May The Feed Be With You: how algorithmic feeds create the best content experience for users
Last week, Meta announced a new “Feed” tab in the Facebook app, which will display chronologically ordered posts from only Friends, Groups, and Pages that a user follows. In other words, a feed of content each user assembles manually, one person or organization at a time, that they have complete control over with no intervention by Facebook’s recommendation algorithms. In sharing the reason behind this feature, Meta CEO Mark Zuckerberg said “one of the most requested features for Facebook is to make sure people don’t miss friends’ posts”. As part of this change, the original, algorithmically curated News Feed would be renamed to “Home”, which Facebook now describes as a “discovery engine for you to find and follow new content and creators through recommendations”.
While the Feed tab launch may have been the headline, the real news is that the Home tab would remain the default place where users are taken. Home comes first, not Feed. This launch isn’t Meta prioritizing your small circle of friends to personalize content for you (Feed), but rather Meta doubling down on the wisdom of its vast data and algorithms to personalize content for you (Home).
Free from the responsibility of promoting friend posts, Meta’s sophisticated algorithms can now curate content from across 3 billion monthly users when making recommendations to turn Home into “the starting point for connection, entertainment and discovery on Facebook”. Because even though users may be requesting a Feed tab to keep up with their friends, Meta knows that their best chance to give users the content experience that they really want is with their algorithms at Home.
Social Media Balance
In September 2015, as Instagram had just crossed over 400 million monthly users and surpassed Twitter’s audience, I remember having lunch with an early employee and congratulating his success. To my surprise, my friend wasn’t in a celebratory mood and instead complained how Instagram’s product was better a few years ago than it was right then. I asked him to elaborate:
“In the early days of Instagram, we were in balance. We had the right people, creating and consuming the right content, at the right time. That’s when the product experience was the best. But as the business has grown, that growth pushed us out of balance. Now you don’t consistently see the right photos anymore. You don’t have those special moments in the app as regularly as you used to.
I’m happy that we’re growing, that so many people are choosing to use Instagram. But there is definitely a downside to that growth that’s knocked our content out of balance. And each day we’re spending more and more time just trying to figure out how to get back in balance.”
Balance here doesn’t mean presenting both sides to an issue or argument. Instead it means creating a content loop where you have the “right people, creating and consuming the right content, at the right time”. Here’s another way to describe this important point of balance in media: think of a TV show that you like. When the show is totally obscure, it’s not that interesting. But as more people start to watch it, you take note because there’s now a following, a community you can be a part of. That community grows to become the perfect size with the perfect relationship with the show and its creator who is getting the right feedback and validation to improve the content. Everything is in balance.
Then as the show continues to be successful and its following gets even larger, the community becomes too big with a flawed relationship with the show. You have the wrong people sharing the wrong feedback that makes the show less interesting and special. The relationship between the creator, the consumer, and the content that ties them together is no longer in balance and the experience for everyone suffers. And then what often happens next is you find another TV show to watch.
For successful social media companies, getting out of balance may be inevitable because success inevitably leads to growth, which applies pressure on balance. In its early years, Facebook was in perfect balance: kids at college, sharing the right content at the right time. But then Facebook’s success led to growth that pushed it out of balance when those kids’ parents joined and started sharing the wrong content at the wrong time. Those kids moved over to Instagram, which itself was in perfect balance until its success brought in the influencers who unbalanced the network by setting too high a production bar for what photos were acceptable to share. That gave rise to the authenticity of Snapchat, which was in perfect balance until its success attracted the brands who gave us the imbalance of Taco Bell lenses and Starbucks Frappuccino Happy Hour. And now we’re onto TikTok and its battle to stay in balance.
But TikTok may be fighting this battle better than anyone with the ideal product weapon to ensure balance in their media content: the universal feed.
The Evolution Of The Feed
When Facebook first launched the News Feed on September 6, 2006, people took note. Over a million people joined protest and boycott groups to pressure Facebook to rollback the feature, before it would eventually go on to become one of the most used products ever invented and the standard for social media UI. In October 2009, Facebook introduced a change to their News Feed that few people noticed even though its impact has been arguably just as profound. That change was the introduction of algorithmic sorting of content.
Before 2009, friends’ activity and actions in the News Feed was chronologically sorted every 30 minutes. You simply saw whatever happened to be shared in the past half hour. But from 2009 onward, Facebook would select and personalize the content you saw with algorithms to have the right people, creating and consuming the right content, at the right time. Algorithms that brought balance.
