People are trained not to call your baby ugly. You need to make them feel safe to do this. My approach was to explain that the worst thing that could happen to me was building something people didn’t care about, so the best way they could help me was absolute, brutal honesty.
My other favorite is what he says to do after the interview: Look for patterns and apply judgment.
Customer development interviews will not give you statistically significant data, but they will give you insights based on patterns. They can be very tricky to interpret, because what people say is not always what they do.
You need to use your judgement to read between the lines, to read body language, to try to understand context and agendas, and to filter out biases based on the types of people in your pool of interviewees. But it is exactly the ability to use human judgement based on human connections that make interviews so much more useful than surveys.
The way I do this in my own work is to collect my notes on a series of interviews and then review them one at a time, recording the incidence of certain problems, requests, needs, characteristics and other concepts. Extracting structured information from unstructured notes, though (as Giff says) not statistically valid, does help to bring out patterns. See my entry on qualitative vs. quantitive research methods for more on this and what to do next.