We got a request for proposal recently from a City Government in Washington State. They said they wanted a scientific sampling of 95%. What does that mean? Well, not really much by itself.
What they said they wanted was a scientific sampling and the way they wrote it means they don't really know what they are talking about. This is why people make seemingly good decisions with bad data. You can have a scientific study with 95% confidence and HUGE variability due to a small sample size.
It is important to look at sampling variability. However, it is easier to remember that larger samples are more reliable and you should probably have n=400 or more surveys. At n=400 you have +/-5% at the 95% confidence level. This means if you replicate the study 100 times, in 95 out of 100 times your answers will not vary by more than plus 5% or minus 5% - that's a ten point swing - there is no statistically significant difference between 45% and 52% in this example. The numbers would be said to be "statistically equivalent."
Or said another way, if someone doesn't have more than n=200 surveys completed and they want to talk about "demographics" then you have my permission to give them a Bronx cheer,