The 5 people rule is only valid for testing, not interviews. It’s because “people are people” when trying to manipulate designs — which inherently offer limited methods of manipulation. As shown in the Nielsen link, you can catch 80% of your problems with only 5 users and 100% with 15 users. Sample-sizes for the types of interviews talked about in this article are generally larger and harder to predict. I’ve always followed “you’ll know you’ve interviewed enough when you stop hearing new answers.”
