Interview your participants. Take an interest in what they're interested in. Spend time building scenarios to help them play a role. Mix a little bit of market research into your user test.
What was most interesting for me was the anecdotes that compliment this article.
Jared also observed that trends still emerge when participants don't do exactly the same tasks and instead have some input into what they do. The added bonus here is that this investment in the activity brings greater passion, or desire to succeed, and therefore more representative behaviour.
I've always employed a similar approach when working with prospective students, but need to employ this approach more often in other usability tests I do.
The design recommendations seemed solid, yet sales had dropped 23% immediately after the changes were made. The recommendations came from a well-constructed set of usability tests... What they didn't know -- what they learned later -- is they had done everything right, almost. They'd recruited the right users, facilitated the test properly, and analyzed the results effectively.
There was only one problem: the tasks didn't match what real users do with the site.
Interview-based tasks: Learning from Leonardo DiCaprio - article by Jared Spool
(The Leonardo DiCaprio bit is a reference to one of Jared's test anecdotes.)