Twitter deal off until fake account numbers are clarified

Elon Musk said his $44 billion acquisition of Twitter will not go ahead until there is more clarity on how many accounts are fake.

Twitter estimated in a filing earlier this month that less than 5% of its daily earning active users — known as mDAUs — during the first quarter were fake or spam accounts.

But Musk estimates that about 20% of Twitter accounts are fake or spam, and he’s worried the number could be even higher.

“My offer was based on Twitter’s SEC filings being accurate.” Musk tweeted early Tuesday morning. “Yesterday, Twitter’s CEO publicly refused to show evidence of <5%. This deal can't move forward until he does."

Musk claims many of the spam accounts are “bots,” but Twitter doesn’t mention the word once in its SEC filing. Fake or spam accounts, often called bot accounts, are automated and not run by humans. But it’s worth noting that Twitter’s automated bots, which are programmed to tweet things at set times, can be good or bad, and Twitter isn’t banning them all from the platform.

Just after Musk voiced his concerns, Twitter announced later Tuesday morning that it is committed to the deal at the pre-agreed price, adding that it has filed a proxy filing with the SEC.

Shares of the company fell 2.46% in premarket trading on Tuesday. A Twitter spokesperson did not immediately respond to a CNBC request for comment.

Musk’s tweet comes just hours after Parag Agrawal, CEO of Twitter, tweeted one long thread about spam on social networks.

Agrawal, a software engineer, said Twitter’s spam ratings are based on multiple human reviews of thousands of accounts that are repeatedly randomly sampled over time.

He said it’s not possible for outside groups to count the exact number of spam accounts on the platform because it requires public and private information that Twitter cannot share.

“From the outside, it’s not even possible to know which accounts count as mDAUs on any given day,” he said.

Analysts at Jefferies said Tuesday that Musk appears to be trying to lower the price due to the recent market selloff.

“Elon Musk’s recent comments suggest he is trying to negotiate a lower offer price,” equity analyst Brent Thill and equity associate James Heaney said in a research note.

“We believe Musk is using his investigation into the percentage of fake TWTR accounts as an excuse to pay below $54.20/share. In reality, NASDAQ COMP is down 25% YTD [year-to-date] and Elon Musk realizes that he can overpay for the asset.”

Musk has said his team is conducting its own analysis on the number of fake accounts on the platform, but experts in social media, disinformation and statistical analysis say his suggested approach for further analysis is woefully lacking.

“To find out, my team will take a random sample of 100 @twitter followers.” Musk tweeted Friday. “I invite others to repeat the same process and see what they discover.”

He clarified his methodology in subsequent tweets, adding, “Pick every account with a lot of followers” and “Ignore 1000 followers first, then pick every 10th. I’m open to better ideas.”

Musk also said, without providing evidence, that he chose 100 as the sample size number for his study because that’s the number Twitter uses to calculate the numbers in its earnings reports.

“Any reasonable random sampling process is fine. If many people independently get similar results for % of fake/spam/duplicate accounts, that will tell. I chose 100 as the sample size number because this is what Twitter uses to calculate <5% fake/spam/duplicate."

Carl T. Bergstrom, a University of Washington professor who co-wrote a book to help people understand data and avoid being taken in by false claims online, told CNBC that sampling 100 followers of each account on Twitter alone should not serve as “due diligence.” to make a $44 billion acquisition.

He said a sample size of 100 is much smaller than the norm for social media researchers studying similar issues and could result in selection bias.

Facebook co-founder Dustin Moskovitz weighed in on the matter via his Twitter account, pointing out that Musk’s approach isn’t really random, uses a very small sample and leaves room for massive error.

– Additional reporting by CNBC’s Lora Kolodny.

Correction: This story has been updated to accurately describe how Twitter categorized fake accounts in its SEC filing.