Usability testing

Usability testing gives representative users real tasks and observes what happens. It is observation, not opinion gathering: what participants say about a design is far less reliable than what they do with it.

Moderated and unmoderated

Moderated sessions have a facilitator present who can probe. They produce the richest findings and cost the most. Unmoderated tests run remotely through a tool, are cheaper and reach more people, and cannot follow up on the interesting moment.

The five users question

Nielsen and Landauer’s 1993 model suggested five users find roughly 85% of usability problems, and the figure is quoted far more often than its assumptions are. It holds for one homogeneous user group and one round of testing. Several distinct user groups need five each, and the model assumes iteration between rounds rather than one large test.

The underlying point survives the caveats: small rounds repeated beat one big study.

Running it well

Set tasks as goals, not instructions. “Find out how much it costs to return this item” tests the design; “Click Returns in the menu” tests obedience.

Resist explaining. The urge to help is strong and every intervention destroys the finding. Silence is data.

In practice

Five sessions in a day will tell you more about whether a flow works than a month of internal debate, which is why it settles arguments faster than any other method.