In information architecture, why should category labels match users' mental models rather than the organization's internal structure?
answer
- don't ship the org chart
- a label predicts what's behind it
- people's words, not jargon
- distinct from its neighbours
- evidence from desk questions and sorts
basics
~20 sPeople look for things using their own expectations and words, not an organization's departments or jargon. Labels drawn from users' mental models predict what lies behind them; labels drawn from the org chart leave existing features effectively invisible.
solid answer
~50 sA **mental model** is a person's internal picture of how something is organised and what things are called. Navigation works when a label **predicts** what sits behind it for the person reading it. Organizations naturally label by their own structure — “Visitor Services”, “Development Office”, “Admissions” — because that is who owns the content, but visitors neither know nor care who owns it. On a museum ticketing kiosk, a member wanting to renew will not guess that renewals live under “Development Office”; they look for “Members”. The result is features that exist but cannot be found, plus desk queues for things the kiosk could do. Good category labels use people's words, are **distinct** from their neighbours and are **consistent** across screens. The evidence comes from questions visitors ask staff, group names from open card sorts and help requests, checked with a tree test.
go deeper
Recall that a label is a prediction of what lies behind it, and that labels copied from departments or jargon hide features people need.
Explain the properties of a working label — predictive, in people's words, distinct, consistent — and where evidence for people's terms comes from.
Show how you resist org-chart navigation in a real product: gather desk and help evidence, cross-list for different groups, and test risky labels first.
Negotiate top-level structure with departments that each want a slot, grounding the decision in tasks and evidence rather than ownership.
## Mental models A **mental model** is the internal picture a person holds of how a system is organised, what things are called and where they should be. People do not read a navigation structure from scratch; they bring expectations from similar places — other museums, other ticket machines, other apps — and scan for the label that matches what they have in mind. A category label is therefore a **prediction**: it promises what the person will find if they choose it. Information architecture succeeds when those predictions come true for the people who actually use the product. ## Why internal structures leak into navigation - **Ownership**: each department owns some content, and the easiest structure to agree on is one section per department. - **Jargon**: staff use internal terms every day — “concessions”, “Category B admission”, “development” for fundraising — and stop hearing them as jargon. - **Politics**: a top-level slot signals importance, so teams argue for their own section rather than for the visitor's task. - **Accretion**: new features are added wherever their owner sits, not where visitors would look. The result is a structure that mirrors the org chart. It is perfectly logical from the inside and opaque from the outside. ## What makes a label work | Property | Meaning | Kiosk example — works | Kiosk example — fails | |---|---|---|---| | Predictive | Tells the reader what lies behind it | “Collect tickets booked online” | “Pre-sold fulfilment” | | In people's words | Uses the terms visitors use | “Members” | “Development Office” | | Distinct | Does not overlap its siblings | “Buy tickets” vs “Members” | “Tickets” vs “Admissions” | | Consistent | Same term for the same thing everywhere | “Annual pass” on every screen | “Annual pass” here, “Membership” there | | Scoped | Covers what it contains, no more | “School and group visits” | “Learning” for group bookings | Wording style, tone and microcopy rules are a writing concern; the IA concern is whether a label **predicts its contents** and **fits people's grouping**. ## Finding people's words Evidence usually already exists before any formal study: - **Questions at the admissions desk** — what visitors ask for, in their own phrasing. - **Help and complaint messages** — the words people use when they cannot find something. - **Search or voice queries**, where a product has them. - **Open card sorts** — the names participants give to their own groups. - **Other places visitors know** — conventions from similar venues set expectations. The candidate labels are then checked with a **tree test**: do people choose the right label first when given a task? ## When mental models differ or mislead - **Different groups, different models**: members, tourists and teachers may look in different places. Serve the primary users at the top level and **cross-list** items that genuinely belong in two places. - **Novices and experts**: frequent users learn internal terms; walk-up kiosk users never do, so first-time comprehension decides. - **Outdated models**: sometimes people's model reflects a constraint that no longer exists. Designing against it is possible, but it needs a familiar entry label and a deliberate bridge to the new concept, validated with users. ## A quick review checklist 1. For each top-level label, ask what a first-time visitor would expect behind it, then check whether that is what is there. 2. Flag every term that appears only in internal documents. 3. Check each pair of siblings for overlap. 4. Confirm the same thing has the same name on every screen. 5. Test the riskiest labels with first-click or tree-test tasks. A museum kiosk makes the cost concrete. When “Collect tickets booked online” sits under “Admissions — Pre-sold”, visitors who already paid join the desk queue, staff time goes on work the kiosk could do, and the analytics show the feature as rarely used — which can wrongly suggest nobody needs it. Mislabelled features look unpopular, so label problems also corrupt the evidence teams use to prioritise. The principle holds on every platform: a native mobile app, a website and a kiosk all fail the same way when the structure is organised around the people who built it.
- What if different user groups hold different mental models?Find the dominant model among the primary users and build the top level around it. Cross-list items that genuinely belong in two places, so each group finds them where they look, and test labels with each group rather than only the largest one.
- Should you ever design against users' current mental model?Sometimes — when their model reflects a constraint that no longer exists or the product introduces a genuinely new concept. Then you bridge: a familiar label as the way in, the new concept explained inside, and a test showing people actually make the crossing.
saying these in an interview costs you the question
- Navigation should mirror the organization's departments so content has clear owners.
- Internal terms are fine because users learn them after a visit or two.
- Clever branded category names make navigation easier to use.
- If a feature appears somewhere in the menu, users can find it.