Twitter miscounted its daily users for three years straight | Hacker News Reader

Twitter miscounted its daily users for three years straight

theverge.com

30 pointsdkobia14 comments

> crack down on spambots, and verify all users

This is the part that terrifies me. I enjoy my 'alt' accounts that have an anime waifu avatar and an assumed name. I don't want Twitter outing my real identity. That would ruin Twitter in a devastating way.

Outing your identity to whom? You must know that Twitter already knows about your alt accounts right? I wouldn't panic just yet, they will probably just derank unverified users in the replies of tweets so spam bots can finally die.

> Outing your identity to whom? You must know that Twitter already knows about your alt accounts right?

They aren't concerned about Twitter, Inc. knowing. They don't want other people on Twitter knowing.

But verifying users doesn’t require publishing their names on their profile page.

> You must know that Twitter already knows about your alt accounts right?

As the article says, the whole miscalculation comes from Twitter counting all your alts as active even when you were only using one of them...

Sure, just a "miscalculation". They would never want to mislead people into thinking they have more users.

Wouldn't that fact be very important for a valuation of the company?

Perhaps this was the source of the panic? Would user numbers be subject to auditing in cases like this?

I wouldn't be surprised if Musk turns around and lowers his bid. At least that's what I would do. But maybe he wants the company so much that he won't contest it.

Musk’s offer explicitly didn’t demand this kind of data. I assume he figured that Twitter’s statements about its finances, number of users, growth, etc. were reliable and he didn’t need to do any additional due diligence. I thought it was reasonable at the time, and now I think he’s feeling some regret.

DAU and MAU are tricky metrics partly because even when they're known to be problematic (perhaps the definition of "active" is too permissive to be useful for certain purposes) they are very hard to change because they are reported at such a high level (frequently without relevant context such as whether or not they are counting accounts or users) so historical changes are trickier to get support for.

Also people want that number to be defined to yield the highest number because Goodhart's Law and we ship what we're incentivised to ship.

Does this explain why followers have been fluctuating in a very unexpected way for high-profile accounts?

It's weird that it has been in both directions, with some accounts losing large numbers of followers but others making sudden large gains.

The reports I’ve seen on this implied that (in general) left-leaning accounts were losing followers, while right-leaning accounts were gaining them.

Given that the coverage we’ve seen in various quarters (where the left generally fall into the “Elon buying Twitter is bad” camp and the right fall into the “good” camp), the changes in follower counts seem to be easily explained by a rash of rage-quitting left-wingers and a corresponding rush of right-wing Gab/Parler users giving Twitter another chance.

It could explain it, but it could also be something else. Without twitter disclosing there's no way to know.

Reply on news.ycombinator.com