Each of the last two years, Facebook has posted videos from its F8 developer conference that reveals (some of) News Feed's inner workings. Last year's video is 45 minutes of content, this year's version is under ten minutes, but contains some key elements.
How News Feed works is one question; how enterprises should approach it is another entirely. I'll run through the latest on News Feed first, and then give you some quick hits for enterprises.
1. When publishers post content, nothing happens immediately on the News Feed of their audience. It's not until someone logs into their Facebook account that Facebook's News Feed algorithm kicks into gear, surfacing what Facebook feels is the most relevant content to that user.
2. When you sign up for Facebook, your News Feed is a blank slate. The content served up on your feed is algorithmically determined by the people you friend, the links you click on/share, and to a lesser degree the pages you like.
3. Facebook is rigorous about what they call "user experience," meaning they won't persist in showing you News Feed content you don't find engaging (with the possible exception of sponsored stories/ads). That's a big reason why the exposure of "business" Page content in News Feed was lowered, causing an outburst of frustration for those companies that followed Facebook down the Primrose Path (see Facebook's explanation for the decline in "organic reach" of Page content.)
4. Facebook serves up the most relevant content to you in your News Feed, including the content types you prefer (e.g. photos vs. plain text updates vs. videos, etc). All content in your News Feed has a specific "relevancy score." That score is unique to you, and the items in your News Feed are prioritized based on those scores.
5. Your News Feed content changes over time based on your consumption choices, and whether or not you define preferences. Facebook calls those preferences "controls." That means you impact what the News Feed shows you - not just by what you click on - but whether you choose to follow all postings from a Page (you have to click on "all notifications" from a page or group for that to happen). Another "control": whether you put a friend on a restricted or "close friends" list.
6. Facebook takes into account the number of likes and interactions on a post. Adam Mosseri, VP of Product Management for News Feed, cited the example of two hypothetical posts from BBC Sports in the UK. One has thousands of likes and comments and the other has only a few - Facebook's algo will infer importance based on that engagement.
7. Recency of content is an important News Feed consideration, but not the only one. As Mosseri put it:
Recency is a really important signal for relevance, but it's not the only important signal. I have a cousin. Her name's Margaret and she recently got engaged. If she'd posted that on Facebook last Friday, and I hadn't been to News Feed since then, and this morning my brother posted a picture of - I don't know - a breakfast sandwich, I'm probably more interested in Margaret's engagement story than my brother's breakfast sandwich, even though her story's a bit older. Hopefully, if we're doing our job at News Feed, Margaret's story would show up at the top, maybe followed by my brother's sandwich.
8. Facebook alters its News Feed algorithm based on structured user feedback. As Mosseri explained, you might be profoundly impacted by a post (e.g. the death of a friend's dog), but choose not to comment or engage with that post. Facebook tries to correct overlooked variables via two programs. One is the "Feed Quality Panel," where they ask users to organize stories based on their interest. Facebook also pulls in tens of thousands of online surveys to determine interest in a particular story - across 30 languages worldwide.
9. Other Facebook controls include the ability to "unfollow" someone, or "hide" a story (hiding a story will also hide similar stories in the future). "See first" allows users to designate which content they want to see first, either from friends or publishers. Mosseri actually recommends unfollowing someone to improve your News Feed:
If you've never unfollowed someone, I highly recommend to you trying it. Maybe you have a kooky aunt who posts - I don't know - funny photos that you're not interested in. If you unfollow them, your News Feed experience will get more interesting.
10. Facebook is prioritizing Facebook Live video on News Feeds, but it has also given users the ability to surpress all Live video content.
Mosseri made these recommendations to publishers:
- Write compelling headlines, not click bait. "[Write] headlines that give people a real sense of the content that's behind that click."
- Avoid overly promotional content - build credibility/trust over time instead.
- Experiment: "What's best for your audience is probably not best for a different audience... Try long form. Try short form. Try video."
- Course correct based on the analytical data from publisher tools. Double down on content your audience finds engaging. Try different tones, and then look at your publisher tools to figure out what's resonating with the people that follow your publication.
- Use the "audience optimization" tool in publisher tools to tell Facebook the demographics of the audience you think will be most interested in your post. Alternately, use "audience restrictions" to designate groups that won't be interested in the content.
Mosseri didn't mention "consider paying us for targeted advertising given your organic content is
pretty much screwed swimming upstream," but he probably should have.
Changes are expected with the reveal that trending topics involve human curation. Facebook is also freaking out about the reduced personal sharing numbers.
Facebook will continue to push News Feed envelopes based on its own agendas (e.g. sponsored stories, Live videos). That leads businesses down a path towards paid advertising on Facebook, as they attempt to make up for the loss of organic Page reach. It's not the only way forward.
Enterprise audiences do exist on Facebook. With advertising and sponsored ads, they can be reached in relatively precise ways given Facebook's demographic ad tools. The enterprise content most likely to succeed on Facebook will be more personal, cultural, and edgier. A link to a position paper on cloud security seems doomed to obscurity. A series of videos on working as a woman in tech sounds promising.
Tackle Facebook on multiple fronts:
- Continue to compile Likes on Pages, but NEVER pay for Likes, which results in a diluted target audience.
- Aside from special events, don't expect big traffic on business Pages. Use the Likes on Pages to enhance targeting on ad campaigns.
- Make organic content as compelling and Facebook-specific as possible, but don't expect excellent News Feed reach unless you are a brand with huge consumer affinity (e.g. my piece on the Boston Celtics).
- Encourage employees who like to post work-related content on Facebook, but provide them with guidelines and flexibility to do this in a way that suits their audience.
- Explore Instant Articles as a content distribution channel. Determine a way to evaluated that content's effectiveness, given there are no outbound links (outside of Facebook) allowed within Instant Articles.
- Take advantage of Facebook's current high weighting to Facebook Live video by hosting live informal/interactive events, streaming them on Facebook, and embedding them on your own pages afterwards (unless you are a verified media organization, a person will need to broadcast the live event, NOT your business Page). See Den's pieces on Facebook Live for the enterprise.
- If anyone is on the fence about posting work-related content, suggest lists. Lists are a bit tedious to set up, but they provide an effective way of sharing work content with those friends who are inclined to like that content, sparing high school chums from work content they have no interest in.
Given how many companies are posting enterprisey content on Facebook in some form, it makes sense to weigh options - while hedging against the downsides with clear guidelines and low scale experiments.