He keeps referencing total number of user accounts. This is a bad number to track, precisely because of fraud, duplicate, and unused accounts. A better number is something like monthly or daily active users. Then you can randomly sample a subset of these users and manually determine the number of fake accounts to give you a confidence interval on the total number of real daily active users, which for investor/shareholder purposes, is much better to use than total accounts anyway.
I go into more detail in the report I wrote in January, which also points out that the perfectly constant DAU/MAU relationship Facebook reports cannot possibly be true as growth continues to slow/reverse. See https://www.plainsite.org/realitycheck/facebook.html.
