Time to trot out http://www.dtc.umn.edu/~odlyzko/doc/metcalfe.pdf again. (Yes, I am one of the co-authors.)
Multiple lines of evidence point to the basic value of a social network scaling like `O(n log(n))`. So getting your first million users is hard. But after you've got there, you don't have to be that much better than the competition to win. The conclusion is that social networks should be a lot less sticky and durable than it looks like.
We concluded that back when MySpace was the dominant social network. MySpace was overtaken by Friendster during the usual delays between a paper being submitted and published. Not long after, Facebook became king of the hill.
And then..Facebook remained there. In large part because they seem to have recognized their vulnerability, and so made a practice of buying new social networks like WhatsApp and Instagram before they grew up. However I concluded years ago that it was only a question of time until someone dethroned them.
Maybe it will be TikTok. Maybe TikTok gets shut down for political reasons. But the existential threat to Meta's dominance is NOT going away. Simply being king of the social media hill is not a sustainable moat.
Has there been research done on negative value from scaling, the uncool participants effect? Where teenagers/young adults leave (or reduce their activity/start looking for other networks) when mom and grandpa join?
One negative effect of scaling I’ve noticed is that mature social networks (eg Instagram) seem harder to break into. Only established influencer content is surfaced so a new creator has trouble.
TikTok seems much better at promoting newer creators, possibly because it doesn’t have as many large existing creators. This means all the new talent goes there, and then the established talent eventually follows.
TikTok nailed that. It’s well established now but back when it was still growing and sort of small, creators who were on both Instagram and TikTok praised TikTok for how easy it was to break in (with good content and not merely follower count).
I'd love to see actual research on "the uncool participants effect" but anecdotally it feels like a real thing that's impossible to avoid. Any social network will start to favor a specific culture - enforced either by the algorithm or by its participants.
Once that culture starts to go out of fashion (which is inevitable) either the algorithm will have to shift to meet the culture (because the existing participants are probably not going to be fast enough to change) or new users will find another platform to start using. That's why even though people are saying Instagram will die if they start to prioritize Reels, I feel like they're going to die if they don't and this is a last ditch effort.
Btw I have wondered for a long time why search on Facebook is such an atrocious mess. But then it struck me: it must be intentional, so you would spend more time scrolling feeds to find that one post you've missed.
For this alone they deserve to be dethroned and forgotten.
Remember the ”dumb fscks” quote? Facebook was founded to be hostile towards its users.
It was a while ago that this happened, so there's a possibility that some people on here are too young to remember this. It was a leaked exchange between Mark Zuckerberg and a friend during the very early days of Facebook - when it was only available to Harvard students:
Mark Zuckerberg : Yeah so if you ever need info about anyone at Harvard
MZ: Just ask.
MZ: I have over 4,000 emails, pictures, addresses, SNS
[Redacted Friend's Name]: What? How'd you manage that one?
MZ: People just submitted it.
MZ: I don't know why.
MZ: They "trust me"
MZ: Dumb fucks.
I actually still remember when my university was added to Facebook - I was studying/procrastinating for my final exams.
> MySpace was overtaken by Friendster
Did that ever happen? Was it a US thing? I've heard of Friendster, but I thought it was something that MySpace killed and then MySpace went global and then got overtaken by Facebook.
My peer group was on Friendster first then MySpace
In some countries, Friendster was extremely popular before Facebook! In others, there were other dominant social networks, like Orkut. Facebook was the first to dominate in all countries.
> In others, there were other dominant social networks
I remember Hi5 in Latin America and India, Hyves in the Netherlands, Orkut in Brazil, VK in Russia, StudyVZ in Germany, Tuenti in Spain. Finland and Sweden had a popular photo log websitr but I can't remember its name.
>> MySpace was overtaken by Friendster
> Did that ever happen?
No, it didn't. Friendster was older than MySpace.
Age isn't relevant. Friendster might've come out before MySpace but not become a dominant network until after MySpace had become dominant.
Not saying that's what happened, I've barely heard the name Friendster before.
I got it backwards. :-(
You are missing a huge point. MySpace and Friendster didn't got out of fashion suddenly because social is not "sticky". Friendster was heavily burdened with scaling issues, design complexity and bugs that basically force moved users to FB which remained fast, slick and extremely usable. Similarly, MySpace just couldn't engineer features faster that users wanted. The big point here is that social is very hard engineering problem and FB excelled it in early days. They are currently failing, in my opinion, because their newsfeed algorithm shows too much content from groups and much less from friends. The group content is something people will like initially but becomes monotonic later and so they quit.
