Most Twitter (TWTR +0.00%) users are familiar with "bot" accounts that automatically pump out tweets and retweets throughout the day. At the end of 2014, Twitter admitted that 8.5% of its users were bots, or programs which "contacted its servers for regular updates" without any human interaction. It also warned that calculations of monthly active users (MAUs) "may be affected" by bot activity. Despite those issues, Twitter hasn't updated its bot tally since then, leaving advertisers and investors in the dark regarding how many of its users are actually human.
Image source: Pixabay.
Why bots exist
The key difference between Twitter and Facebook is that the former thrives on strangers following you, while the latter enhances connections with family members, friends, and real-life acquaintances.
Since Twitter is a virtual soapbox which grows taller with more followers, it's an ideal platform for brands, celebrities, or journalists who want to attract attention. To gain credibility, many users buy software-generated users from gray market sites. As of this writing, one such site is selling 18,000 followers for just $15. These sites also sell bulk packages of likes and retweets, which are automatically executed by bot accounts.
Image source: Author's screenshot.
Twitter's dirty secret
These bots are worthless to paying advertisers, but they basically form the foundation of Twitter's soapbox ecosystem. According to Twitter Audit, only 56% of President Obama's 5 million followers are likely human. The rest are either software generated accounts or bots that mindlessly retweet his messages.
However, Twitter won't flush out these bot accounts for several reasons. Last quarter, Twitter's data licensing business, which sends a "firehose" of tweets to certain companies, generated 10% of its revenue. If Twitter starts killing bots, retweets could sharply decline and reduce the total number of tweets it can sell to businesses. Killing off bot accounts would also reduce its closely watched MAU count, which grew just 11% annually last quarter.
Twitter is so desperate for MAU growth that it classifies automated software, like Apple's Shared Links feature on Safari, as "active users." In the past, Shared Links automatically aggregated links from followed Twitter accounts onto a bookmark-like page, and Twitter counted every automatic update as an "active" user. But after the iOS 8 update, Shared Links only updated the links whenever it was manually opened. In response, Twitter blamed the iDevice maker for its loss of 4 million "MAUs" during the fourth quarter of 2014.
Twitter's attempt to handle bots backfires
For companies that use Twitter as a free advertising platform for viral tweets, bots are helpful. But for companies that pay Twitter for Promoted Tweets, Accounts, and Trends, bots deliver useless interactions which make it impossible to gauge how many real people see an ad.
That's why Twitter replaced its "cost-per-engagement" model -- where advertisers paid whenever users clicked, replied to, retweeted, or liked a promoted tweet -- with a selective one which let advertisers pick the engagements they paid for. In addition to the original interactions, advertisers could choose to only pay if a user did something that a bot couldn't do, like downloading a promoted app, signing up for marketing emails, or buying a product from an external website.
Twitter thought that this strategy would encourage smaller businesses to buy cheaper promoted tweets while convincing larger customers to buy additional engagements. Instead, it backfired and let customers buy fewer types of engagements, resulting in fewer clicks and less ad revenue. That's why analysts only expect Twitter's full-year revenue to rise 58% annually this year, down from 241% growth in 2014.
Is Twitter built on quicksand?
If CEO Jack Dorsey wants to save Twitter, he needs to face this bot problem. If we assume that Twitter's 8.5% figure is accurate, the official MAU count should be reduced by 27 million. If that percentage has risen higher since last December, Twitter should update that figure quarterly so average revenue per human user is more accurate. Turning a blind eye to these problems for the sake of short-term data licensing and MAU growth could cause advertisers -- and investors -- to flee for the exits.
