In the most layman of terms, when the data variance stops doing this:
View attachment 283668
And does this:
View attachment 283669
You have an adequate sample size, assuming the data is arranged like this:
View attachment 283670
One problem with survey data is that it doesn't jive that well with parametric statistics, and non-parametric analysis is weaker. I don't have a problem with survey data, it can be very useful. It's just extra easy to misinterpret inference and results. It is true that at a certain point more data does not make an analysis more accurate in a detectable or meaningful way. It depends on what the cause of the variance is.
I personally consider survey data like triage. It's great for highlighting large scale changes or points for which deeper study is warranted.