If you collect nothing else in a usability test it should be a list of problems encountered by users.
It seems so simple yet there is a rich history of how many users you need to test, what constitutes a problem and which method to use.
Via Peter Shanks