Facebook’s video suggestion algorithm is complex and constantly evolving in order to provide users with engaging and relevant video content. The goal is to keep users watching videos they enjoy and interacting with the platform. There are a few key factors that go into Facebook’s video suggestion system.
Watch History
One of the main signals Facebook uses is a user’s watch history. This includes videos they have watched fully or partially, pages/profiles they watch videos from, types of videos watched (genre, length etc.), and how long they watch. If a user frequently watches cooking videos from a certain chef’s page, Facebook will suggest more videos from that chef. The more data Facebook gathers on a user’s preferences, the better job it can do suggesting new videos aligned with their interests.
Page Suggestions
Facebook aims to recommend videos from new pages that it believes will be relevant to a user. So if a user watches videos about mountain biking, Facebook may suggest videos from new pages about extreme sports or outdoor activities, even if that specific page has not been watched before. This helps users discover new content creators that align with their tastes.
Trending/Viral Content
Facebook also wants to make sure users see entertaining videos that are trending or going viral at the moment. These suggestions help capture people’s attention and increase engagement. Even if a viral video is not closely related to a user’s typical interests, Facebook will still suggest it if it thinks the content is intriguing enough to get the user to watch.
Connections
If a user’s connections, such as friends or pages they follow, have interacted with a video, Facebook sees this as a signal that the user may also be interested in that content. For example, if a user’s friend likes and comments on a video, there is a higher chance of it being suggested. Connections provide insight into what types of content someone may value.
Demographics
A user’s demographics like age, gender, location etc. also impact video recommendations. Facebook conducts studies to determine what types of video content resonate most with certain demographics and factors this data into its algorithm. For example, younger male users may see more suggestions for gaming or sports videos.
Uploaded Videos
The videos that a user uploads to Facebook provide a strong signal of their interests. Facebook closely analyzes the content of uploaded videos to better understand a user’s preferences and suggest new videos it thinks they will appreciate. Uploading parenting vlogs would likely result in baby-related video suggestions for instance.
Responses
How users respond to suggested videos also teaches Facebook’s algorithm over time. If a user consistently ignores or hides suggestions from a certain page or genre, Facebook will pick up on this feedback and avoid recommending that type of content. On the flip side, if a user frequently watches, likes, comments on or shares suggested videos, the algorithm doubles down on those preferences for future suggestions. User responses allow the system to continually refine recommendations.
Video Details
Facebook also looks at the objective characteristics of videos in order to make smart suggestions. Details like keywords, title, description, subtitles, content tags, author details and metadata offer clues into what the video is about. Facebook can scan these signals to surface videos related to a certain topic, even if the page posting them is new to a user.
Third-Party Integrations
In some cases, video content that originates outside of Facebook can influence suggestions. This includes videos a user has watched on YouTube or interactions with videos on third-party apps. If Facebook can gain insights into what types of non-Facebook videos a user enjoys, it may incorporate this info into its recommendation algorithm as well.
Ad Targeting
For users who enable personalized ads, Facebook relies partially on their ad interests/categories to recommend videos. If Facebook identifies cooking as one of a user’s ad interests, it will likely suggest more cooking videos organically too. Serving users content aligned with their ad preferences can increase engagement.
Page Suggestions Based on Likes/Follows
Facebook will recommend videos from pages that you have previously liked or followed. For example, if you like the Facebook page for a sports team or celebrity, Facebook may suggest live videos or new video posts from those pages since it knows you are already interested in their content.
Location
For users who enable location services, Facebook can factor geographic data into video suggestions. This could include recommending local news, events, neighborhoods, and city-specific pages to help users see what’s happening around them. Location indicates what type of regional videos might be relevant.
Friends’ Likes/Follows
Facebook will sometimes suggest videos based on the pages your friends have liked or follows. For example, if several of your friends follow a gaming personality, you may see videos from that personality even if you haven’t watched their content before. Friends’ interests provide potential video discoveries.
Group Recommendations
If you are part of certain Facebook groups related to your interests or profession, Facebook may recommend videos popular within those groups, even if you haven’t watched them. For example, a user in a gardening group could get tips from videos that are trending in that group.
Platform/Device
Facebook optimizes video suggestions based on the platform and device being used. Video recommendations may differ for mobile vs. desktop for instance. Longer, TV-style videos often be suggested on TV apps and connected devices. Platform factors help serve relevant videos.
Social Context
For surfaces like Watch Party that have a social component, Facebook will recommend videos it thinks will spark conversation, engagement, or shared interest among the users involved.
Publisher Monetization Goals
Facebook partners with video publishers and creators who monetize their content via in-stream ads. Facebook may suggest monetized videos more frequently because increased views also increase publisher revenue from ads. This incentivizes publishers to create quality content for the platform.
Timeliness
Facebook is more likely to suggest fresh, timely videos that align with current events or trends vs older videos. For example, around Halloween Facebook may recommend more seasonal videos to increase engagement.
Objectionable Content Filters
Facebook has filters in place screening out videos with vulgar, offensive, misleading or dangerous content. These objectionable videos are suppressed in suggestions to provide a safe viewing experience.
Business/Sales Objectives
Facebook has a business incentive to recommend videos that keep users engaged with the platform for longer periods. This increases ad revenue. Suggesting addictive but low-quality content can accomplish this goal.
Affiliate/Influencer Marketing
Facebook may suggest videos from pages participating in their affiliate/influencer programs because it has a revenue share agreement with influencers driving traffic. However, these recommendations may not always align with a user’s interests.
Cross-Platform Recommendations
Facebook can identify videos trending on other platforms like YouTube or TikTok and make those suggestions on Facebook as well. Cross-platform trends indicate rising popularity.
Emotional Response Prediction
Using advanced AI, Facebook tries to gauge the emotional sentiment and reaction certain videos will produce. It recommends content predicted to generate positive emotional responses.
Dark Posts
Pages can pay Facebook to “suggest” their videos to specific demographics as dark posts. These paid recommendations help pages promote videos and boost views even if the content may not be relevant to users.
Funneling Into Ad Content
Facebook’s algorithm may funnel users towards videos that lead into paid ad content to increase ad performance. This benefits Facebook’s bottom line through added ad revenue.
Political/Social Agendas
Critics argue that Facebook prioritizes suggesting videos from pages aligned with certain political or social agendas in order to influence public opinion.
Conclusion
In summary, Facebook uses a complex array of factors to recommend videos to users including watch history, page suggestions, trending content, connections, demographics, responses, video metadata, third-party signals, likes/follows, location, friends’ interests, timeliness, monetization goals, emotional analysis, funnels into ads, and potential political motives. The system is optimized to keep users engaged for longer periods on the platform. While not entirely transparent, the algorithm aims to enhance personal relevance and satisfaction through customized video suggestions catering to individuals’ preferences and interests. However, critics argue the system can also promote low-quality viral videos or have biases based on Facebook’s business incentives or political leanings.