Fragmentation & Network Viability
Why market fragmentation is the difference between high margin networks and low margin distribution channels
In my last few posts, I have put across a screening framework for network effect-based startups based on defensibility and scalability, and also explained how SaaS integration can change those assessments. Before evaluating any of that, we have to first determine the viability of developing network effects in the first place. In order to do that, we not only have to evaluate the strength of the value proposition, but also the fragmentation of markets that the network or marketplace aims to connect.
Network effect-based models come in many different variants. They can be 1-sided (connecting users to each other), 2-sided (connected demand to supply), or many-sided (connecting multiple demand-side and supply-side participants). But a common theme among all network types is that every side has to be fragmented for the network or marketplace to be viable, regardless of categorization. As I’ve shown in previous posts, Airbnb (tier-1) and Uber (tier-3) are very different companies when we compare their network effects through the lens of defensibility and scalability. However, both companies still connect a fragmented base of demand (riders or guests) to a fragmented base of supply (drivers or hosts).
On the other hand, networks that are reliant on a handful of participants on any side of their network face the same roadblock, i.e. they become overly reliant on those participants and turn into just another distribution or acquisition channel for them. This is because dominant participants become the most critical part of the value chain, and are free to use multiple channels to reach their audience, buyers, or sellers. Meanwhile, companies built on the premise of connecting these participants to others become unviable without them. As a result, it becomes impossible to create network effects that scale, i.e. adding a new user to the fragmented side of the network does not organically attract or make it more valuable for the dominant side.
Let’s take a look at a few specific examples of this. Unlike my previous posts, I won’t be comparing companies with network effects. Instead, I’ll explain why certain companies or products that have some characteristics of networks or marketplaces could not develop any network effects and the consequences of that.
Moviepass: Doomed by Supply Side Concentration
MoviePass, once branded the “Netflix of cinemas”, may be the best case study of supply-side concentration. Launched in 2011, Moviepass allowed consumers to watch a limited number of movies at participating cinemas in exchange for a monthly subscription. In order to implement this model, Moviepass bought tickets from cinemas to make them available to subscribers. Moviepass then priced its monthly subscriptions lower than the cost of tickets in the hope that driving traffic to cinemas would convince them to share ticket revenue with Moviepass. In other words, Moviepass wanted to transition from a middleman buying inventory to a subscription-based marketplace connecting consumers and cinema theaters. Unfortunately, things didn’t quite work out the way they hoped.
Low pricing certainly helped attract demand as Moviepass hit 3 million subscribers at its peak. However, its planned monetization strategy flopped spectacularly because leading theatre chains like AMC declined to share any of their revenue. The top three theater chains, AMC, Regal Cinemas and Cinemark owned more than 60% of theater screens in the United States and viewed Moviepass as just another customer acquisition channel. This killed any opportunity for Moviepass to establish a network effect between cinemas and consumers. Without another avenue to monetize its service, Moviepass shut down after burning through nearly $70M in venture capital funding.
Supply Side Concentration in the Music Industry
Beyond MoviePass, the music industry is a treasure trove of case studies with a concentrated supply side. In general, the industry is a poor fit for most network-effects based models because (1) a small minority of artists draw the most listeners/fans, and (2) an even smaller group of music labels control the vast majority of licensed music. Let’s take a look at how this has affected Apple and Spotify, two companies that operate at the intersection of technology and music.
The Demise of Music-focused Social Networks
Apple launched iTunes Ping in 2010, a music-oriented social network integrated with iTunes 10. While it allowed users to connect with each other and share music, its core value proposition was allowing fans to follow and connect with the artists themselves. It was shut down just two years later due to a lack of user engagement. Then in 2015, Apple launched Apple Music Connect, a very similar social network integrated with the first release of Apple Music. This was again shut down in 2018 for the same reason.
The struggles of Ping and Apple Music Connect were a direct outcome of supply-side concentration. The music industry follows a power law, with the most popular artists drawing in the vast majority of fans. In fact, a recent study showed that just 1% of artists account for 60% of all concert ticket revenue. The presence of these artists was critical for Ping and Apple Music Connect. But for artists, Ping and Apple Music Connect were merely just another engagement channel. As a result, artists were free to multi-tenant and engaged with fans on multiple networks, including less specialized ones like Twitter and Facebook. This made Ping and Apple Music Connect redundant for both fans and artists. In other words, Ping nor Apple Music never had any potential to create network effects and engagement never took off.
Elusive Margins in Music Distribution
What about iTunes, Apple Music, and Spotify? Each of them rode secular shifts in music consumption (offline to digital downloads, and then downloads to streaming) to become market leaders. Aren’t they examples of successful marketplaces in the music industry?
