Ask people about a product and they will usually be kind. They like the idea, they would probably use it, the design looks clean. Then the product launches and they do not come back.
What people say and what they do are different kinds of evidence. Opinions are shaped by politeness, by the question and by the moment. Behaviour is harder to fake.
So we test by watching. We give someone a task the product is meant to solve, one goal per session, and ask them to complete it without help. Then we stay quiet.
We note where they stop, where they go back, where they hesitate before a click. Those moments point to the parts of the product that do not yet work, even when the person tells us afterwards that everything was fine.
A few sessions like this are often enough to show the pattern. The fixes that follow are specific: a step to remove, a label to change, a feature nobody looked for. Praise rarely leads to a change; hesitation almost always does.
A product people use is built on what people do. Watching them is the most direct way we know to get there.