Not as big as you might think. Your sample must be large enough that you can claim with a high degree of confidence that the results are representative of the entire group under investigation. Three factors affect sample size: the degree of confidence you want to be able to place in the results, the margin of error you are willing to accept, and finally, the total size of the population being researched.
While higher is often better, there is no acceptable response rate, as such. In fact, more important than response rate is the representativeness of your sample. Of greater concern than response rate is whether, due to some bias, respondents differ in some significant way from those that have not responded. You could, for example, have a 50% response rate and discover that for some reason you hadn’t heard from any women. In this case you would be better off with a 20% response rate that included a balance of men and women.
Even a very low response rate can be acceptable if the sample is representative of the population you need to hear from.
For detailed information see:
“How to conduct Customer Surveys” – Appendix B: “Sampling – More Advanced Topics”