"Spam/fake accounts do represent less than 5% of users" This claim raises lots of open questions. If they are able to determine the fake users, why they did not just delete these users? Don't they have such an spam detection and prevention process?
Also I think it is a lot more than %5. Also, again, I think it is still OK only if the half percent of the users represent the real (not-spam) users. The spam users are increasing and becoming much harder to determine when the user base grows. Even if you're running a simple blog and you notice lots of spams in your comments when you get some traffic. %5 is just not a realistic value.
Also lets not forget about "the paid" users. There is such a thing behind the scenes and they are not spam/fake.
If you can do a statistical sample of a very small number of users, say 0.1%, and do very expensive detailed investigation of them that determines 5% are fake, you can easily extrapolate that 5% to the entire userbase with small confidence bounds, but have no idea which of the rest of the users are fake.
5% of monetizable users - a crucial distinction, since it completely ignores bots which might represent 80% of the accounts while not being monetizable
You can in principle conduct user interviews to determine the likely percentage, without having a clear way to ide tify every individual account. Similar in principle to how you conduct pre-election polls.
Also I think it is a lot more than %5. Also, again, I think it is still OK only if the half percent of the users represent the real (not-spam) users. The spam users are increasing and becoming much harder to determine when the user base grows. Even if you're running a simple blog and you notice lots of spams in your comments when you get some traffic. %5 is just not a realistic value.
Also lets not forget about "the paid" users. There is such a thing behind the scenes and they are not spam/fake.