We collect data to evaluate usability using a variety of methods. Quantitative methods can produce useful metrics such as average times to carry out certain tasks or average ratings of satisfaction. Qualitative methods produce richer insights.

Lab Testing

The key idea behind lab testing is to control the environment in which the user is operating, and observe directly what they are doing. This generally provides stronger evidence than asking them afterwards to recall what they thought they were doing.
Video recording of their actions.
Audio recording as they talk through performing a task.
Eye movement tracking, for finer-grained analysis. This produces a map showing where on a screen people are looking most frequently, and the sequence of eye movements. This highlights both what people are looking for and what they are missing.

Interviews

Structured interviews can provide in-depth, qualitative insights.
 

Focus Groups

Focus groups can collect feedback more quickly than individual interviews, in a more natural setting, and allow people to 'bounce' ideas off each other. They can provide a quick means of sanity checking suggestions for change. A video recording is usually taken for off-line analysis. Focus groups can also be used to brainstorm new ideas for an interface.
 

Questionnaires

Questionnaires are a more formal approach, and can be used to collect comments as well as quantitative ratings from users that can be aggregated to produce usability indexes.