Facebook admits its 10th measurement mistake since September
Facebook has messed up on its math again. More notable than the error itself is the fact that, by my count, this is the tenth measurement mistake that the company has disclosed since September 2016 and the first that resulted in Facebook mischarging advertisers.
Over the course of a year, Facebook incorrectly charged advertisers for link-based video carousel ads that ran on its mobile site when clicked on by smartphone users, the company announced on Tuesday. Facebook is refunding the affected advertisers in full.
Before getting into the nitty-gritty of the latest error, some context. Given Facebook’s series of measurement mistakes in 2016 and the consequential erosion of marketers’ trust, no mess-up is minor, especially one that affects advertisers’ wallets. But this one could have been way worse.
Of the total ad impressions that Facebook served during the year-long period that this error was in effect, 0.04 percent of impressions were affected by the error, according to Facebook. And the median amount of money that individual advertisers are being refunded is $10, according to a person familiar with the matter.
Here’s why the blast radius stayed relatively small despite the prolonged period during which the error was in effect.
1) It only affected video carousel ads, not standalone video ads.
2) It only affected video carousel ads running on Facebook’s mobile site. To underscore how small Facebook’s mobile web audience is, when Facebook decided to fight back against online ad blocking last year, it opted to ignore mobile web ad blocking despite Apple’s iOS enabling developers to build software for iPhone and iPad owners to block ads on mobile sites, like Facebook’s.
2) The error only occurred if someone was using Facebook’s mobile site on a smartphone. That’s likely the majority of Facebook’s mobile web audience but rules out whatever share are accessing the site on tablets.
3) The error only occurred if someone clicked on the ad. Advertisers were billed in error if they had chosen to pay for the video carousel ads based the number of times people clicked on a link attached to the ad to visit the advertiser’s site. In other words, any advertiser who bought a video carousel ad from Facebook based on the number of impressions or video views was not affected, whether the ad was shown to someone using Facebook’s mobile site on a smartphone or not.
So how did this happen? Facebook counted the wrong clicks. When someone clicked on a video in the ad carousel to play it at full screen, Facebook counted that as if it were a click on the link to the advertiser’s site, even though the actual click didn’t send the person to the advertiser’s site.
The error mirrors a measurement mess-up that Facebook had disclosed last year. In November, Facebook said that its analytics tool, Analytics for Apps (recently renamed Facebook Analytics), had incorrectly measured referral traffic sent from Facebook to a developer’s mobile app or website. In that case, when people clicked to expand a photo or video in a Facebook post that linked to a developer’s app or site, Facebook counted those clicks as link clicks, even though the person clicking did not click on the link and never left Facebook.