Answer:
The answer to this question depends, in part, on the kinds of questions that you want to ask them. If the purpose of the survey is to try and figure out what concert patrons are thinking, then you want to try to get a good variety of ideas. In that case, you're no so interested in having a sample of individuals that's exactly representative of the population of concert goers. You only want to get a wide variety of the ideas that are out there. This is called heterogeneity sampling. That is, in this case, we'll be trying to get a sample not of people, but of ideas. In this case, we'd use brainstorming groups, panel sessions, and other group discussion methods to get all the ideas out there.
This approach is not an appropriate one if the purpose of the questions is to determine the preferences of the population of theater goers. That's because it doesn't communicate the popularity or prevalence of ideas in the sample. If this is a problem, and it likely is, it seems more appropriate to use some sort of purposive sampling method like modal interest sampling. This approach requires determining what the typical concert goer is like . If the company determines that the typical (or modal) attender is a young single person in college, then those people are sought out and interviewed. This is the sort of polling that is used in election polls where they interview "typical voters." For different genres of music and concert, a different typical concert goer could be postulated and different surveys and interview styles determined.
Probably most appropriate to this problem, though, is quota sampling. This form of purposive sampling uses demographic and other information to determine how many of different kinds of people to interview. For instance, if it is determined that only 20% of concert goers for a particular concert were women, then only 20% of the nonprobability sample should be made up of women. As long as these quotas are met, the sample can be sufficiently random.
I recommend using a combination of quota sampling and modal instance sampling. It seems to me that the kinds of people who attend concerts are idiosyncratic in some ways. Inasmuch as we can define universal characteristics of concert goers, we should seek to include only those people in the sample, but since we have demographic data as well, we should use these proportions as quotas as we select typical concert goers. For example, suppose we determine that the vast majority of people who attend operas make over $80,000/year, then we should limit our sample to people with that income level, but if we also know that 80% of opera attenders are women, we should work to insure that not more than 20% of our sample is men.