How large should my survey be?

I need to conduct a 10min long marketing survey with 6 to 7 questions . How large should my sample size be to have results representative of a 1000000-large population group, considering I want to achieve a 95% confidence level?

I have read the formulas:

$\displaystyle n = ({z \sigma \over E })^2$

but how can I reconcile the concept of a questionnaire with the standard deviation of the sample population $\displaystyle \sigma $ and the error E? Is there even a mean or proportion to talk about? The only unambiguous part is z = 1.96 for 95% confidence.