Usability Testing

Usability testing is a method in which people who resemble the intended users attempt defined tasks with a design while an observer watches and, usually, asks them to think aloud. It is evaluative rather than exploratory: the design already exists in some form, and the question is whether it works. The facilitator gives a task and a realistic scenario, then stays quiet, resisting the strong urge to help. What is recorded is behavior — where people hesitate, what they click, what they misread, where they give up — alongside their commentary. The output is a prioritized list of problems, not a verdict or a score.

Testing exists because designers and engineers cannot un-know their own product. Having built the structure, they can no longer perceive an ambiguous label or an unexpected sequence, because they already know the answer. This is not a matter of skill; expertise reliably removes the ability to simulate a newcomer. Testing exists too because internal disagreement is otherwise unresolvable: arguments about whether a flow is confusing can continue indefinitely until three people in a row fail at the same step, at which point the argument ends. And it exists because fixing a discovered problem before release is dramatically cheaper than supporting it forever afterward.

• →Five people attempting to change a delivery address in a checkout, with the facilitator recording where each hesitates.
• →A test on paper before any code exists, checking whether the sequence of questions makes sense.
• →A remote unmoderated study where participants record themselves completing tasks, useful for scale but weak on why.
• →A test with participants using screen readers or switch access, which routinely finds failures sighted testing misses.

Testing method comes from experimental psychology and human factors, where measuring operator performance on defined tasks was standard practice. The think-aloud protocol used in usability work derives from protocol analysis in cognitive psychology, notably the methods described by K. Anders Ericsson and Herbert Simon. Jakob Nielsen argued that small samples find most problems and that frequent cheap tests beat rare expensive ones, a position that made testing practical for ordinary teams. Steve Krug's Don't Make Me Think and Rocket Surgery Made Easy pushed further, advocating regular sessions with a few participants, run by the team itself, with fixes applied immediately rather than compiled into reports.

Remote testing platforms have made recruitment fast and geography irrelevant, and unmoderated studies allow larger samples at low cost. The trade-off is context and depth: you see a screen and hear a commentary, but you do not see the room, the interruption, or the document the person needed and could not find. Analytics, session replay and experiments have partly displaced small-sample testing in organizations that prefer numbers, though these answer different questions — what and how often, rather than why. Consent, recording and privacy obligations have grown, and accessibility testing with people who use assistive technology is increasingly treated as a required part of the practice, not an add-on.

Steve Krug's central argument is that testing does not need to be an event. A team can test three people on a Thursday morning, watch together, agree on the single most serious problem before lunch, and fix it the same week. The value is not statistical; five participants will not tell you what proportion of users struggle. What they will tell you, reliably, is that a label is ambiguous or a step is missed, because the same failure appears in the second and third session. The common failure mode is the opposite pattern: a large formal study, a long report circulated weeks later, and no change to the product.

• →Test tasks, not opinions — ask people to do something, never whether they like it.
• →Recruit people who resemble real users, including those who struggle with technology.
• →Stay quiet; the instinct to help destroys the finding you came for.
• →Small samples find problems; they cannot measure how widespread a problem is.
• →Watch what people do and weigh it above what they say about what they would do.
• →Test early and often with fixes applied between rounds; a report nobody acts on is wasted effort.