So, the social platforms have failed historically because of engineering issues that their competitior excelled at.
Aren't you just rephrasing the GP's point? If social media was "sticky", users would suck up the poor UX and missing features and stick with platform X because 'that's where everyone is'. The fact that users are willing to 'start over' their accounts over such issues just shows that the moat offered by being the place where everyone is is more shallow than one would think.
> MySpace was overtaken by Friendster
It's the other way around - Friendster was the incument that Myspace unseated.
Awesome, I read this years ago and have always kept it in mind ever since. Just wanted to say thanks for writing/co-authoring it!
It's an interesting thesis, but I'm a bit confused by your methodology - how can you propose a law without looking at any data yourself? You just seem to collect a few unrelated observations ("librarians have observed X", without giving any reference there) and laws that were developed for different contexts (Zipf's law), and from there you argue your way quite informally (mathematically speaking) to your proposed n log(n) formula.
Sorry for being so critical, but I would find it quite interesting if you had hard evidence to support a scaling law like n log(n) for the value of a social network.
We were applying a variety of other scaling phenomena which have been long observed, and then applied it in terms of networks. Most of the actual analysis of data, such as mail delivery was done by Andrew Odlyzko. There is also another form of the paper out there with Bob Briscoe added, and he had quantitative data not previously published from British Telecom. (See https://spectrum.ieee.org/metcalfes-law-is-wrong for a link, but you'll need a subscription.)
That said, the argument and result is intentionally informal, and intended to be a better rule of thumb than Metcalfe's law. Which was itself even more informally reasoned. However its key flaw we explained with:
Metcalfe’s Law is intuitively appealing, since our personal estimate of the size of a network is based on the uptake of that network among friends and family. Our derived value also varies directly with that metric. We therefore see a linear relationship between the perceived size and value of that network.
That said, different social networks may well have different scaling laws. In fact I pointed out privately that eBay likely was a counter-example which is closer to Metcalfe's law for the simple reason that you derive value there from connections to random strangers. And indeed eBay successfully maintained a market leading position for a long time despite charging significantly higher fees than competitors like Amazon Marketplace.
TikTok is not the same offering as Facebook though right? It competes in the Instagram and YouTube space. One thing about FB though is that it was the first social media platform that my mom and dad joined. FB was big enough for the boomers to join it. And I figure that is the moat, the boomers who for the first and only time in their lives joined a social platform and they ain't changing it up now.
A moat based on an aging population isn't much of a moat
No moat can change the fact that kids wouldn't want to be caught dead on the social network that their parents or grandparents use.
Maybe they will leave it on like a tv news channel in the background.
If those they came to follow are gone they will still scroll a feed increasingly filled with ads. When they complain, their friends/family will install something and say touch the dancing penguin/barfing unicorn/whatever to see the latest, it's just like Facebook used to be.
Ostensibly their core use cases don’t overlap 100%, but as long as you accept that textual medium is just a means to information and is replaceable with video, you can actually get everything you get from Facebook on TikTok — news, commentary, entertainment, friend and family updates, messaging, etc. Plus, on a deeper level, everything competes with everything for your time and attention.
> But after you've got there, you don't have to be that much better than the competition to win.
I don't follow, did you mean it's O(log(n))? Otherwise a social network with 2mil users is still more than twice as good as a social network with 1mil users.
It's often said the value of social networking is O(n^2), which is basically impossible to overcome, versus n lg n, which makes the difference between 2 million and 20 million users a lot smaller (too tired for the math, sorry, maybe 20x?) instead of 100x.
If the total value of the network is O(n log(n)) then the value per user is only O(log(n)). So comparing a 2 million user network with a 1 million user network, if the second one is on a product that is 10% better, users should find switching an improvement.
Thanks for explaining! I'd always heard that O(n^2) value for social networks but never really thought about who that value is for. Now that you mention it, the value of a network to an individual user really is only O(n) or even O(log(n)). So while the company still has a massive cash advantage, it's much harder to deepen that moat and secure your customers with it than it seemed.
So, did Zuckerberg read your paper and take it to heart, or did he have some _intuition_ or paranoia that helped guide him?
No clue.