Home Recommendation Systems How does the Facebook News Feed work?

How does the Facebook News Feed work?

by Rovin

Facebook recently hit 2 billion MAU with an average user spending around 27 minutes a day scrolling through their news feed. With millions of apps and services trying to get your attention, how does Facebook top the list for engagement globally across all demographics and ages? As a Curious PM, I take a peek under the hood of one of the most valuable billboards on Earth, News Feed.

The Mission

The objective of News Feed is to show it’s users the stories that matter to them the most. On any given day, each user has an upward of 2000 new stories that can be served to them via News Feed. An average user will scroll through a few hundred stories at most. Displaying the posts in an unranked fashion using all but a chronological order will likely lead to you missing the most important stories which will, in turn, reduce your engagement.

If you could rate everything that happened on Earth today that was published anywhere by any of your friends, any of your family, any news sourceā€¦and then pick the 10 that were the most meaningful to know today, that would be a really cool service for us to build. That is really what we aspire to have News Feed become.

Chris Cox, Chief Product Officer at Facebook

Ranking becomes supremely important since the more you are engaged the more you will scroll through your news feed, and the more you scroll the more ads Facebook can show you.

The Mystical elements of Ranking

Adam Mosseri, VP of Product for News Feed has recently revealed the four basic elements that go into ranking all the content that is available to a user on Facebook.

  1. Inventory – The list of stories that you have not seen from your friends and the publishers that you follow
  2. Signals – Information Facebook has that helps them make an informed decision. Signals could be information like how old the story is and who posted it. It could be smaller things like how fast your internet is and what kind of phone you are on. Facebook employs over 100,000 signals which they keep updating on their blog here. Signals are also used to understand problematic content like spam when users Hide or Report stories.
  3. Predictions – Using Signals, Facebook makes predictions like how likely are you to like or share a story, or how likely are you to hide or report a story.
  4. Relevancy Score – A relevancy score is created for each story based on all the predictions. All stories are then sorted based on this score for all the posts. A really interesting fact about relevancy score is that for any given post the score will be different for each user since signals are different for each user.
Source: TechCrunch

Since recency is a very important factor in the stories that we engage with, all four steps are done for each post, for each user, every time you visit or refresh Facebook.

The Main Signals that form your News Feed

In one of the recent conferences held by Facebook for developers, Adam Mosseri, VP of Product for News Feed said that although there are thousands of factors that influence what you see on your news feed, there are four really important ones.

Who posted it – The more you have interacted with a friends or a publishers post, the more likely you are to see their posts on your feed. Interactions could be a like or a share. Facebook learns over time when your interactions with a friend reduces and they stop showing their posts from old friends.

When it posted – Recency is one the most important factors in showing a post. However, if you have not logged onto Facebook for a really long time then you will be shown the biggest stories first.

Interactions – The more your friends like, share or tag a post, the more likely it is that you are shown that post.

Type of Content – You will be shown types of posts (images, articles, videos, statuses) that you are most likely to engage with. I usually engage with videos so most of my news feed has videos.

Interesting Signals

Like I said before, Facebook uses a ton of signals to determine the relevancy score and there is nowhere they mention every signal explicitly. After much research, I found a few really interesting one.

Cleaning up News Feed spam – Identify stories that ask you to like or share and promote them lesser on your news feed.

Trending stories – Identify topics that are trending and ranking them higher

Time spent on Stories – Identify the time spent on reading a story, including the time spent on reading comments. You may not push the like button or comment on the story, but the amount of time you spent shows interest and that means you will be interested in similar stories.

Actions on Videos – Identify if you turned on sound for a video, watched it in hi-def, make the video full screen. You may not be inclined to not like or comment on certain videos but other interactions show interest in similar videos.

Reactions – Identify your emotion for the content to help publishers understand more than your likes and dislikes.

For more interesting signals, follow this blog by Facebook – https://newsroom.fb.com/news/category/inside-feed/

Related Posts

Leave a Comment

The Curious PM