The basic deal here is this: Walmart did a customer survey with a faulty question. It had bad wording and implied the answer that WM assumed to be true. So people answered Yes, and Walmart acted on the info, spending millions to revamp stores -- only to discover that sales went down, not up. Down by a billion. Whoops.
Some of the comments dispute the story, others cite articles that uphold it. At any rate, the lesson stands: If you want to do a customer survey, the way you word questions is extremely important! It's not enough to get a few staffers together, write some questions, and send them out. (OK, if you're asking teens whether they'd prefer pizza or pretzels at their next movie night, or something equally simple, go for it. However-- ) If you're trying to find out about something really important, something that's going to change your business model or your strategy, something that you're putting a lot of money into, then you need some expertise in this area to do it right.
At the very least, you need library outsiders to test your questions on before you send your survey to the masses. Do people interpret them the way you think they will? Are you using lingo that they don't understand (or that they have different perceptions of than you do)? Are the queries phrased in such a way as to invite a certain answer? If so, you could be polluting your own data before you even get it.
Walmart can lose a billion dollars and survive. You can't. Be very careful about how you ask for, and interpret, customer data!