Definition
Information Architecture (IA) is the discipline of structuring, organising and labelling content within a digital product — including navigation, taxonomy, hierarchy and search — so users can find what they need without thinking about how the system is organised.
Bad IA is invisible — until users start failing tasks. They can't find a setting; they don't know which menu houses what they need; search returns the wrong thing. None of that is the user's fault; it's an IA problem. Good IA makes the system's structure align with the user's mental model of the domain, not with the engineering team's database schema.
The classic IA workflow combines top-down (start from the domain model, design the structure) and bottom-up (start from existing content, organise what's actually there). Most large systems need both — top-down for the strategic shape, bottom-up for the messy reality of legacy content.
Origin
Richard Saul Wurman coined the modern usage of 'information architecture' in 1976. The web era formalised the practice through landmark texts by Rosenfeld and Morville (Polar Bear Book, 1998).
How it works
- Inventory existing content — what's actually there?
- Define user mental models — how does the audience think about this domain?
- Card sort with target users to understand grouping logic.
- Design the hierarchy — primary navigation, secondary, breadcrumbs, search.
- Define a labelling system that uses user language, not internal jargon.
- Tree-test the proposed IA with users before locking it.
When to use it
Use when
- Designing any system with more than ~20 distinct content items.
- When user research shows people can't find things.
- Before any major redesign or new product launch.
Skip when
- For very small single-page products.
- When IA is being driven by org charts instead of users.
Key metrics
- Tree test success rate (target 70%+)
- Search-to-navigation ratio (heavy search use signals weak IA)
- Time-to-find for representative tasks
- Support tickets about 'where do I find...'
Examples
- We tree-tested the proposed IA with 25 users; success rate was 42%. Rebuilt it; second round hit 84%.
- The IA was organised by the team's org chart, not the user's mental model. That's always a tell.
- Search volume on our help docs told us the IA was failing — users were giving up on navigation.
In practice at Makreate
Makreate UX engagements for content-heavy products always include an IA phase. We card-sort with real users, design the proposed structure, then tree-test before any visual design work begins. The discipline catches IA problems while they're still cheap to fix — once visual design and engineering are underway, IA changes get expensive fast.
UX Design →Common mistakes
- Organising IA by internal teams instead of by user mental model.
- Using internal jargon for labels.
- Skipping card sorting and tree testing.
- Designing the IA after visual design is locked.
- Adding categories instead of structuring properly — bloat is a sign of IA failure.
Frequently asked
What's a tree test?
A usability test where you give users a task and ask them to find it in your proposed IA hierarchy — without any visual design or navigation chrome. It isolates whether the structure works.
How is IA different from navigation design?
IA defines the underlying structure; navigation is one of the surfaces that exposes the structure. Same IA can be exposed via multiple navigation patterns.
When do I need a card sort?
When you're creating or restructuring categories. Card sorting reveals how users naturally group your content, which becomes the basis for the IA.