Testing five users uncovers roughly 85% of usability problems — the highest-return research bargain in product development. Information architecture — how content and flows are organised — determines whether users can find anything at all. Together they answer the two questions every product lives or dies on: can people find it, and can they use it?
Usability testing: the method
Show real people real tasks, watch where they struggle, fix, repeat. The discipline that makes it honest:
- Task scenarios, not feature tours: 'book a table for four tonight' — never 'try the booking feature'; users behave naturally toward goals, politely toward demos
- Think-aloud protocol: participants narrate their thinking; the narration is the data
- The facilitator's rule: never help, never lead, never apologise for the design — the struggle IS the finding
- Record everything: the screen, the face, the timestamp of every hesitation
Moderated versus unmoderated
- Moderated (a facilitator present): richer data, immediate probing, 5–8 users per round — the gold standard for discovering *why* users struggle
- Unmoderated (remote tools, recorded sessions): faster, cheaper, 10–20 users — better for validating specific flows across devices
The sequence that scales: moderated tests when questions are open, unmoderated when questions are specific.
How many users, honestly
Five users per round is the classic answer and remains right — but the nuance matters: five per audience segment. A product serving doctors and patients tests five of each. And rounds beat volume: three rounds of five users, fixing between rounds, finds more than one round of fifteen.
Information architecture: the organisation
IA is the structural design of a product: how content is grouped, labelled, and navigated. Users experience IA every time they decide where to click — bad IA means everything else fails quietly.
- Organisation systems: how items group — by task, audience, topic, or chronology; most products mix several
- Labelling systems: the words that stand for categories — 'Products' versus 'Solutions' versus 'Shop' is an IA decision with conversion consequences
- Navigation systems: the paths — global nav, local nav, search, breadcrumbs — see the UX audit checklist for the diagnostic layer
The IA research methods
- Card sorting: users group content into categories that make sense to them — the input for structure
- Tree testing: users find items in a proposed structure without visual design — the validation of structure before pixels
- First-click testing: where users click first predicts task success at roughly 80% — the cheapest predictive test in UX
The sequence: card sort to draft the structure, tree test to validate it, first-click tests to check live navigation.
Why the pair belongs together
Testing surfaces symptoms ('users cannot find settings'); IA names the disease ('settings live in two places with different labels'). Teams that test without IA work fix symptoms forever; teams that redesign IA without testing ship opinions. Our design services run the pair as one practice — research informing structure, structure informing testing. For how the artefacts flow into design work, see wireframes versus mockups; for a scoped engagement, book a consultation.
Frequently asked questions.
How many users do you need for usability testing?
Five users per round uncover roughly 85% of usability problems — but five per audience segment, and rounds beat volume: three rounds of five with fixes between find more than one round of fifteen. Unmoderated remote tests scale to 10–20 for flow validation.
What is information architecture in UX?
The structural design of a product: how content is grouped (organisation), named (labelling) and reached (navigation). Users experience IA every time they decide where to click — poor architecture makes every other design effort fail quietly.
What is card sorting?
A research method where users group content into categories that make sense to them — the raw input for information architecture. Open card sorting lets users create their own categories; closed sorting assigns content to predefined ones. Follow with tree testing to validate the structure.
Moderated or unmoderated usability testing?
Moderated (facilitator present, 5–8 users) delivers richer data and probes why users struggle — right for open questions. Unmoderated (remote tools, 10–20 users) is faster and cheaper — right for validating specific flows across devices. Scale: moderated when exploring, unmoderated when confirming.
What is first-click testing?
Users click where they would start a task before navigating — and the first click predicts eventual task success roughly 80% of the time. It is the cheapest predictive test in UX: quick to run, immediate to analyse, and it validates navigation without any design work.
Enjoyed this? Let's talk about applying it to your product.
