We talked quite a bit about the advantages and disadvantages of platforms over traditional pipelines, but there is one aspect that was glaringly absent. Network effect is a critical element of any platform, not only does it allow platforms to grow rapidly, but also allows it to defend itself against competition. It is such a powerful phenomenon that it has been estimated that 70% of value in tech derives from it. The ability of a platform to leverage this powerful effect allows it to not only disrupt pipelines, but protect itself from any future disruptive forces.
What are network effects?
Network effect is a economic phenomenon, where the per user value of the network is nonlinearly affected by the number of users on the network. The term typically describes positive network effects, where more users increases the value per user, but negative network effects are not uncommon. It can also be described as demand side economies of scale and diseconomies of scale, where aggregating demand nonlinearly affects value for users.
In the more common positive case, the company responsible would typically be able to capture more of the value and convert that into increased revenues and profits. This would be similar in nature to the more commonly known supply side economies of scale, where increasing supply allows fixed costs to be spread over more units of goods, driving costs lower and profits higher.
It is important to note that effects from branding and supply side economies of scale are not network effects. They may exhibit similar behaviours, where more users generates more for customers, but the fundamental cause of the value does not come from the interaction between the users.
Why are network effects important?
Strong positive network effects allow value to users to grow exponentially with each additional user. The value is typically modeled with either Metcalfe’s Law or Reed’s Law. Metcalfe’s Law assumes each user would interact with all users in a network, and each additional interaction generates an equal unit of value for the user, thus value increases at the rate of n^2. Reed’s Law goes even further, and assumes that in larger networks, subsets of smaller networks form and generates more value for each user. Reed’s Law therefore estimates value to increase at the rate of 2^n. These estimates likely overstates the power of network effects, as in no real network does each user interact with every other user, but they are illustrative of the power of interactions.
Network effects, along with brand, supply side economies of scale, and IP, creates a moat that allows platforms to defend itself against competitors. Defensibility, or durability, is particularly important in tech. Based on analysis from Peter Thiel, three quarters of the discounted cash flow value of tech firms come from its later stages (10 years later).
Strong positive network effects also allow platforms to scale up and reach profitability quickly, critical in tech where companies burn massive amounts of venture capital cash in the early stages of development.
The advantages of strong network effects are so large that it often creates a winner takes all dynamic. In economic terms, this would be a natural monopoly. Take Google for example, due to the number of users and websites on its platform, as well as its superior pagerank algorithm, it produces by far the best results for your search. The minimal negative network effects resulting from its architecture ensures consistent value for its users.
Types of network effects
There are many ways to categorize network effects, the most broad categories are the aforementioned positive and negative effects. In the typical case, positive effects dominate in the early stages of a network, where every marginal user generates significantly more value for the company. As the company scales up, negative effects start to take hold, and the growth of value per user slows. Take Yelp for example, at the beginning, every new user on the platform will allow for more accurate ratings and reviews of restaurants. As the number of users hit a certain point, additional ratings and reviews become less useful and even counter productive, making it more difficult for you to figure out where to eat tonight.
We can also categorize them into same sided and cross sided effects in a two sided supplier consumer network. As the names imply, same sided effects are network effects where the size of the consumer (or supplier) network affects the value of the platform to itself. Similar logic works for cross sided effects. This is helpful as the effect of more consumers on the value for consumers are typically very different than the effect of more consumers on the value for suppliers. Take Yelp again for example, the effect of more consumers on value for consumers is much greater than those for suppliers.
There are many other ways to categorize network effects, local effect is one that is commonly cited. These other categories are all subsets of the types of effects described above. The basic idea is, depending on the platform, certain constraints, such as geographical or psychological, changes the dynamics of the interactions between the users. It would be necessary to consider all constraints, so that you can more accurately predict value to the users.
Examples of network effects
The most classic and intuitive example of positive network effect is with telephones. The value of one telephone is zero to its user. As the number of telephones increase, the value per user scales up non-linearly. If both of your best friends have telephones, you would spend more than double on that telephone than you would if only one of your best friends have a telephone.
A more modern example would be Facebook, where the value of the platform derives from the fact that your friends are also on the platform. However, despite the prevalence of network effects in technology platforms, they certainly existed before modern technology. One example that predates the age of information are certain foods. The value of certain foods, let’s say avocadoes, increases as more consumers start eating the food. Consumers would make them for dinner parties and create new recipes, which would make the food more valuable to consumers.
Measuring network effects
Measuring network effects is not easy, particularly to outsiders with little data about the platform. Network effects do not appear on the balance sheet or income statements, as they are even more intangible than brand effects. The stock price plotted against the number of users might be helpful in giving you a rough estimate, but this would be conflated by intangible internal assets such as brand, and external shocks such as regulations or epidemics.
If you are privy to detailed information about the platform, you might be able to tease it out of the data. The smart folks from Andreessen Horowitz does a nice job at outlining 16 ways here, so I won’t repeat them. The general ideal would be to plot some sort of proxy for how valuables users find the platform against the number of users. You would also need to account for influence from marketing, product improvement, or anything else that the users might find valuable. Any deviation from constant returns to scale could be attributed to network effects.
Sources: Platform Revolution, Andreessen Horowitz, internal analysis