Farewell to the network effect

My biggest qualm with the network effect has always been that there is an innate limit to scale. Eventually, we reach a state of diminishing returns; for each marginal unit gained (e.g. onboarded user), there will be a point at which platform utility ultimately starts to decline for everyone, and everything: be it content, product, or people.

And yet: we see centralized social platforms vying for more, more and more; symptomatic of extravagant shareholder expectations (the ostensible scapegoat for all things capitalistic), profit reigns supreme. With each new viral app, incumbent platforms scramble to keep up, in a panicked endeavor to retain users before they churn. Instagram Reels and Youtube Shorts, as desperate attempts to maintain market share, are primarily haphazard regurgitations of short-form content first published on TikTok.

Mainstream social platforms have become bloated, outdated, and overcomplicated. Working to own the entire vertical media stack forces the product to lose its original shape; a square peg stretched thin to squeeze into a round cylinder.

And ironically, it is this very act that generates churn in the first place: platform adoption is not synonymous with network migration; TikTok’s short-form videos are not a replacement to Instagram posts—or to longer-form Youtube content. Rather, they’re a complement; these new content primitives supplement the media on existing social platforms—not only do they provide the experimental container for divergent creativity—and new creators—but these new outputs consequently inspire original content across the entire social media ecosystem.

Humans inherently stan categorization: we want to keep things in separate boxes, with different labels for where each of our items are stored. We visit different social media sites to fulfill different content needs; we ascribe different contexts to the different media we consume. What you seek on Instagram is not the same as what you watch on Youtube, or what you engage with on Twitter. And who you follow varies just as much.

Because it’s not just content we categorize; it’s community, too.

There’s almost no overlap between my web3 community on Twitter and my professional network on Linkedin; Facebook is a stagnant archive of my high school classmates but Instagram is a bustling hive for my closest IRL friends; and while content-first social media platforms (meaning that users engage with content, but not necessarily each other) are less network driven, there are still very few friends I know who also indulge in commentary Youtube channels.

We’re multi-dimensional people; the more opportunities we have to express and explore as many different facets of ourselves — and of each other, the more we get to grow and evolve alongside one another. It’s cheesy, but true: we shed outgrown artifacts of ourselves across the platforms we hop between online.

And in turn, those platforms see us as entirely separate people; if Instagram were to bump into Youtube-you on the street, would that algorithm recognize you?

It’s not just that we share different contexts with different people across different platforms: we also have distinct relationships with the platforms themselves. These platforms see us as separate people, because to them, we are: the context for engagement differs between each of the platforms we exist on. And consequently: we code-switch based on how these contexts shape our personal relationships with each social media platform we call home.

And so, when Instagram tries to poach marketshare from TikTok through Reels, the algorithm often misinterprets what we want when watching short-form videos. Every ad I get on Instagram totally nails me, but every Reel I’m shown is equally as off-base. Even Youtube, which shares the same overarching content primitive as TikTok—i.e. video—can’t seem to get it right with Shorts. With the obscurity of data across centralized social networks, TikTok will always know who I am as a short-form video content consumer better than anyone else.

Nathan’s latest essay in the Every brings up structural platform decline as one of the main drivers of network churn. Twitter is fragmenting, and fractal platforms (powered by onchain distribution channels) are stepping up in response—fragmented social-knowledge networks with very distinct cultures, who draw entirely separate social crowds. The new “downtown” won’t look like Twitter, true, but there will only be a few platform-cities that reach the same relative scale as Twitter.

Power might be shifting towards the edges; market share will not.

Here’s why: because all data is social data, decentralized networks don’t have the same barrier to information as the incumbents do. And because decentralized social platforms are grounded in underlying social graph protocols, scale is about how well you shape the data: how highly demanded your algorithm is in the open algo-marketplace.

Centralized social networks try to bypass asymmetrical information by compensating with more, more, and more data, lowering the barriers to — and context for — content engagement, because they need to mass-produce these proxy insights at scale. That’s why network effects matter so much in a web2 social world: the more data you collect, the better you get at interpreting that data—and the platforms that know their users best will always win in their native content primitive.

And by racing to replace the platforms that service different types of content with your own versions of their primitive — you are literally trying to beat an algorithm at its own game. The relationships you do build feel forced and inauthentic — once again, the very act that leads to churn.

But in web3 social, you (as a platform) not only have your own unique relationship with each user you onboard—you also know the shapes of their relationships beyond the confines of your own platform. Instead of being dependent on network scale to know your users best, you’re now capable of focusing on making the native content created on your social graph as valuable as possible.

While every centralized network hunkers down on scale, onchain social protocols get to fixate on quality: how do you win at creating the stickiest rabbithole algorithms—the types of thought trains that take you down windy content paths, bringing you out of your filter bubbles, and leaving you hanging with more questions than answers?

Meanwhile — how do you keep building for intimacy at the network-layer, owning your own corner of the social internet?

Let’s recap:

Which means, in layman's terms: how good is your algorithm?

Published Jul 14, 2023