And really, one can learn so much about what people want without having them use anything. User context interviews are hugely helpful. But my favorite thing is real-world tests even before there's a product. If most of your sales will come via people clicking on an ad and looking at a landing page, test that first! If people won't sign up, there's no point in building anything after that.
A good example comes from my cofounder at a startup we did in 2010. At the time, we had theories about an app that people would mostly discover through their Facebook feed. We used Greasemonkey to do user tests and found out that although some liked our idea, most people hated it. So we threw that out and did something else. But two other startups went on to build the same idea and fail with it. We estimate we saved $2m in learning what they learned. Ignite talk here: https://vimeo.com/24749599