How Facebook’s News Feed Algorithm Really Works

facebook-f-logo-1920-800x450Late last year, I attended an education session on Facebook’s News Feed algorithm, conducted by a social media lecturer of relatively high standing in the field. The session sounded great – insight into how Facebook’s News Feed algorithm actually works, the ‘hows’ and ‘whys’ of what appears in your News Feed and what brands can learn and implement in order to boost their organic reach. Organic reach, as anyone with any exposure to social knows, has been declining at a rapid rate – brand Pages these days are lucky to reach 10% of their total fans with each of their Facebook posts.

The info session sounded like a great learning opportunity, a great way to get some insight into how to work with the algorithm to maximize Facebook performance.

Except, the information presented was largely wrong.

This person, who speaks and presents to a great many people on social media best practices, outlined strategies that were either out-dated, ill-informed or just plain incorrect, yet stated them as total fact. And as other attendees narrowed their eyes and nodded along, I felt like standing up and saying ‘no, that’s not right’. But then that would assume I was right, and given Facebook’s secrecy around the specifics of their News Feed algorithm and how it works, maybe I actually had it wrong. Maybe what was being presented here was the correct info.

In order to get to the bottom of this and clarify for all those looking to maximize the performance of their Facebook content, I did some research into what’s known about Facebook’s News Feed algorithm and how it selects what content will be shown to each user, every time they log on. And while we can’t know every specific factor that plays a part in how content is distributed on the platform, there are quite a few well established principles that clearly indicate the path to best performance.

Seeking Attention

First off, a bit of history.

When Facebook launched News Feed back in 2006 it was a straight-up, chronological feed of all the activity of your connections.

fbn1

Remember that? The basic looking blue links, the green speech bubble comments.

The ‘Like’ button was introduced a year later, giving Facebook its first insight into what users were actually interested in, and as Facebook became more popular, and more people started using the service – and the News Feed, logically, got more cluttered – Facebook started using those Likes (along with other measures including shares, comments and clicks) as indicative signals to prioritize the content appearing in each users’ News Feed to ensure posts from Pages they’d indicated interest in appeared higher in their stream.

This worked for a while, but there were a couple of problems with this basic approach.

The first issue was that people clicked ‘Like’ for different reasons – funny cat pictures were getting heaps of Likes, and thus, flooding peoples’ News Feeds, while more serious news content, which people weren’t clicking ‘Like’ on (because they didn’t necessarily ‘Like’ it), was being totally buried. Publishing click-bait style headlines became a key tactic as these garnered lots of Likes and clicks, pushing them higher in News Feed ranks – eventually Facebook was at risk of losing their audience because people’s Feeds were being crowded with junk and there was no way, under that system, for Facebook to filter and uncover better, more relevant information for users.

In 2013, Facebook acknowledged it had a problem on this front and sought to correct it with a new algorithm that would uncover ‘high quality content’, the first iteration of the News Feed algorithm.

The second issue confronting The Social Network was that Facebook was getting big. Really big. People were adding more friends and Liking more Pages, meaning there was more and more competition for attention within the News Feed listings. But people only have so much time in the day to check their Facebook updates – according to Facebook, an average Facebook user is likely to have around 1,500 posts eligible to appear in their News Feed on any given day, but if people have more connections and Likes than average, that number could be more like 15,000.

Given this, it’s simply not possible for every user to see every single relevant post, based on their connection graph, every day. Facebook’s challenge with the algorithm was to create a system that uncovers most relevant content each day to provide every user with the best possible experience in order to keep them coming back. But that would also, necessarily, mean that Facebook would have to show some content people had indicated an interest in while excluding others which may also be of interest. The system needed to be incredibly clever to get this balance right.

“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 source, 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, Facebook’s chief product officer (to Time Magazine in July 2015)

These were the two major challenges facing Facebook in developing the News Feed algorithm, and despite the protestations of brands who were forced to sit idly by as their organic reach slowly declined (and who were rightly annoyed at Facebook for promoting Likes as a means of reaching audience, then reducing their relevance), the numbers show that Facebook’s machine learning curation process for the News Feed is actually working. In their most recent earnings report, The Social Network reported that engagement was now up to 46 minutes per day, on average, across Facebook, Instagram, and Messenger, with Monthly Active User numbers also continuing to rise.

FBNF2

The continued rise of Facebook shows that they’re getting the user-experience right – brands don’t like it, many users don’t even know it’s happening, but the News Feed algorithm is working as a means of rationalizing and boosting user activity.

This finding, in itself, highlights just how much Facebook understands about their users and their likely preferences.

Inside the Machine

So how does the News Feed algorithm actually work? While the company’s understandably tight-lipped about the specifics of the News Feed calculations (largely because it’s continually evolving), the basics have been communicated by Facebook several times over the years.

Back in 2013, when Facebook introduced the first version of the News Feed algorithm, The Social Network noted four key points of focus for people creating content on the platform:

  • Make your posts timely and relevant
  • Build credibility and trust with your audience
  • Ask yourself, “Would people share this with their friends or recommend it to others?”
  • Think about, “Would my audience want to see this in their News Feeds?”

Those core principles remain the fundamentals of the News Feed – in a 2014 interview with TechCrunch, Facebook News Feed Director of Product Management Will Cathcart outlined a similar listing for the ‘most powerful determinants of whether a post is shown in the feed’:

  • How popular (Liked, commented on, shared, clicked) are the post creator’s past posts with everyone
  • How popular is this post with everyone who has already seen it
  • How popular have the post creator’s past posts been with the viewer
  • Does the type of post (status update, photo, video, link) match what types have been popular with the viewer in the past
  • How recently was the post published

Cathcart’s advice lead to development of this equation, which is a basic overview of how News Feed prioritizes content:

FBNF3

(Image via TechCrunch)

Of course, as noted, there are many more factors than these at play, but at its most basic, this is the logic behind how Facebook chooses and displays the most relevant content to each user. But that system is always being refined.

