The exact same sample size that would be appropriate for 20 million plus or 2 million plus accounts or 20,000 plus accounts.
It's the sample size itself that matters and the degree to which the sample size is representative i.e. non-biased.
You are viewing a single comment's thread from:
And what size would that be?
It depends on how confident you want to be in your estimate.
You can play with the numbers here, for example:
https://www.surveysystem.com/sscalc.htm