Facebook News Feed Algorithm Updates
Image by Jim Wilson/The New York Times
Facebook has announced several updates to its News Feed algorithm over the last couple of months. In this post I’ll cover some background context on what the News Feed algorithm is and what the recent updates mean for marketers.
What is the News Feed algorithm?
The News Feed algorithm’s goal is to ensure that Facebook users are seeing the highest quality, most engaging content in their News Feed. The News Feed algorithm was originally launched under the name ‘EdgeRank’. While Facebook hasn’t used the term EdgeRank in some years, you may still hear the term from some marketers. And, understanding the context behind EdgeRank is helpful in understanding the News Feed algorithm today.
In order to create a balanced equation that can surface personalized content to each user, Facebook normalizes each potential input by classifying each input as an ‘Object’. Each Status Update, Image, Video, Link, etc. is considered an Object. When an individual (or brand) first posts an Object on Facebook, an ‘Edge’ is created. Each subsequent interaction with that Object, such as a Like, Comment, Share, or Tag creates an additional Edge on that Object.
The EdgeRank algorithm (or News Feed algorithm) attempted to weigh the Edges in a matter that can surface the best content for each user. In 2010, Facebook revealed the high level formula for EdgeRank seen below:
The elements of the formula are as follows:
u = the Affinity Score. This score weighs in how frequently you interact with the individual that is posting, how many mutual friends you have, whether or not you are related, etc. The more factors you share with the individual, the higher the affinity score. The Affinity Score is also one-way in that for your feed, it only takes into account how frequently you interact with the other individual; it does not account for how frequently the other individual interacts with you. This is because the algorithm is all about ensuring personalized relevance to you, the user.
w = the Weighting. Not all Objects are created equal. Users tend to engage more with visual content such as Images than they do with, say, text only content in a Status Update. Thus, Facebook weighs these Objects based on users’ engagement behavior. Also, not all Edges are created equal. For example, a Comment represents a higher level of engagement than a Like — given the relative effort a user puts in on each action. So, Facebook weighs these actions (Edges) differently.
d = the Time Decay factor. Again, Facebook wants to ensure that users are seeing relevant, engaging content. An article from last week is likely outdated and less relevant than an article that was released today. Thus, EdgeRank factors in when the Object was originally posted.
∑ = the sum of the Affinity Score, Weighting and Time Decay to score a piece of content. The higher the score, the more likely that content will surface in your News Feed.
How the News Feed algorithm has evolved
Much like Google with its search algorithm, over the years, Facebook has continued to evolve and improve on its EdgeRank (News Feed) algorithm.
In mid-2013, Facebook tested a new algorithm with over 1,000 different factors to gauge the quality of content that a Page on Facebook posts. When this algorithm proved to increase engagement in the News Feed, Facebook, incorporated the algorithm into the core News Feed algorithm.
Later in 2013, Facebook added more weight to posts (Objects) that included links to quality news sites, as more users were tuning into Facebook to keep up with current events.
Facebook also took notice that the more text based Status Updates that a user sees from friends, the more likely the user is to post her own text based Status Update. Thus, in early 2014, Facebook added more weight to Status Updates with text.
Also in early 2014, Facebook increased the weighting of a Tag on a Facebook post — increasing the chances that the audience of the individual that was tagged would see the post.
Later in 2014, Facebook took a page out of Google’s playbook and took action to reduce clickbait — a term used to describe sensationalist headlines or eye-catching thumbnail pictures that entice audiences to click on the related link, which drives the audience to low quality content.
In fall of 2014, Facebook began considering when posts were getting engagement and incorporated Trending Topics — surfacing content that was trending in engagement amongst users.
Also in the fall of 2014, Facebook made an effort to reduce the number of promotional posts that Pages were creating, pushing people to buy product, enter a promotional code or tune into a show/event at a certain time.
Today, Facebook’s News Feed algorithm takes into account over 100,000 ‘signals’ (think Edges and other factors described above) to determine what content is most relevant and engaging to each of its users. Facebook moved away from using the term EdgeRank when it began incorporating these signals into a machine learning algorithm. Now it’s referred to simply as the News Feed algorithm.
