If your question is asked entirely free of bias from a truly randomly selected population, you can use standard statistical methods to come up with a number based on your desired confidence interval. On the other hand, if you're an organization with your funding tied to the perception that people support your mission, and you're surveying whether people support your mission, I'm going to need damn near everyone in the population to respond to your survey before I'll buy it.

There isn’t one. Some things are not measurable by this standard and the stats relatively meaningless. 3000 people from a whole continent? Cmon