I imagine one of the biggest problems they have with the whole survey thing is selection bias and actually asking the people who are most likely to be their customers. If they had an online survey asking "How important is it that your phone run an open OS?" and that got linked from slashdot, they'd get an incredible response in the affirmative, but they'd have no way to tell if this is really that big of a deal to the majority of their target market. I'm sure a huge part of the work that their "survey contractor" does is making sure the survey gets into the hands of a diverse group of people who represent their target market.

-John