A usability test observes people attempting representative tasks with an interface.

The goal is not to ask whether participants like the design. It is to discover where they hesitate, misunderstand the system, take inefficient paths, make errors or fail to complete the task.

A test needs a task goal stated without giving away the interaction path. “Change the shipping address for this order” is more informative than “click Edit, then open Address.”

Useful observations include task completion, time or effort, errors, recovery behavior and comments that reveal the participant's current mental model. The facilitator should avoid rescuing the participant too early, because the point where help becomes necessary is evidence.

Participants should resemble the users and context that matter for the design. Testing only the development team measures familiarity with the implementation rather than ordinary learnability.

A small test does not estimate every population statistic, but repeated observation can expose high-impact interaction failures early enough to revise them.