Three music labels, Warner Music Group, Universal Music Group, and Sony Music, together own the rights to music catalogs that account for 65% of the industry’s revenue. So iTunes, Apple Music, and Spotify could only acquire supply by signing licensing deals with these music labels. It was critical for them to have all the labels on board because the absence of even a single label would result in a significant hole in their library and handicap their products. As a result, iTunes, Apple Music, and Spotify have no network effects at all. They are not marketplaces. They are merely digital distribution channels for music labels.
There are two telling signs of a distribution channel. The first, which I have already explained is supply concentration. The second is a lack of profit potential, as suppliers selling through the distribution channel accumulate value for themselves. In this case, it is the music labels that keep the vast majority of revenue. Studying gross margins makes this abundantly clear. Spotify’s gross margins are roughly 25% while Apple Music’s gross margins have been estimated at 15%. As a frame of reference, Uber, a tier-3 marketplace that has been criticized for its weak profit potential, has gross margins of roughly 50%. Meanwhile, iTunes was previously reported to operate at or near break-even and could not generate substantial profit until the emergence of the app store (which had a fragmented supply side, with gross margins estimated at 90%).
Unlike the other cases we have seen, iTunes, Apple Music, and Spotify were still viable because of other components of their business models. Apple was able to leverage iTunes to fuel the growth of its wildly profitable hardware business (specifically, the iPod). Similarly, Apple Music is purely an added switching cost to keep users locked into Apple’s lucrative device ecosystem. Spotify, on the other hand, developed same-side network effects between listeners by allowing them to create curated playlists which made it more valuable for other listeners. However, since the identity of curators was unimportant, the value of this same-side network effect declined over time and did not result in long-term defensibility (allowing Apple Music to scale). Also, Spotify’s (limited) network effects were restricted to the demand side, leaving it with no avenues to profit from its status as a distribution channel. Spotify has attempted multiple pivots to counter this structural disadvantage. First, it attempted to reach artists directly but then shut down the program because of pressure from labels. Spotify then turned its attention towards podcasting and made multiple acquisitions in the space in an attempt to diversify its business away from being a distribution channel for music labels.
Implications for B2B Networks and Marketplaces
As we have seen so far, supply concentration can prevent network effects from developing, strain economics, and sometimes doom products to failure. This is a key consideration for B2C marketplaces or networks as their demand side is always fragmented (consumers). However, the same dynamics can manifest themselves on the demand side for other types of marketplaces or networks. This will become more obvious as network effect-based models expand deeper into the B2B space, particularly those targeting customers in highly concentrated industries.
Computational health startups are one such example. This is increasingly viewed as a “hot” space by VCs because the cost of sequencing the human genome and microbiome has fallen precipitously over the past decade. This has led to a wave of startups built on the premise that genomic and microbiome data can help pharmaceutical companies develop targeted and more effective drugs (precision, not personalized, medicine). This is the primary reason behind the eye-watering valuations of many genomics companies, including 23andMe, Helix and Ancestry. Consumer sales of DNA testing kits are already slowing and are not nearly sufficient to justify their multi-billion valuations. Instead, the long-term bull case is that their data can improve the efficacy of drug discovery research and clinical trials.
In theory, their data bank becomes larger and more diverse as consumers take more DNA tests, which should make them more valuable for pharmaceutical companies. However, the top 15 pharmaceutical companies own nearly 70% of R&D spend on drug development (and prescription drug revenue). While there are other network participants involved, including contract research organizations (CROs), university research centers, etc., the vast majority of revenue potential (demand) is tied to pharmaceutical companies. So these startups largely target the same, small set of customers.
This is likely to result in the same set of challenges that I have highlighted throughout this post. Pharmaceutical companies will be free to multi-tenant and source data from many of the growing number of genomics and microbiome startups (who gather data from consumer test kits or participants in clinical trials conducted by CROs). So unless genomics and microbiome startups expand to a broader (and perhaps less valuable) customer base, they are at risk of becoming simple data acquisition channels for big pharma. As we have seen, that is not a recipe to create a defensible or profitable business.
Market fragmentation is best viewed as a necessary, but not sufficient condition for building strong network effects, i.e. the absence of market fragmentation prevents the formation of network effects, but its presence does not guarantee it. For both investors and entrepreneurs, fragmentation should be viewed as a first-level filter, to weed out product concepts that may not have the potential to create strong structural advantages, before evaluating other facets.
What you should do next…
1. Pitch me: No warm intros required
I invest in pre-seed/seed-stage startups as a Venture Partner at Speedinvest and also via the Atomico Angel Program. If your startup is built on network effects and based in Europe, pitch me here.
2. Apply for my Reforge course
Want to learn more about network effects? I run a 3-week Reforge course on identifying, bootstrapping, measuring and enhancing network effects. Sign up here!
3. Take the network effects assessment
Benchmark your network effects by answering a 1 min quiz about your startup or a potential startup investment. Try it here.
4. Book a consultation
Do you have a specific question about creating, scaling, or monetizing network effects? Book a consultation or sign up for my free weekly office hours (they tend to get booked out fast).
