Re: Personally
Sample size requirements are hard to "intuit"
For instance, if 14 people out of a randomly selected sample of 70 are X, and the remaining 80% non-X, you've already got a 95% confidence that the true population frequency of X people is between 12% and 28%, however big the population.
This is why you can get reliable polls even if you ask fewer than 1 in 100,000 people. The randomness of the sample is vastly more important than its size.