I might have been the first person to write academically about competition in networks back in the late 1990s, with networks being actually defined as such (using graph theory) rather than group size effects as Katz & Shapiro did in their “Network Externalities” papers.
One of the realizations I had to go through — other than pointing out that neighbor effects are much more relevant than group size effects, which had been my original contribution — was that there are two kinds of competition in networks. The first one is influence networks: consumers' individual utility functions contain other consumers' signalled adoption choices in some form. This is just a generalization of Katz & Shapiro by invoking network structure. The second one is physical networks: the marketplace is not a single locus but transactions occur distributed over the network. I ended writing on the former academically but working on the latter professionally.
Most businesses built on “network effects” tend to invoke Katz & Shapiro reasoning for business model validation. This is a huge problem because even in influence networks their findings are at best half-right. But most of the current network business models are of the physical network type, and physical networks are simply much harder to monetize — they are mostly pure war of attrition plays (Uber is playing a dozen wars of attrition globally) — and even if you win such a war of attrition and try to monetize your monopoly gains, you’re vulnerable to selective entry. De novos might just focus on serving your most profitable routes. A deadly mix for an enterprise as leveraged and as stretched thin geographically as Uber.
So while I might quibble with Horan’s reasoning I agree with his findings. 40 years of deep industry experience shine through .