Math is just too damn difficult for Facebook. What to know about two more measurement errors

Facebook is having a difficult time counting.

The social network revealed new measurement errors Friday related to engagement on live videos and any links to direct users outside of Facebook. That's the third time since September Facebook has admitted to miscalculations, drawing ire from publishers and advertisers and reemphasizing a need for third-party verification.

Facebook reported that none of the errors directly affected advertisers' spending on the network, a.k.a. how the company makes the majority of its revenue. Yet, companies look to these metrics to strategize their content.

"What I would tell you is we are not going to be perfect," Carolyn Everson, Facebook's VP of global market solutions, told Mashable last month. "What we are going to strive to be is as accurate as possible and fully committed to being transparent if we find things and certainly as we update our metrics." 

Facebook disclosed the updates in a blog post within "Metrics FYI," a new category that was created in the wake of other miscalculations.

The first error relates to the breakdown of Facebook reactions to live video. In the category "reactions on post" Facebook only included one reaction per person even though a user can generate multiple reactions during a live video. That's fine.

The error arose in that those extra reactions were misallocated to the category "reactions on shares," which should include reactions based on a re-shared video.

Original image has been replaced. Credit: Mashable

To fix the error, Facebook will count extra reactions in "reactions on post." Facebook said that "reactions on post" may increase by 500 percent on average and "reactions from shares of post" could decrease by 25 percent.

Additionally, Facebook had been calculating inconsistent metrics for links off of Facebook. Tim Peterson of Marketing Land reported that he identified the error with outside links when analyzing an investigation by BuzzFeed on fake news on Facebook and brought it to the network's attention.

Peterson found that the number of shares, likes and comments a link receives as shown through Facebook's Graph API does not equate those same numbers in Facebook's mobile app.

Facebook has currently not fixed the problem. It is still investigating whether the Graph API is over-counting engagement or the mobile search results are under-counting them.

Original image has been replaced. Credit: Mashable

Lastly, Facebook is changing how it measures the expected reach of ad campaigns. According to Facebook, the metric will be less reliant on sampling and extrapolating. Advertisers may see a 10 percent change in their audience sizes.

Facebook isn't alone in facing heat against measurement. For years, users of Twitter's live-streaming app Periscope have been requesting for more transparency on how the app defines a view. The company finally provided a definition Thursday, following a report from Inc on software bugs in the app.

"The measurement, validation, and valuation of social data (digital views, impressions, engagements and time spent) will be the biggest debate in the ad media world in 2017," Nick Cicero, CEO of Delmondo, a social video analytics software company that now partners with Facebook.

Facebook's Everson told Mashable that third-party verification is the "most important thing" for the company going forward.

At least, Facebook is being transparent. That commitment came after the backlash the company received for its error in September, which was surfaced by a Wall Street Journal report.

"Facebook would do well to let publishers independently measure their native vid​eos using neutral third-party analytics providers," Andrew Montalenti, chief technology officer of Parse.ly, wrote in an email. "If the media industry increases the amount of video it's producing, and visitors continue to prefer video content, having standard metrics will be crucial for companies to judge success."