Facebook Announces More Miscalculated Metrics Reported to Advertisers
Social media network to form measurement council and provide more data from third-party verifiers
Facebook has announced that it has discovered numerous additional flawed measurements of how users interact with content, raising further questions regarding the metrics that marketers depend on when deciding whether or not to purchase ads on the social media network.
Errors Upon Errors
The Wall Street Journal (WSJ) reports that the company discovered so-called "bugs" by conducting a comprehensive audit of internal metrics. These discrepancies resulted in either the under-counting or the over-counting of four measurements. These include the weekly and monthly reach of posts made by marketers, the number of full video views, and the amount of time spend with publishers' Instant Articles.
According to Mark Rabkin, vice president of the company's core ads team, none of the affected metrics impact Facebook's billing.
The announcement was made two months after Facebook acknowledged that it had been overestimating the average amount of time users viewed video ads on the platform for about two years. That discovery compelled the company to make a more thorough examination of all of its metrics and data related to advertising.
Now that it has found additional instances of miscalculations, Facebook is planning to provide more independent reviews of certain measurements.
The news of the errors may damage faith in the platform for some advertisers and web publishers who depend on it for distribution and monetization. Many of these had already called for more validation of the platform's data by third parties.
Rectifying the Problems
In an attempt to bolster advertisers' faith in the reliability of its metrics, Facebook has committed to provide viewability data from third-party metrics companies for display ad campaigns. This information was previously limited to video campaigns only. Such companies include Moat and Integral Ad Science.
Carolyn Everson is Facebook's vice president of global marketing solutions. "We are doubling down on our efforts at third-party verification," she said.
The company also claimed that it is collaborating with ratings firm Nielson to count its video views, including both on-demand views and live viewing, as part of the firm's Digital Content Ratings metric.
Finally, Facebook is also planning to put together a Measurement Council comprising marketers and ad agency executives, and it will also provide a blog to provide more regular measurement updates.
Most of Facebook's metrics are unaffected by the bugs, as they affect only four out of more than 220 metrics counted by the platform.
Business Insider reports that Everson notified its agency partners about the bugs and the resultant changes in an email. "Some are overstated, some are understated, but most importantly, none are billable metrics," Everson wrote.
"Fairness and transparency"
One example of an impacted metric, says WSJ, is the over-counting of the number of people exposed to organic posts created by marketers. This meant that normal posts were not paid ads because the metric was adding up the daily reach measured over certain periods without taking repeat visitors into account. Facebook said that, on average, the corrected metric will be approximately 33 percent lower for the seven-day period, while it will be 55 percent lower for the 28-day period.
Another example is the under-counting of the number of people who watch video ads until they finish. Although some people do watch such ads all the way through, because the audio and video tracks on the videos are not always synced perfectly, the network was not counting some views as finished at 100 percent. The result is that Facebook now expects the number of finished video views to rise by 35 percent.
And due to another calculation mistake, the amount of time spent by users with Instant Articles had been over-reported by an average of 7 percent to 8 percent since August 2015.
Consumerist reports that the network will also use more descriptive terms to ensure that metrics are clear. For example, "view content" will be changed to "website view of content," and "video views" will change to "3-second video views."
"Some of [the bugs] will have a bigger impact than others," Mr. Rabkin told WSJ. "The range of impact varies. We wanted to err on the side of fairness and transparency."