Yet again, Facebook is being sued over inflated video metrics

Marketing companies suing Facebook over inflated video-viewing stats can seek a court-ordered outside audit of the social networking service’s numbers, a California federal judge ruled last week.

“Plaintiffs have alleged that these various errors were the result of deficient auditing and verification procedures and the inference in the [complaint] is that these underlying procedures have not been corrected,” wrote Judge Jeffrey S. White. “These allegations, which must be taken as true for purposes of the pleading stage, give rise to a sufficient inference that Plaintiffs will be unable to rely on Facebook’s representations regarding its analytics in the future.”

Facebook didn’t immediately respond to an inquiry from Fast Company.

The company acknowledged in September 2016 that the data points in question, which tracked time spent viewing video clips, were inflated by 60% to 80%, MediaPost reports. It’s not the only time Facebook has revealed errors in computing ad-related metrics. The company has disclosed at least a dozen such errors in the past two years, MarketingLand reports, including

• Double-counting people who saw Facebook Page’s posts over a particular period of time.

• Counting people as having seen posts that didn’t actually appear on their screens.

• Overstating when people clicked from a Facebook post to a link to a business’s website or app by calculating when they merely opened a Facebook image or video embedded in the post.

• Incorrectly counting multiple emoji reactions to Facebook videos as reactions to shared posts of the video.

• Charging advertisers for various video views that weren’t actually visible to users.

• Claiming to reach more people in various demographic groups than official census figures say actually exist.

Facebook has vowed to boost transparency by hosting online and in-person events to educate advertisers about how its metrics work, culling reporting of metrics that are confusing or not actively used, and clearly labeling measurements that are estimated or still undergoing development.