I recently finished Andrew Chen’s The Cold Start Problem: How to Start and Scale Network Effects and would like to recommend the book to anyone interested in digital assets or technology in general. It’s been a while since I was enrolled in the CFA program or university, but to my knowledge this is not a topic covered in curriculums about business or investment analysis. That said, I believe that the material will become mandatory reading in short order.
It could be that we’ve needed to see many iterations of network effects in action to arrive at a point where we can begin to study and classify the elements of the phenomenon, but it’s been an analysis over a century in the making. For example, in AT&T’s 1900 Annual Report, the company President, Theodore Vail remarked:
This might be the first canonical reference to network effects, and it is similarly applicable to social networks, chat platforms, or even ride hailing applications. The network effect is a phenomenon where the value of a product or service increases as more people use it. As the subscribership grows, there are more opportunities for interactions, which can lead to increased benefits and positive outcomes for each customer. These enhanced engagements provide additional prospects for developers to innovate and encourage more platform utilization. The result is a flywheel where patrons reap additional utility from the marketplace whose tech stack and brand become inherently more valuable.
Organizations that leverage network effects can demonstrate rapid and persistent growth. This can lead to significant market share capture, deterring new entrants and stymieing competition. It is often believed that the largest companies will “win” in their markets by virtue of having the broadest network. Looking at the proportionate market capitalization of Meta Platforms (Formerly Facebook), Amazon, Netflix, Microsoft, Apple, and Google (each of which demonstrate network effects) to the aggregate U.S. equity market cap you might be tempted to think that size is a key determinant:
Given the definition of network effects outlined above, it’s counterintuitive to think that platform size doesn’t matter (this is a recommended read). Frankly, it does, but what’s more important is the idea of network density. A wide, but disparate network is susceptible to disruption by a smaller entrant who does a better job of creating tight relationships between members of a platform. Beyond the tech giants, many of today’s most popular companies enjoy significant network effects:
Notice that there are many firms in the list above that compete against one another. Yes, Uber has network effects, but so does Lyft and now that both have established ridesharing as a prominent way to get around, a new entrant would theoretically be able to reap those benefits, too. Having said that, it’s interesting to see the change in market share between the two companies since the pandemic:
Uber has launched several ancillary services like food delivery and pet-friendly transport. The former was a boon during COVID lockdowns and has since maintained a regular audience for those who value the convenience of getting their meals dropped off. The result is a greater density of the Uber network since it provides an additional source of revenue for drivers while giving users another reason to interact with the app. An important element of network effects which will come up over the course of the coming analyses is the idea of the “hard side” of a network. This relates to the group that creates disproportionate value and by extension has more power. Network operators covet these users. In the case of ridesharing, the drivers are the hard side, so an additional source of income and a larger pool of clients are helpful to keeping them engaged.
In time, it might be the case that there’s a winner-take-all effect in verticals like ridesharing or food delivery, but for now the networks are sufficiently dispersed that platforms are engaged in healthy competition and this benefits both sides of the market. It’s often presumed that network effects alone can provide a competitive moat, but we can see how that’s not true. To stave off new entrants, a platform should constantly seek greater density, by studying its users and finding ways to drive more engagement. The economic value will follow.
Many have put forward ideas for quantifying the value of a network. For a given number of nodes in a network (n), Reed’s law will produce the highest values and is said to be suitable for platforms where groups and individuals come to interact and the user community can extend the original functionality to unlock new capabilities. Reed based his work on the idea that subgroups of network nodes can scale faster than single nodes alone. This seems important when considering companies like Amazon & Shopify whose various verticals have evolved as new business lines sprouted out of creativity and scale. Academics have published reports showing that Metcalfe’s law works best for social networks – you can check out one of the papers here. Finally, Odlyzko’s law (sometimes called Zipf’s law) is the most conservative as it tries to adjust for the idea that just because each node in a network can connect with another, there are reasons why they won’t (language, preference, etc.).
Andrew Chen breaks network effects into three distinct forces. The Acquisition Effect allows products to tap into their network, driving user growth more efficiently as the network grows. To measure this, we look at the percentage of organically acquired users and hope to observe a decreasing customer acquisition cost over time. The Engagement Effect involves increased levels of participation as more people join. To gauge success here, analysts will consider usage metrics and retention rates of various cohorts. Finally, the Economic Effect refers to higher monetization rates as the network expands. We will return to these metrics through the application of this framework to the analysis of different blockchains.
Another important factor to consider when assessing “groups or systems of interconnected people or things” is the potential for negative network effects. Following Metcalfe’s Law above, if you start with 100 users and double to 200, then your value has increased from 10K → 40K. Unfortunately, that math works the same in reverse, so in the example above losing 50% of your users would imply a 75% reduction in value. If a network fails to appeal to users through design flaws, supply/demand imbalances, or congestion, then users will disengage, and this would collapse the value of the platform. Andrew Chen outlines this idea in his post – The Social Media Death Spiral.
This might be the first Digital Dive that doesn’t focus on web3 outright, but I hope you can see where we’re going with the discussion. We’ve defined network effects, considered some of their nuance, and looked at how they continue to foster competition despite a general belief to the contrary. Over the coming weeks we’ll take a more granular approach to how network effects are pervasive across the web3 landscape and why this is likely to accelerate digital asset adoption.