This video of Adam Mosseri (VP of Product Management for Facebook News Feed) at the F8 Facebook Developers Conference in April is a good primer for how Facebook News Feed works today.
About the most recent updates
On June 29, 2016, Facebook announced that it would increase the weighting of posts by users’ friends and family based on feedback from users that they are still worried that they may miss posts from people that they care about. I guess #FOMO is alive and well with Facebook users. From Facebook’s announcement:
For people with many connections this is particularly important, as there are a lot of stories for them to see each day. So we are updating News Feed over the coming weeks so that the things posted by the friends you care about are higher up in your News Feed.
On August 4, 2016, Facebook announced that it is taking action to further reduce clickbait. From Facebook’s announcement:
[Clickbait] are headlines that intentionally leave out crucial information, or mislead people, forcing people to click to find out the answer. For example: “When She Looked Under Her Couch Cushions And Saw THIS… I Was SHOCKED!”; “He Put Garlic In His Shoes Before Going To Bed And What Happens Next Is Hard To Believe”; or “The Dog Barked At The Deliveryman And His Reaction Was Priceless.”
Facebook’s has categorized tens of thousands of headlines as clickbait by considering whether or not:
- the headline withholds information required to understand what the content of the article is, and
- the headline exaggerates the article to create misleading expectations for the reader.
Facebook’s algorithm then reviews the headline of a post and compares the post’s language with phrases commonly used in clickbait headlines. If the post is determined to be clickbait, then it is less likely to surface in users’ News Feeds. Furthermore, if the post comes from a Facebook Page that has been flagged as frequently posting clickbait, posts from that Page will be less likely to surface in users’ News Feeds until that Page reduces its clickbait posts.
On August 11, 2016, Facebook announced that it is taking action to surface more personally informative stories in users’ News Feeds. From Facebook’s announcement:
To better understand how we can show people the most informative stories to them, we talk to people and ask them how we can improve what they see when they check Facebook. This is our Feed Quality Program, which includes global crowd-sourced surveys of tens of thousands of people per day, as well as people who answer more detailed questions about what they like seeing in their feeds. We ask people through this Feed Quality Program to rate their experience. For stories that people rate highly, we then ask them why they enjoyed seeing those particular stories. One of the most common reasons people give us is that the story made them feel informed about the world around them.
Facebook is using its Feed Quality Program to survey users, asking them how informative they rank a story on a scale of 1 to 5 (1 being “really not informative” and 5 being “really informative”). Then, Facebook is combining this that story’s score (i.e. ‘signal’) with signals that determine how relevant the content would be to you personally, using signals such as the ones in the Affinity Score described above.
What do these changes this mean for marketers?
These changes are really just a continued focus on providing users with higher quality content — quality as defined by Facebook users’ common behavior and each user’s personal behavior.
Increased emphasis on stories posted by Facebook friends vs. publishers just means that publishers (and brands) need to ensure that they are posting content that Facebook users are likely to engage with via Likes, Comments, Shares and Tags. De-ranking clickbait means that stories posted to Facebook must have substance. And, improvements on qualifying informative content provides marketers and publishers direction on one characteristic of content that is highly engaging: being informative.
Indeed, Facebook has previously posted its News Feed Values, where Facebook describes the values on which it makes its algorithm improvements. The values are listed below. You can click here to read the full values statement.
- Friends and family first
- Your feed should inform
- Your feed should entertain
- A platform for all ideas
- Authentic communication
- You control your experience
- Constant iteration
Over the last several years, especially since Facebook’s IPO in 2012, publishers and brands have increasingly had to pay to reach Facebook audiences — even those that have liked the brand’s Facebook Page. Indeed, I’ve seen social media practitioners’ role on Facebook evolve from primarily community managers — drafting and publishing Facebook posts, and moderating and engaging with audience’s comments — to primarily media planners and buyers — managing Facebook ad buys. This trend will continue. Successful brands will combine an emphasize on creating quality content that is highly informative and/or entertaining to their audiences with ongoing optimization of audience targeting via Facebook’s ad platform. I’ll cover this in depth in a future post.
(This post originally appeared on ReciprocityTheory)