The n for the group is only 246 though. That’s very biased depending on where the stats are from.

Very bold to claim it's a biased sample without any evidence or even a theory of how.

The evidence is the n being 246 lol

jmye a day ago | parent [-]

What, specifically, would be an optimal sample size and why, specifically, is 246 insufficient, in this case? “Lol”

	▲	sanex a day ago \| parent [-]
		I'm not a statistician but when I know when I've only gotten 250 users through a split test it's not even worth looking at the results. So just for you I'm going to say 10,000 and geographically spread from more than one county in Ohio which contains a University.