I’m a firm believer that incentives and systems govern behavior at scale. If you want to understand why companies make the decisions they do, look at how their executives and boards are rewarded. If you’re trying to grok an unfamiliar sport, see how points are scored and wins are tabulated.
Follow the incentives, and you’ll understand the system.
If you’re a marketer who uses Facebook, Twitter, LinkedIn, Instagram, et al. to promote your messages and content, you’ve undoubtedly read and watched countless pieces about how these social networks optimize for engagement. Looking at the incentives, it’s easy to see why:
- More users on the network leads to more advertisers bidding against each other for user attention, which leads to more revenue
- The more (and more active) users a network earns, the higher their stock goes (financial upside for these companies’ leadership teams is almost all in stock price growth)
- To succeed at scale, social networks need hundreds of millions of heavily addicted users spending more time on their apps, inviting more users to it, and creating content (for free) that hooks their friends into coming back
- To create those addictions, the feed they show you needs to be highly engaging, drawing you back to the platform again and again
- THUS: Social algorithms are intentionally designed to reward highly-engaging accounts with greater future visibility
If you’ve grasped the above, and you know how machine learning works (data + model trained on data + desired outcome = algorithm optimized for outcome), what social networks choose to show us becomes pretty obvious.
As an individual user of any network, we’ve been trained (perhaps subconsciously) to understand that earning engagement is how we get people to see our content. If you’re a social media marketer, or someone who uses social posts to drive traffic and conversions to your brand, this knowledge is crucial to your work.
What I’ve found is that while engagement algorithms are pretty well understood, two other aspects of how social networks choose whether to show our content, are not. Those are:
- Social algorithms inherently reward engagement streaks.
- Social algorithms inherently punish attempts to serve multiple audiences.
Let’s break down both of these, and see how our social reach can benefit from taking them into consideration.
What’s an Engagement Streak?
I bet this’ll sound familiar:
- You post a photo on Instagram. It gets a surprising number of comments and likes.
- You think about why that photo worked so well and try to do it again. BOOM. Lots more views, likes, and comments
- You go for a third, similar post, and if you’re lucky, it really takes off, getting you loads of engagement. But the dopamine hit of likes and comments feels less important, so after a while, you barely even check your stats.
- A week later, the streak forgotten, you post something unrelated, which initially shows promise (a few early likes), but ends up with little overall engagement
You’ve just experienced an engagement streak. Instagram’s algorithm saw that your first post did better than your posts usually do, so it bolstered the reach of your next post. When that one also performed well, Instagram gave you even more reach. Your photo(s) were the first posts folks saw at the top of their feeds. You might even have been featured highly in the Discover sections or to friends of friends.
When the fourth post went up (the “something unrelated”), Instagram initially extended its reach as well. But when a low percent of viewers interacted with it, the algorithm pushed you back into lower visibility territory. Now, if you want to reach Instagram users like you did with the 2nd and 3rd posts in your streak, you’ll have to restart another string of high-performing content.
I’ve visualized this process in action:
This phenomenon isn’t new. Instagram, Facebook, Twitter, LinkedIn, and others have been following this model with their algorithms for at least the past 3-4 years. But the impact is even greater these days than 2015/16 because:
- There’s even more competition for engaging content
- The machine learning systems have grown vastly more sophisticated
- The learning models have trillions more input data to work with
- Both software and UX design techniques have become better at addiction generation
The takeaways for a social media marketer are clear:
- You should intentionally optimize your social posts to discover high resonance, high-engagement-earning content
- Once you’ve found a system that works, it’s wise to double down until you see that engagement fading (at which point, more experimentation may be worthwhile)
- When you’ve got a streak going, it’s very unwise to sabotage it with content that might throw the algorithm off from promoting you
- If your goal is to drive traffic to your website, you want to design a feed that builds up algorithmic equity through engagement, then spends that equity on traffic-driving posts with links (since posts with 3rd party links are almost always pushed down in the feed by social networks who want to keep you on their platforms).
These may not be the ways we personally want to use our social feeds (I certainly hate feeling constrained by The Tyranny of Optimizing for Amplification), but they’re undoubtedly effective. If you’re coaching the team running your brand’s social account, you want them to at least understand these mechanisms.
Now, let’s talk about the other overlooked byproduct of these systems.
Why Do Social Algos Bias Against Serving Multiple Audiences?
On LinkedIn, I post almost exclusively about marketing and startups-related stuff. On Instagram, it’s nearly all personal life, travels (and yes, smooches) with Geraldine. On Twitter, however, I’m all over the place. I’ll post about marketing, startups, pasta recipes, politics, a fascinating article I found via Pocket, a friend’s charity fundraiser. It’s all authentic, and true to what I want to share, but it isn’t focused.
Because of how social network algorithms operate, that’s invariably hurting my content’s potential reach.
In some circles I’m sure this concept is well-understood. But in most of the discussions around optimizing for social reach I see, it’s surprisingly missing. The reality is that social networks, as sophisticated as they are, don’t try to optimize *who* to show your different content pieces to. The algorithms as they stand today (mostly) assume that either an account’s content is consistently engaging for someone or it’s not. They’re designed, or at least *optimized* for single-focus accounts.
If you’re a celebrity or a personal brand, this matters much less. People follow you because they’re interested in everything you do, because you’re the one doing it. Those accounts flawlessly match the algorithms’ design.
If you’re trying to build a personal or company brand, though, the focus vs. audience problem becomes much more challenging. Shake up what you’re talking about and you might lower the engagement you receive from your usual followers, which then renders your posts nearly invisible. Even if it might earn you new kinds of followers over time, those people won’t be as likely to engage if/when you go back to posting about your old topic.
That’s why I have so much respect for accounts like Dr. Pete Meyers, who’s willing to send dozens or hundreds of tweets he knows most of his followers (apart from me and Nancy) will ignore. If he were more like Cyrus Shepard (whose content is relentlessly consistent about SEO), he’d be building up high engagement streaks and likely earning far more visibility on his “professional” tweets. But, like me, he doesn’t want to compromise who he is. Pete wants to be his whole self on Twitter, and I love that, even if the algorithm hates it.
Contrast with Nandini Jammi, cofounder of Check My Ads (and prev. Sleeping Giants). She’s laser focused on brand safety, wasted ad spend, and holding companies that enable disinformation, misinformation, and hate speech online’s feet to the fire. That relentless focus and the emotional responses she’s generated around it mean that despite her account having <10% the raw count of followers that mine does, I’m certain her tweets usually reach 10X more impressions and engagements than mine do.
Twitter is far from the only place this happens. The same patterns holds on Instagram, on Facebook, on Tik Tok, and it’s starting to happen on LinkedIn (though that network’s feed algorithm is pretty obviously behind the other players in terms of engagement-optimization).
There is no right answer here. For some, optimizing to these algorithms may be entirely wrong. For others, it’s exactly what you want to do. But I’m confident that this knowledge is better out in the open than locked behind closed doors. We’re all on the social web all the time (doubly so with the pandemic’s lockdowns), and we should understand the systems that govern our behavior and our feeds.