Twitter is getting better at finding and purging fake accounts, but the company has suspended so many bots in the past two months that it’s spooking investors.
A disclosure from Twitter last week said the company’s technology was able to identify more than 9.9 million potential spam accounts per week. That echoes a recent Washington Post report, which said Twitter suspended 70 million accounts in May and June.
Taken at face value, Twitter’s efforts means more than 20 percent of the platform’s 336 million strong active users could have been purged from the site. But as Twitter shares dipped as much as 9 percent on Monday, Ned Segal, its chief financial officer, explained the numbers aren’t quite what they seem.
“If we removed 70M accounts from our reported metrics, you would hear directly from us,” Segal said in a tweet. “This article reflects us getting better at improving the health of the service.”
Byers Market Newsletter
Get breaking news and insider analysis on the rapidly changing world of media and technology right to your inbox.
The clarification seemed to induce a sort of sigh of relief on Wall Street, as shares of Twitter began to slowly rebound from the day’s steep losses.
Twitter had been enjoying one its best runs as a public company, with its stock recently touching a three-year high.
Twitter has said its success is due to improvements in the automated systems that identify and eliminate bots, which are automated accounts that can be used to artificially boost news topics and hashtags. Twitter said it is now removing 214 percent more spam accounts than this time last year.
Alex Taub, CEO of the social media analytics company SocialRank, said he had seen evidence of Twitter’s efforts in going after fake accounts. His company has been working on building out an algorithm to tell fake Twitter accounts from real ones. To do this, the company had been buying fake followers in order to get an idea of what these accounts looked and acted like.
But before SocialRank could build its model, Twitter banned the bulk of the accounts.
"We’ve been trying to build a fake follower formula for Twitter for a long time, and we keep running into the same problem,” Taub said. “We keep buying fake followers, and then we go to make the formula and basically two-thirds are gone.”
Twitter has been under pressure to clean up its act. While the platform has long had a problem with harassment, it also served as a key launchpad for bots and trolls to meddle in the 2016 U.S. election and during subsequent elections and breaking news events.
During Twitter’s last earnings call, in April, Segal warned investors that removing spam accounts would affect the number of monthly users that Twitter reports.
“We're always going to do the right thing to make sure that the service is great for those that should be on it and that's going to be removing spam and suspicious accounts whenever we can,” Segal said, “and we continue to improve at that.”
Twitter is also seeing results on the number of spam reports it receives. In March, Twitter saw 25,000 reports per day, a number that dropped to 17,000 in May.