Here's What Facebook's Video Inflate-Gate Is All About

By Published on .

Facebook says no marketers were overbilled as a result of a glitch affecting how it reported the performance of videos.
Facebook says no marketers were overbilled as a result of a glitch affecting how it reported the performance of videos. Credit: Courtesy Facebook
Most Popular

Madison Avenue was flooded with questions from clients about Facebook's inflate-gate on Friday, after brands were alerted that the social network reported ads were being watched longer than they technically were.

Facebook had to correct how it reports on the performance of videos, because it discovered an error in the way it had been calculating consumers' average time spent on them.

On Friday, agencies and media executives said they were hearing from clients worried the problem also impacted their wallets, leading them to pay more than they should for video ads.

First things first: No. No one was overbilled as a result of this glitch, according to Facebook.

Second, there are still concerns among advertisers with Facebook's transparency, and even though this wasn't about money, the industry said there could be other ramifications.

"When data is of the utmost importance," said one agency social media exec. "You can't [mess] up for two years and not catch your own problem."

So, what happened?

Facebook offers advertisers and ad agencies a website where they can check on how well their posts perform. It's called Facebook Insights and it shows all sorts of metrics such as the number of views for a video, the number of unique views and the average time spent. It turns out, the average time spent was inflated 60% to 80%.

How could this happen?

Here are the basics: Facebook reports on two key numbers here, views for a post and the average time spent on a particular post. It only registered a video "view" when someone spent more than 3 seconds on it. But it was calculating the average time spent using the full total of video watch time, including even those people who didn't generate a "view" because they watched for less than 3 seconds.

So if 2 million people saw a video for a total of 10 million seconds, that should be a 5-second average view. But Facebook was dividing that total watch time only by the number of people who stayed for longer than 3 seconds. If that was 1 million people, the 10 million seconds would generate a misleading 10-second average view.

The Wall Street Journal on Thursday reported Facebook's revelation to marketers.

How come the problem didn't impact billing?

The number that Facebook uses to bill advertisers is different than what it reports in performance. Advertisers can pay for video views two different ways, per 1,000 impressions or per 10-second views. The 1,000 impressions are calculated each time a video ad ever starts playing, if even for a fraction of a second, in the Facebook News Feed. The 10-second view speaks for itself. Averages don't come into it.

So, no money, no problem, right?

Not so fast. Media planners and agencies rely on these performance numbers when spending clients' money. Many of them were looking at Facebook video like the new TV, and impressed because people were actually watching.

"It was weird, because it didn't line up," said a major agency tech exec, talking on condition of anonymity. "For the exact same content, across three different platforms, we'd put it on YouTube, we'd put it on Twitter, and Facebook was always higher. It was fishy."

"The impact is more than whether there was inappropriate billing or not," another agency exec said. "It's about how the industry thinks about the platform. If it works better than YouTube, and they want us to take money from TV, then we have to believe the numbers they come up with."

Madison Avenue is pretty pissed then?

Facebook video has always drawn some skepticism, especially around duration of video views. A year ago, WPP's Martin Sorrell called the 3-second standard for a "view" ludicrously low. Now, Facebook is by no means alone, but it has been one of the slower ones to open up to some verification, allowing third-parties to come in and double-check what its saying is correct about what it reports to brands.

What is the solution?

Facebook already changed how it reports average duration spent on each video, so it no longer inflates that number. It also has been opening some windows into its "walled garden" to get third-party measurement. Facebook allows Moat, Nielsen, ComScore and Integral Ad Science, for example, to plug in to verify certain metrics. It also provides deeper analytics tools form advertisers to show how effective campaigns are driving sales, numbers that matter beyond video views.