Facebook attempted to respond to recent criticisms regarding the dominance of conservative pages on its platform with a Newsroom post by recently named chief marketing officer (and vice president of analytics) Alex Schultz that appeared to highlight the limitations of analytics tool CrowdTangle—which Facebook acquired in November 2016—as well as running counter to the social network’s emphasis on quality of interactions versus number of interactions.
Schultz also sought to minimize the presence of political content on Facebook, saying that it makes up just 6% of what users in the U.S. see in their News Feed, and adding that the increase in posts on Halloween was double the corresponding rise on Election Day, despite the social network’s prompts encouraging users to post about voting.
Kevin Roose, a technology columnist for The New York Times, drew much attention to the performance of conservative pages with his @FacebooksTop10 Twitter account, which shares the top 10 link posts from Facebook pages every day, ranked by total interactions and culled through CrowdTangle.
Schultz wrote that CrowdTangle was built to provide insights on content that generates likes, comments and shares, but not to determine which content is seen the most, adding, “Ranking top page posts by reactions, comments, etc., doesn’t paint a full picture of what people actually see on Facebook … Likes and comments don’t equate to reach. Our ranking models include much more than just engagement. For example, they can include survey results, like when we ask people whether a post was worth their time.”
He called Roose’s lists “a subset of page posts,” and shared two charts comparing the top performing U.S. pages by engagement on posts with links and engagement with all posts and spotlighting the differences, from Oct. 23 through 29.
Schultz also shared data on reach, writing, “To show the full picture, though, we think it’s important to look at how many people actually see the content (vs. liking or sharing a piece of content).”
He illustrated this point with two more charts, showing the contrast between the top U.S. pages by reach from all posts and the top U.S. publisher domains by reach of their links.
This would appear to run counter to the tendency at Facebook—and at social platforms, in general—to view reach as more of a vanity metric and put more of an emphasis on interactions with content.
Schultz addressed the spikes in engagement for pages such as CNN, The New York Times, NPR and The Washington Post Saturday, Nov. 7, when the election was called in favor of former Vice President Joe Biden, saying that some of those publishers saw massive spikes on that day but then returned to normalcy, and adding, “A smaller effect appears to have been due to the temporary measures we put in place to address potential misinformation and delegitimizing content on our platform related to the election.”
He admitted that the points he presented were “not meant to be a perfect analysis,” pointing to Facebook’s partnership with researchers from several universities through the Facebook Open Research and Transparency project and nothing that the first research papers from FORT are expected next year, with the aim of providing a better understanding.
Schultz concluded, “Following Cambridge Analytica, it is clear how careful we need to be about partnering with researchers and giving them access to data. It is also clear that following the last presidential election, we need to have independent research to understand our role in elections. This partnership through FORT is a step in that direction and (although nothing is perfect) one I am really proud of. I hope it can serve as a base from which we can build for the future, to thread the needle on the needs of privacy and research.”