🗂️Information Architecture
How content is structured, named, and findable. The unsexy layer that determines whether your product feels coherent or chaotic.
Bad IA shows up as users complaining 'I can't find anything' or support tickets stuck on the same questions. Good IA is invisible. Getting it right early is far cheaper than refactoring it after a million users.
Information architecture is the structural design of an app's content and navigation. The PM owns the IA at the same level as they own the user flows — bad navigation hierarchy is bad product, not just a UX nit.
What IA includes
- Navigation structure — top nav, side nav, breadcrumbs, modals
- Naming — how features and sections are labeled
- Grouping — what goes together, what stays separate
- Hierarchy — what's at the top level, what's a sub-page
- Search and findability — how users locate things they can't navigate to
The two failure modes
Too flat. Everything at the top level. 30 nav items. Users can't decide where to go.
Too deep. Five clicks to get to a common task. Users abandon.
The right answer is usually 3-7 top-level concepts, with 2-3 levels of depth max. Use card sorting (literal or digital) with real users to validate the grouping.
Diagnostic signals
- Support tickets asking "where do I find X?" — IA failure
- Users using search instead of navigation for common tasks — IA failure
- Multiple paths to the same place — usually fine, often necessary
- Feature areas that nobody uses despite the feature being good — could be IA hiding it
How to refactor IA without breaking everyone
- Map the current state. Inventory every nav element and content type.
- Card-sort with 10-15 users. Watch where they group things.
- Test the proposed structure with another cohort using a tree test (Optimal Workshop is the standard tool).
- Roll out gradually. A/B test for power users, give them the option to revert, watch task-success metrics.
- Don't change labels and structure simultaneously. That doubles the cognitive load.
IA refactors are deeply unfun politically — every team has opinions about where their feature lives. Get exec sponsorship before you start.
Real-world examples
Slack's left-rail refactor in 2021 — moving DMs, mentions, threads, and apps into a more structured hierarchy — was a multi-quarter IA project. The PM team ran extensive card sorts and tree tests. Some power users hated the change initially; metrics improved task completion materially over 90 days.
Go deeper — recommended reading
Interview questions (1)
Q1Your product's navigation has grown to 25 top-level items over 5 years. How do you fix it?executionsenior▼
Six-step plan:
- Inventory. List every nav item, who uses it (data), and how often.
- Customer research. 10-15 card sorts. Watch how users naturally group your concepts.
- Propose a new structure. Aim for 5-7 top-level concepts, 2-3 levels deep.
- Tree-test it. Optimal Workshop. Test task success against the current structure.
- Get exec sponsorship and team buy-in — every team owner of a current nav item will resist.
- Roll out with a "switch back" option for 30 days; track task success and complaints; commit fully when the data supports it.
The hard part isn't design — it's politics. Don't start without sponsorship.