![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
walkitoutI ask this question with a _degree_ of sincerity, but mostly as a way to frame thinking about the Musk / Twitter back and forth.
I would hope that no one is still wandering around thinking that every account posting on the interwebs is a distinct adult human being engaging in thoughtful discourse. I mean, that cartoon about on the internet no one can tell you’re a dog thing is oooooold at this point. I had multiple accounts on a bb at the UW back in the 1990 +/- time frame. I helped my son set up a FB account when he was too young to do so. This is old, and not all of it is a problem per se.
Twitter statements about daily active user calculations have _already_ been modified once in the Musk / Twitter deal era. That has largely been treated as irrelevant / unrelated to the deal, however, I suspect that revision occurred as a result of some discussion between Twitter and Musk, and follow-on internal investigation by Twitter. The impact on the numbers was minor, and the adjustment very technical (basically, Twitter lets at least some users link accounts and makes it easy to switch between them. I hope this is for brand managers and people managing institutional accounts for government agencies, corporations, etc. Twitter’s adjustment to the daily active count was an error they were making. _Any_ update to a _single_ account counted as a distinct update to _all_ of the linked accounts. That’s clearly wrong. I don’t know anything other than what I read in the filing.)
Matt Levine has been skeptical — to be fair, he was probably pretty angry about having to create a column on one of his days off — since Musk skipped due diligence that anything could possibly come up that would let Musk out of the deal. I’ve really wondered about that 5% +/- calculation, and have been spending some time trying to figure out whether I believe it or not. I _don’t_ believe it. If I had to pick between 0% fake accounts, 50% fake accounts and 100% fake accounts, I’d pick 50% as the closest guess, because I think the actual number of fake accounts is pretty close to 1/3rd. This is _daily active_ — not all accounts, which of course I could have no visibility into at all, whereas daily active is the sort of thing that you could at least pretend to get a feel for over time.
If you showed me some mildly plausible data collection and mildly plausible calculations that led you to believe the fake accounts was closer to 2/3rds, I’d read with interest and probably say, sure, that makes sense.
Levine’s attitude in general is, in order for this thing to be _material_ it couldn’t be off _slightly_ from the 5% — the filing says that it could be larger or smaller. However, I think that if it’s anything more than a quarter, the deal is off. Period. And the reason why the deal is off is pretty simple. It’s because the financing walks. And the financing walks because at a quarter fakies, twitter’s ad rates collapse, and there is literally _no business there at all_.
Also, twitter has not been forthcoming with academics and others who are trying to quantify the fakies, and some very reputable academics with relevant expertise are out there saying it really could be like 35% of daily actives are fakes. There is reporting saying that internally, there is strong resistance to doing anything meaningful about fakes, and it makes sense that that would be true. If you can get a quarter or a third or a half or two thirds or double triple quadruple you name it more eyeballs to convince ad buyers to pay more, why would you chase them away?
Advertising business models are super weird. Especially during the years of transitioning from paper to online, the paper side really _wanted_ numbers receiving papers to prop up ad rates. That’s largely gone by on the paper side. But the problem with the model remains. There’s a powerful incentive to misrepresent eyeballs. If Twitter has a really big bot problem — not just a crypto bot problem, but everything bot problems — in its daily active user number, then the deal is off, twitter’s stock price will nosedive, the SEC will take an interest and there will be shareholder lawsuits.
On the other hand, maybe _then_ we’ll finally get some kind of meaningful approach to dealing with bots.
ETA:
Last year, when a whistleblower drew the SEC’s attention to related issues at FB.
https://arstechnica.com/tech-policy/2021/10/facebook-misled-investors-about-shrinking-user-base-ex-employee-alleges/
In 2017 when Snap did it:
https://www.cnbc.com/2017/04/12/snap-lawsuit-ceo-evan-spiegel-said-misreported-metrics-no-big-deal.html
I would hope that no one is still wandering around thinking that every account posting on the interwebs is a distinct adult human being engaging in thoughtful discourse. I mean, that cartoon about on the internet no one can tell you’re a dog thing is oooooold at this point. I had multiple accounts on a bb at the UW back in the 1990 +/- time frame. I helped my son set up a FB account when he was too young to do so. This is old, and not all of it is a problem per se.
Twitter statements about daily active user calculations have _already_ been modified once in the Musk / Twitter deal era. That has largely been treated as irrelevant / unrelated to the deal, however, I suspect that revision occurred as a result of some discussion between Twitter and Musk, and follow-on internal investigation by Twitter. The impact on the numbers was minor, and the adjustment very technical (basically, Twitter lets at least some users link accounts and makes it easy to switch between them. I hope this is for brand managers and people managing institutional accounts for government agencies, corporations, etc. Twitter’s adjustment to the daily active count was an error they were making. _Any_ update to a _single_ account counted as a distinct update to _all_ of the linked accounts. That’s clearly wrong. I don’t know anything other than what I read in the filing.)
Matt Levine has been skeptical — to be fair, he was probably pretty angry about having to create a column on one of his days off — since Musk skipped due diligence that anything could possibly come up that would let Musk out of the deal. I’ve really wondered about that 5% +/- calculation, and have been spending some time trying to figure out whether I believe it or not. I _don’t_ believe it. If I had to pick between 0% fake accounts, 50% fake accounts and 100% fake accounts, I’d pick 50% as the closest guess, because I think the actual number of fake accounts is pretty close to 1/3rd. This is _daily active_ — not all accounts, which of course I could have no visibility into at all, whereas daily active is the sort of thing that you could at least pretend to get a feel for over time.
If you showed me some mildly plausible data collection and mildly plausible calculations that led you to believe the fake accounts was closer to 2/3rds, I’d read with interest and probably say, sure, that makes sense.
Levine’s attitude in general is, in order for this thing to be _material_ it couldn’t be off _slightly_ from the 5% — the filing says that it could be larger or smaller. However, I think that if it’s anything more than a quarter, the deal is off. Period. And the reason why the deal is off is pretty simple. It’s because the financing walks. And the financing walks because at a quarter fakies, twitter’s ad rates collapse, and there is literally _no business there at all_.
Also, twitter has not been forthcoming with academics and others who are trying to quantify the fakies, and some very reputable academics with relevant expertise are out there saying it really could be like 35% of daily actives are fakes. There is reporting saying that internally, there is strong resistance to doing anything meaningful about fakes, and it makes sense that that would be true. If you can get a quarter or a third or a half or two thirds or double triple quadruple you name it more eyeballs to convince ad buyers to pay more, why would you chase them away?
Advertising business models are super weird. Especially during the years of transitioning from paper to online, the paper side really _wanted_ numbers receiving papers to prop up ad rates. That’s largely gone by on the paper side. But the problem with the model remains. There’s a powerful incentive to misrepresent eyeballs. If Twitter has a really big bot problem — not just a crypto bot problem, but everything bot problems — in its daily active user number, then the deal is off, twitter’s stock price will nosedive, the SEC will take an interest and there will be shareholder lawsuits.
On the other hand, maybe _then_ we’ll finally get some kind of meaningful approach to dealing with bots.
ETA:
Last year, when a whistleblower drew the SEC’s attention to related issues at FB.
https://arstechnica.com/tech-policy/2021/10/facebook-misled-investors-about-shrinking-user-base-ex-employee-alleges/
In 2017 when Snap did it:
https://www.cnbc.com/2017/04/12/snap-lawsuit-ceo-evan-spiegel-said-misreported-metrics-no-big-deal.html