How the January 2018 Facebook algorithm update works
So, let’s get into the nitty-gritty of what Facebook actually prefers now. Everything here is based off a webinar from Facebook, so huge thanks to Matt Navarra for publishing the slides on his Twitter. We’ve put them all the ones related to the algorithm in this Google doc here.
The aim of the webinar was to help explain the algorithm, and also explain Crowdtangle, a Facebook acquisition that helps track the performance of content. If you just want the Crowdtangle slides, you can find them here (thanks to Derek Silverman).
The news feed
Facebook’s news feed and how it’s ordered and presented is based on four things: inventory, signals, predictions, and score.
This is how Facebook refers to all the available content on Facebook, whether it’s posts from your friends, family, groups you’ve joined, or pages you’ve liked.
These are what Facebook uses to help choose which content goes out. It includes a list of criteria. We’ve listed all the ones we know here, with the bolded ones being more heavily-weighted by Facebook.
- Comments and likes on a person’s status of photo
- Engagement with publisher content posted by friends
- Shares on Messenger
- Replies to comments on a video
- Who posted the content
- When was it posted
- What time is it now
- Technology (what type of phone, how strong the internet connection is)
- Content type
- Average time spent on content
- How informative the post is
- Completeness of a profile
It’s also worth noting that Facebook will give more weight to conversations between people than they will to those between a page and a person.
This is where Facebook uses your profile and previous behaviours to decide what to show you. They attempt to work out how likely you are into like or interact with content, keeping stuff they think you won’t engage with out of your timeline.
This refers to a value assigned to a piece of content that will determine its ‘relevancy’ to the user. The higher the score, the more likely it will appear in the feed. Obviously this means content will get different scores for individual people.