance issues, see bills and charges, and more.
Find out more about Top Tasks and buy the book
Finding your organising principle
Whatever methodology you use, the key thing you’re looking to identify for your IA is an organising principle. An organising principle is the main pattern or logic that you’re going to use to group your content.
These are some of the most common organising principles (but this is by no means an exhaustive list):
- Topic: Content is grouped by subject matter. For example, GOV.UK uses topic categories like ‘Benefits’ and ‘Housing and local services’.
- Task: Content is grouped by things people want to do. This doesn’t mean that the labels need to be written like ‘I want to…’, just that groupings match things users want to do. For example, Monzo uses categories like ‘Current accounts’ and ‘Savings and ISAs’ that match tasks users come to the site with.
- Services: Content is grouped by the services or products an organisation offers. I’m cautious about this approach, and only use it if it also matches user needs and language. For example, the IA I designed for DACS uses categories like ‘Artist’s Resale Right’, ‘Licensing’ and ‘Advice’ that match both services the organisation offers and user needs.
- Audience: Content is grouped by the audience it’s for. This is another one I’m cautious about, as I find it can create some extra cognitive load for users to work out what group they are in. It can work where groups are clear, for example for universities, where grouping content by undergraduate and postgraduate users is logical and helpful.
- Format: Content is grouped by what type it is, for example guides, tools, opinion. This is a principle that I might use within a resources section of a website (and it’s what I opted for in the Resources section of my own website).
- Location: Content is grouped by place. For example, within the visit section of the National Trust website, places to visit are organised by area.
- Time: Content is grouped by date or period. A what’s on section for a cultural organisation might group events by when they are taking place.
As you can see from the examples above, sometimes an IA will combine multiple organising principles. You might have one organising principle at the top level, and then another to help users find content at a lower level. For example, The National Trust uses a task based approach at the top level (Visit, Discover and learn, Membership) and then location within Visit.
A warning about IA and org structure
If your organising principle mirrors the structure of your organisation or teams and departments, be wary. Internal organisation structure rarely matches the way users understand your organisation or want to use your website. If you find yourself going down this route, question whether the evidence shows this works for users. It can be a battle to get people to give up ‘their area’ of the website, and harder still to stop ourselves grouping information the way our organisation does, but it’s often necessary.
Heuristics for good information architecture
When you’re designing a new IA or reviewing an existing one, these heuristics can help you understand what to look out for. A good IA should be:
- Distinct: The groups that content is organised into should be clearly differentiated. There shouldn’t be too many examples where content could belong in more than one category. For example, if a charity website had groups called ‘Get involved’ and ‘Support us’ volunteering could belong in either group.
- Meaningful: The labels should be clear and carry meaning for the user. Someone who has never been to the site before should be able to predict the contents. For example, labels like ‘Discover’ and ‘Explore’ are very common in the cultural sector, but make the user work pretty hard to understand the contents.
- Exhaustive: Your IA should give everything a place. There shouldn’t be any straggly topics or pages that don’t belong. For example, putting everything in the blog, or coming up with an ‘Other’ category are signs that your IA is not exhaustive.
- Durable: The IA should be able to accommodate growth in content, and (reasonable) changes to your organisation’s activities. For example, if you were to launch a new service, a durable IA would be able to accommodate that.
- Consistent: The wording in the labels should follow a consistent format and structure, so it’s easy to follow. Some of the common approaches (which are often combined, or used inconsistently) are:
- Verb: Find services, Get advice
- Noun: Services, Advice
- Gerund: Fundraising, Campaigning
- Organisational voice: Our services, Our research
- Question-based: What we do, Who we are
- Balanced: There should be a manageable number of categories and levels, and the categories should be reasonably even in size, to make it easy to use. Will a user be able to take in all the options? Can people reach content without moving through too many levels? Do the categories have a fairly consistent amount of content?
Key takeaways
- Information architecture (or IA) is how content is organised on a website, or in an app or digital product.
- IA is not the same as navigation. Navigation is a way that you express the IA and that users experience it.
- IA design is a small investment at the start of a project that acts as a foundation for everything else that follows: usability, SEO, GEO, and goal performance.
- The key thing you’re looking to identify for your IA is an organising principle: the main pattern or logic that you’re going to use to group your content.
Reading list
Top Tasks – A how-to guide, Gerry McGovern
Information Architecture Heuristics, Abby Cover
Eight principles of information architecture, Dan Brown
How to Make Sense of Any Mess, Abby Covert
Yippee IA: Six principles for creating a successful information architecture, Chris How
Everyday Information Architecture, Lisa Maria Marquis
From Information Architecture to GEO: Rethinking IA for the AI Era, Svetlana Neškoviċ
Generative Engine Optimization (GEO): A Practical Guide, SEM Rush
Services are verbs combined with nouns, Duncan Stephen
Information Architecture: Study Guide, NN/G
Related resources
Designing the IA for my website
