Re: Less than 5%!!
Musk claims that Twitter uses a sample size of a hundred accounts. Twitter claims that it's quite a bit more complex than that. It's hard to assess as an outsider as Twitter has the benefit of having lots of backend data to use in its analysis.
For instance, without access to backend data it's hard to know whether an account is active or not. There will be quite a few accounts of people who log in daily and follow topics and hashtags, DM with other users, click on ads etc. (and are therefore considered 'monetizable' by Twitter) but that rarely post themselves. To outsiders those accounts look dead, Twitter knows they are not.
Similarly there might be accounts that look perfectly normal to us but where Twitter knows the account is part of a batch of accounts that are harmless but that they know all come from the same IP address at a university research lab and so might not be actual 'monetizable' users.
Also, there should be a lot more consideration for the fact that the set of 'non-monetizeable' accounts are not homogenous at all. For instance, El Reg has a Twitter account. It has over 100,000 followers so clearly many people believe the account adds value to their experience on Twitter. That means that accounts like that add value to Twitter as a company too. Yet, it makes no sense to consider these types of accounts as 'monetizable' as the people/teams operating such an account will not respond to ads. Twitter will not include them in their count of 'Monetizable Monthly Active Users' but will, at least internally, not lump them together with spam accounts, political bots etc.
All in all this means that Twitter has the ability to do much more accurate analysis than any outsider ever has. Whether Twitter Inc. are truthful in their statements or not is a separate matter but analysis by outsiders, which can be very interesting and insightful, will likely always suffer from a greater inaccuracy than Twitter's internal data.
Biting the hand that feeds IT
