I have one concern with collection of data for problem validation; how to avoid that the data is biased?
For example, the people who agree to talk with you is likely those that have a problem. So if you talked with 5 people and they all say they have a problem, you might still risk that the 95 people that said no to a conversation do not have a problem. Can you still conclude there is a problem?
Or when I walk at a conference about security, where people will be who are aware of security and concerned about it, can I then really conclude that there is a problem?
I need an initial representative set. Normally you would choose 1000+ by random. Selected without bias. Then interview as many as possible, assuming the ones that won't talk have no problem.
Or maybe I'm over analysing here?