Credit: CC0 Public Domain Data-driven Internet companies that boast they can precisely target advertising down to users' laundry detergent and their favorite TV shows are getting a reputation for overstating their reach.
On Thursday Twitter admitted it had miscounted its monthly user numbers for three years, raising tough questions for social media companies about how accurate this data is.
The revelation came as Twitter, Facebook and Google face intensifying scrutiny over hundreds of fake accounts from Russia that purchased political advertising to influence the outcome of the U.S. presidential election.
As they prepare to be hauled before Senate and House intelligence committees next week to testify on how Russia may have advertised on their services to sway public sentiment, these colossal information portals are being cast as the Wild West of the Internet, with too few checks and balances on their growing power over people's time and advertisers' pocketbooks.
"I believe it's an honest mistake but it doesn't look good because it's an overstatement and when they have mistakes it's always overstating and not understating," Dennis Yu, chief technology officer of BlitzMetrics, said of the Twitter measurement error.
The challenge: It's not as simple as you might imagine to figure out how many people use a social media service.
Counting errors can occur as they did with Twitter. Social media companies also include all kinds of users, from businesses to publishers.
Duplicate accounts—when a person has more than one Facebook or Twitter account, say one for themselves, one for their business, one for their child and another for their pet—can also inflate numbers. And then there are the fake accounts and accounts that are run by bots, not humans, that these companies are constantly trying to root out and remove.
Duplicates and bots
That's why third-party measurements often conflict with measurements provided by the companies themselves, says eMarketer analyst Debra Aho Williamson.
For example, eMarketer factored out bots, duplicate accounts and spam accounts to estimate that Facebook had 1.37 billion monthly active users in 2016. Facebook calculated it had 1.86 billion monthly users at the end of 2016.
"Duplicate" accounts (an account that a user maintains in addition to his or her main account) represented 6% of monthly active users and fake accounts represented 1% of its user base in 2016, according to Facebook.
"Digital media measurement has never been as exact a science as you would think," Williamson said. "People used to call the Internet 'the most measurable medium,' but there have been many instances over the years where things like user counts have been under- or over-stated."
Twitter, which has struggled to add users amid withering competition from Google, Facebook, Instagram and Snapchat, on Thursday revised down its monthly active users by 1 million to 2 million for each of the past three quarters. It said it miscalculated monthly users because it included third party applications that should not have been counted.
"It's important to note that since it's a small absolute number on a base of over 300 million (monthly active users), that historical growth rates and absolute numbers really don't change very much," Twitter's chief financial officer Ned Segal said during a conference call with analysts.
Investors shrugged at the disclosure. But this isn't the first time Twitter has had a problem with how it's measuring its audience. In 2016, it briefly miscalculated metrics on video ad campaigns, which it blamed on a technical error.
Twitter's also constantly trying to weed out fake or spambot accounts. As much as 15% of Twitter's user base may be bots, according to research from the University of Southern California and Indiana University.
Automated systems catch more than 3.2 million suspicious accounts a week, Twitter says. Spam accounts that Twitter has identified are not included in the active user numbers reported to shareholders, the company says. And Twitter estimates that fake or spam accounts represent less than 5% of its monthly active users.
Twitter is far from the only company facing criticism over measurement errors.
Facebook vs. the U.S. Census
In September Facebook was accused of overestimating its reach with young people. That's when Pivotal Research analyst Brian Wieser noted that Facebook claimed to advertisers it could reach 25 million more young Americans than the U.S. Census reports exist.
According to Facebook's Ad Manager, a planning tool for marketers to consult before purchasing ad campaigns, advertisers could reach 41 million 18- to 24-year-olds in the U.S. and 60 million 25- to 34-year-olds. Wieser pointed out that the census counted just 31 million 18-to-24-year-olds and 45 million 25-to-34-year-olds last year.
Facebook said the numbers are self-reported by users and include non-residents and it said the estimates are based on "Facebook user behaviors, user demographics, location data from devices and other factors."
The estimated reach advertisers are given does not determine how much they pay for the ads and advertisers mostly focus on how well their campaigns perform.
Still, the discrepancy has shaken the confidence of advertisers who are keen on Facebook's targeting power and extensive reach but are frustrated by a series of audience measurement misstatements including how long and how often people watch videos and read articles on Facebook.
"Advertisers are frustrated, and the underlying drivers may be the source of the concerns that regulators and politicians have about them in the U.S., Europe and elsewhere," Wieser said. "However, we're not seeing it in the numbers yet, which is why investors haven't expressed much concern."
Those misstatements have escalated calls for independent oversight and for the companies to submit to measurement conducted by third-party companies.
This week Twitter agreed to be audited through the Media Rating Council, the media industry's measurement watchdog. The audit will evaluate how accurately the social media company reports ad impressions and video views.
The decision follows Facebook's move earlier this year to allow the information it provides advertisers to be audited by the Media Rating Council. It also says it's offering new tools for advertisers to judge the performance of their campaigns.
"Any company that does business with a digital publisher shouldn't let them grade their own homework," Williamson said. "It's always going to make sense to look to other sources of information about usage, and to ask questions if something doesn't seem right."
Explore further: Facebook security chief warns of dangers to fake-news solutions
