sts working on testing people, but we were in the same small group. So sometimes I'd get to sit in the room behind the glass and watch somebody being tested. So I learned a lot about that process. And the most surprising thing I learned is if you're testing whether people can use a user interface, what do you think is the right number of people to test? Let's compare that to an opinion poll where I don't know what the number is, but you probably want a thousand people. Would you feel comfortable if you talk to a thousand people? That would be a pretty good opinion poll, I think, nationally. But if you're testing people for a user interface, how many do you have to test? Do you have to test a thousand? Nope. Five, maybe one. Because if that person says, "Ah, I'm looking all over this page and I don't see a button." And then you bring in the next person and they say, "I don't even see the button." And then you bring in the third, fourth, and fifth, and they all say, "I don't know where the button is." You don't have to wonder if those five people are coincidentally messed up and they're the only ones who can't use your user interface. You have guaranteed that it's unusable. Four out of five, three out of five guarantees it's unusable. You have to redo it.

So if you're looking at how people feel, and this is maybe the analogy is not perfect. I realize that. But if you look at how