Twitter estimates bots by hand sampling 100 accounts per day...
"In a May 6 meeting with Twitter executives, Musk was flabbergasted to learn just how meager Twitter’s process was. Human reviewers randomly sampled 100 accounts per day (less than 0.00005% of daily users) and applied unidentified standards to somehow conclude every quarter for nearly three years that fewer than 5% of Twitter users were false or spam. That’s it. No automation, no AI, no machine learning."
Too bad for Musk that (1) the risk of inaccuracy of this estimate is highlighted in their SEC filings in which the estimate was mentioned, and (2) Musk explicitly agreed in the acquisition contract any risk identified in the SEC filings cannot constitute a materially adverse event along with waiving due diligence.
I a sure lawyers will argue that while the SEC filing was accurate, Twitter is not making a reasonable effort. Could Twitter estimate its revenue in an SEC filing by sampling 100 paid invoices and putting in CYA statement saying the estimate could be wrong?
Regardless, seems like Musk will have a hard time reaching a material adverse standard.
> I a sure lawyers will argue that while the SEC filing was accurate, Twitter is not making a reasonable effort.
A reasonable effort at what?
> Could Twitter estimate its revenue in an SEC filing by sampling 100 paid invoices and putting in CYA statement saying the estimate could be wrong?
It could. It might be a problem between Twitter and the SEC. It might be a problem that due diligence would founder on. It might be a problem that would constitute an materially adverse effect without an explicit contractually agreement that risks disclosed in SEC filings could not do that. These are among the reasons why no sane buyer would sign a merger agreement like Musk signed waiving due diligence and excluding a whole of of specific things from being considered materially adverse events without having done a thorough investigation on their own first.
From a purely statistical perspective, might 100 accounts a day be sufficient? For a quarterly estimate, 9000 samples (assuming the 100 samples are accumulated) seems like it can give fairly tight error bars - if you have 90 samples with a sample mean of 5 and a sample standard deviation of 2 (no idea what the standard deviation is, so consider that a placeholder), then the 99% confidence interval is [4.46, 5.54] according to this website [0]. Not the most precise boundaries, but perhaps good enough for investors.
Granted, it's been a while since I've done stats, so I wouldn't be surprised if this were an inaccurate application, but if it's accurate 100 samples a day doesn't sound that crazy.
> Musk’s lawyers would like to make this case into a complicated factual dispute over spam bots. A complicated factual dispute will take longer;... Taking longer is to Musk’s advantage. ... Musk’s financing commitments expire in April 2023; if he can push the conclusion of this case past then, then he won’t have financing, so he — arguably — can’t be forced to close the deal.
7 comments
[ 2.5 ms ] story [ 19.7 ms ] thread"In a May 6 meeting with Twitter executives, Musk was flabbergasted to learn just how meager Twitter’s process was. Human reviewers randomly sampled 100 accounts per day (less than 0.00005% of daily users) and applied unidentified standards to somehow conclude every quarter for nearly three years that fewer than 5% of Twitter users were false or spam. That’s it. No automation, no AI, no machine learning."
Regardless, seems like Musk will have a hard time reaching a material adverse standard.
A reasonable effort at what?
> Could Twitter estimate its revenue in an SEC filing by sampling 100 paid invoices and putting in CYA statement saying the estimate could be wrong?
It could. It might be a problem between Twitter and the SEC. It might be a problem that due diligence would founder on. It might be a problem that would constitute an materially adverse effect without an explicit contractually agreement that risks disclosed in SEC filings could not do that. These are among the reasons why no sane buyer would sign a merger agreement like Musk signed waiving due diligence and excluding a whole of of specific things from being considered materially adverse events without having done a thorough investigation on their own first.
Granted, it's been a while since I've done stats, so I wouldn't be surprised if this were an inaccurate application, but if it's accurate 100 samples a day doesn't sound that crazy.
[0]: https://www.socscistatistics.com/confidenceinterval/default3...
This one's new for me.