With a chronological feed, social media platforms are solely depending on users to manually balance content. Not only do users have to share interesting content with their friends, they have to do it at exactly the moment when those friends are looking for that content. There’s no margin for error in finding balance. But that’s not the case with an algorithmic feed. Now those platforms can choose which content people see from friends at any time, not just real time. There’s now plenty of margin for error.
In February of 2016, Twitter released a new algorithmically sorted version of their feed. Just one month later, Instagram would do the same, citing the need to solve the problem that users were missing 70% of the content in their chronological feed. Vice Magazine would even call 2016 the “Year of the Algorithmic Timeline” as everyone had come to the same conclusion that an algorithmic feed was the best way to bring balance to social media content.
But balance isn’t permanent. The battle for a social media platform to remain in balance is never ending as forces like growth or influencers or advertisers constantly push content networks out of balance. Social media platforms would respond with increasingly complex and sophisticated algorithms. Facebook for example uses 10,000 different data signals to rank a single post and bring it into balance. But what happens when there are fewer posts to rank?
The Universal Feed
In the past 6 years, there have been numerous reports of Facebook and Instagram users sharing posts less often, with certain forms of sharing declining more than 15% year over year. Fewer posts means less content for algorithms to pick and choose through, which makes it harder to give users the right content at the right time. When users stop posting, gramming, and tweeting, an imbalance is created in the system that even the mighty algorithmic feed can struggle to solve. That’s why TikTok changed the very fundamentals of the feed.
There have been hundreds of investigative reports, technical analysis, research whitepapers, statistical studies, PhD dissertations, and more about how TikTok’s secretive recommendation algorithm works. TikTok’s recommendation system is without a doubt incredibly complex and impressive, but so too are the recommenders that Facebook, Instagram, and Twitter have built. What’s most interesting (and powerful) isn’t how TikTok recommends content, but what content TikTok is recommending from.
Facebook, Instagram, and Twitter all build their feeds from content shared by friends and followers. The pool of content their algorithms choose from is limited or scoped. TikTok’s feed however is built from content shared by any user, regardless if you know them or not. The pool of content their algorithms choose from is unlimited or universal. This universal algorithmic feed has even more margin for error in finding balance compared to a scoped algorithmic feed. With a universal feed, TikTok can not only choose content from any time (not just real time), but also from any person (not just your friends).
The universal feed isn’t a new concept. Since 2017, Instagram has suggested posts from people you don’t follow, but it’s the outlier, not the mainstream content on Instagram. And with all the warnings and disclaimers surrounding any suggested post, it’s as if Instagram was unsure or even embarrassed by the inclusion of this content for many years. That’s not the case for TikTok, who has fully embraced their universal feed, called For You. They’ve set For You as the default, centerpiece content experience for all users. They proudly refer to For You as “one of the defining features of the TikTok platform”. They’ve entrusted For You as their solution to bring balance to more than one billion users.
And now they’ve inspired Meta to do the same.
This Is The Way
On May 31, 2022, the Supreme Court granted an emergency stay request that stopped controversial Texas state law HB 20 from going into effect. Texas legislators had drafted HB 20 to severely limit how social media sites can moderate their platforms. Had the law been allowed to stand, it would have become illegal for any social media platform with more than 50 million monthly users to “block, ban, remove, deplatform, demonetize, de-boost, restrict, deny equal access or visibility to, or otherwise discriminate against expression”.
For now, we don’t have to deal with the wide sweeping impact HB 20 would have had on the entire social media industry. But we can think about the turmoil hypothetically. In its most literal sense, HB 20 doesn’t just prevent content moderation, it prevents content recommendations. Simply removing spam or sorting by relevance or suggesting something you might be interested in requires a platform to “de-boost” or “deny equal visibility” of some content over others. Boosting and de-boosting are core principles for adjusting algorithmic weights in all recommendation systems. You just can’t recommend and organize content without boosting and de-boosting that content, and in a post HB 20 world, you can’t recommend and organize content without committing a crime.
Over the past 2 decades, social media has evolved from primitive sharing to a higher form of communication. Its usage has evolved from a periodic novelty to a high frequency utility. Its audience has evolved from small communities to every community. And instrumental to allowing more people to share more content more often on social media is the millions of engineering hours and billions of overall dollars invested into algorithmic feeds that help make sense of the deluge of information posted every day.
Boosting, restricting, ranking, and recommending content in algorithmic feeds may be the only way that social media can find balance and deliver the content experience that users want.
