You have a problem when you have to explain the 5 users to a marketing person. Marketing people live with statistics and large samples, so they can have hard time buying 5 users.

It is kind of obvious that testing more users will almost definitely bring out more problems. Nielsen himself states at the end of his post that in certain cases you need more than 5 users. I think that Faulkner’s study confirms Nielsen and Landauer’s conclusion – her study shows that a 5-user sample is likely to bring out ~ 85% of the problems. The chart that you use in this post is misleading – the variance in results is shown in dark gray, which is hard to spot on the dark background.

Given the state of software and the web, I readily subscribe to the 5-user rule of thumb. I am a practically-minded person. We’ve done hundreds of usability tests for various clients in different domains. When we have one user group, we usually test with 6 people. When we have two groups, we would test 8 or 10 users, and when we have more groups, we aim at 3-4 users per group. We’ve never had a case not to find any usability problems. Our average report lists 50-60 problems, and our clients address 50 to 80 % of these problems. From a business point of view, testing more users does not make sense.