Those refinements are borne of necessity – more people using Facebook means more content and more variables to take into account to ensure the best possible user experience for each individual. To get an insight into just how complex that equation is, take a look at the documentation behind Facebook’s ‘Unicorn’ social graph indexing system. While Unicorn was built to power Facebook’s Graph Search engine, the way that system works highlights just how many factors can come into play when trying to uncover the most relevant content for each user – particularly when you consider that a typical Facebook user’s relationship graph looks like this:

FBNF4

In the Unicorn documentation, Facebook refers to the amount of ‘nodes’, signifying people and things, and ‘edges’, representing a relationship between two nodes:

“Although there are many billions of nodes in the social graph, it is quite sparse: a typical node will have less than one thousand edges connecting it to other nodes. The average user has approximately 130 friends. The most popular pages and applications have tens of millions of edges, but these pages represent a tiny fraction of the total number of entities in the graph.”

And in the introduction to the document, Facebook notes that:

“Unicorn is designed to answer billions of queries per day at latencies in the hundreds of milliseconds”

Even without a full grasp of the technical complexities of such inter-connectivity, you can still imagine how complex Facebook’s algorithm needs to be to serve up the most relevant content, and how many potential variations need to be taken into account.

This is why it’s almost impossible to explain the full extent of how the algorithm works, and why Facebook largely avoids doing so. It also enables them to make changes without worrying about what they’ve said previously – if Facebook were to say ‘this is how the system works’ then make a change that altered that, brands that had structured their Facebook strategy around the previous rule would be disadvantaged (which is pretty much what happened with ‘Likes’ when they changed the rules). As such, the core principles noted above remain the driving force, and the key elements marketers should logically be focused on. The further complexities and refinements work to support these fundamentals – adhering to them should keep you in good stead.

Constant Evolution

In line with this, Facebook’s always seeking to refine and update the News Feed algorithm to better serve their users and deliver an evermore relevant on-platform experience. Time Magazine recently reported on how Facebook uses two primary devices to help refine and improve the News Feed algorithm – a team of around 20 engineers and data scientists who assess and evaluate the results of tests and updates to determine the best evolution of the system, and a group of some 700 reviewers, called Facebook’s ‘Feed Quality Panel’, who deliver real, human feedback on their News Feed results, which then help the data team make more informed choices.

“…[members of the Feed Quality Panel] write paragraph-long explanations for why they like or dislike certain posts, which are often reviewed in the News Feed engineers’ weekly meetings. Facebook also regularly conducts one-off online surveys about News Feed satisfaction and brings in average users off the street to demo new features in its usability labs.”

Through this process, combining feedback from real people and improved machine learning, Facebook is continually moving the News Feed algorithm forward and uncovering new best practices. This is why we see so many changes and updates to the algorithm rules, newer factors like ‘time spent reading’ are brought in as Facebook learns from user behavior – content that people click ‘Like’ on before reading, for example, is not given as high a rating as content that’s Liked after reading (after a person has clicked through on a link), because if you’ve taken the time to read something and then Liked it, that’s considered a stronger endorsement of of quality than a knee-jerk response to a headline. Such refinements are logical and thoroughly tested, and Facebook’s gone to efforts to underline that the way the system is weighted is entirely dictated by each individual users’ actions and preferences.

The way Facebook’s algorithm defines ‘high-quality’ in this sense is entirely user driven – if you like cat memes but hate posts from The New York Times, you’ll be shown more of the former and less, if any, of the latter.

“…there’s a line that we can’t cross, which is deciding that a specific piece of information – be it news, political, religious, etc. – is something we should be promoting. It’s just a very, very slippery slope that I think we have to be very careful not go down.” – Adam Mosseri, Project Management Director for News Feed

Due to this, it’s up to each individual brand and business to create content that appeals to their specific audience, and caters to that audience’s needs.

It’s worth noting too, in considering Facebook reach and how to worth with the system to maximize reach and performance, that the actions users take after exposure to your content are far more important than them seeing it in the first place.

This was pointed out by Facebook marketing expert Jon Loomer, who noted that even if your Page reach has declined, that’s not really relevant – what is relevant is whether your website clicks have also declined as a result.

“Let’s assume for a moment that reach actually did drop. If all engagement remained healthy — including website clicks and conversions — what does that drop in reach mean? It would mean that Facebook was showing your content to people most likely to engage favorably — which is what we as marketers and users would want.”

It may just be that, as a consequence of Facebook improving their algorithm, that your Page reach will inevitably drop, because your content’s being shown to a more targeted and focused audience based on their behaviors. And that isn’t necessarily a bad thing.

In all, the main thing to focus on in order to maximize Facebook reach and response is quality content, as defined by audience response. The more utility and value you can provide for your audience, the more likely they’ll want to see more information from you, which they’ll indicate through their Facebook actions – be those direct (Likes, shares, comments, clicks) or indirect (time spent viewing, word-of-mouth via off Page comments). Facebook’s tracking all of it, and in this sense, the core fundamentals of Facebook content remain the same as they did the day the News Feed algorithm was introduced back in 2013:

  • Make your posts timely and relevant
  • Build credibility and trust with your audience
  • Ask yourself, “Would people share this with their friends or recommend it to others?”
  • Think about, “Would my audience want to see this in their News Feeds?”

The News Feed is constantly evolving, but its fundamental principles remain the same. Understanding your audience is key to maximizing your Facebook reach.

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s