As I have learned more about what to look for in technology-oriented investment opportunities, I have begun to realize that investing in businesses with strong network effects is a way to skew the odds in your favor. Knowing little about network effects, I recently did a deep dive and the result is this three-part series on network effects.
Part 1: Network effects and how they work
According to a study by James Currier, a four-time CEO and Silicon Valley VC, network effects have been responsible for ~70% of all value (measured in $) created by technology firms since 1994. Simply put, network effects are the key driver of value creation in today’s digital world.
Network effects are important because they are the best form of defensibility. Defensibility is your ability to protect your business from competition. This is different from your competitive advantage, which is what drives your growth initially, although the lines can get blurred. Defensibility is essential to a company’s long-term success because once product-market fit has been established, other competitors will try to invade the space and steal market share. Defensibility prevents this and allows a company’s value to grow non-linearly.
Understanding network effects (what they are, the different types, how to build them, etc.) allows founders to build category-defining companies and investors to fund them. They are also fascinating to learn about given their massive impact on our day-to-day lives. Yet most people don’t know what network effect are or how to leverage them.
And that is where we will start. Part one of this three-part series will explore what network effects are and how they work. Part two will focus on the different types of network effects and how to build them. Part three will touch on a few case studies to explore examples of network effects in businesses today.
Now, let’s get into it!
What are network effects?
Broadly speaking, networks are interconnected systems of people or things. Networks are a common part of our lives and are found almost everywhere, from sewage piping and roads to social media, the internet and human brains. Networks and network effects are most referred to from the perspective of businesses or products and this is the context in which I will be using the term as well.
Network effects occur when the value of a business or product increases with every new user (or increased usage). They occur when customer 2 adds value to customer 1.
But to truly understand what network effects are, it is important to explore how they work.
How do network effects work?
Network effects have six properties worth exploring:
1. Nodes and links
2. Network density
5. Critical mass
6. Value creation (a mathematical perspective)
1. Nodes and links
Networks are made up of two components: nodes and links. Nodes are the network participants. Common examples include devices (e.g., phones), buyers, sellers, and users. Links are the connections between nodes.
To use a tangible example, think of a city as a node, and a highway that connects two cities together as the link.
Not all nodes are the same. Different types of nodes can have different roles within the same network, and as a result, can differ in their levels of influence and value. Nodes can be categorized as either central or marginal. Central nodes have a high number of links and, as a result, are often more valuable to the network. Marginal nodes have less links and are typically less valuable.
Nodes are used to determine the size of a network, but size alone is not sufficient to determine network value. This is because activity (usage) is also a critical component.
Similarly, links differ in value based on two characteristics: directionality and strength (more on this in Directionality).
2. Network density
Network density is the ratio of links to nodes. The higher the ratio, the denser the network.
Denser networks typically have more powerful and valuable network effects because each link reinforces the strength of the other links.
Recall diagram 2? It will be easier for the marginal node to break away from the network. Why? Because nodes with fewer links, or weak links, derive less value from the network, and vice versa, the network derives less value from the node.
A more tangible example can be drawn from friendships. Imagine the following two scenarios:
You are friends with John
All your other friends are also friends with John
You are friends with John
All your other friends are not friends with John
Your bond with John is stronger in the first scenario because the network is intertwined. You are more likely to see him, spend time with him and be closer to him in scenario 1 than in scenario 2.
Network density is rarely uniform. Certain areas of a network are often denser than others, which leads to clustering.
Clusters occur when networks contain pockets of connectivity that are denser than the whole network.
Nodes (recall that these can be users, people etc.) in a cluster get more value from the network than non-cluster nodes. As a result, companies and businesses that have clustering capabilities within their networks tend to be more valuable.
The link between nodes can either be unidirectional or bidirectional.
In unidirectional links the flow of the interaction is directed from one node to another and is not reciprocated back. Flow of interaction can refer to many things, such as money, information, etc.
Unidirectional connections are common in personal social media networks. Twitter is a good example. Kim Kardashian has ~70M twitter followers and only follows 122 people. The flow of tweets is mostly one way. Most of her followers do not have reciprocal relationship with her.
Bidirectional links occur when the interaction flows both ways. A conversation (link) between two people (nodes), by necessity, is bidirectional. Same with Facebook Messenger, WhatsApp and other similar one-to-one communication platforms.
Unidirectional links are easier to scale and grow, but less sticky and defensible. A twitter follower can stop following Kim Kardashian and no side would lose significant value. Bidirectional links tend to be harder to scale, but also more defensible (you are inclined to stay in a 1 to 1 conversation with someone, even if it is terrible).
Networks usually have elements of both directed and undirected connections. But will often skew towards one more than the other.
5. Critical Mass
The value provided to users of products or businesses with network effects can be categorized into two parts: 1) the value of the product or service; and 2) the value of the network effect.
Critical mass in network effects refers to the point at which the value of the network exceeds the value of the product or service.
This can happen at different times for different types of products.
Zoom is a great example of a business that gains critical mass early. A Zoom user gets no value from the product unless someone else uses Zoom. The product value is completely useless without network effects. Once there is a second Zoom user, there is enough network-driven value to achieve critical mass.
Contrast this with an app such as Waze. Waze has stand-alone value: it provides detailed navigation. Users don’t need to learn and memorize how to get somewhere. But the more users use Waze, the more real-time traffic data the app can include to optimize the suggested route. It is only after a large number of people use Waze that the value of the network exceeds the stand-alone value of Waze.
Reaching critical mass should be the goal of every business or product that can have network effects.
When the value of the product is driven by the stand-alone value, the company needs to work hard to constantly improve the product and fend off competition.
When critical mass is reached, the company’s defensibility increases significantly. Each new user adds more value to the network, and the value accumulates much faster than if driven by the company. Competitors need to work harder to compete and keep up with the business.
A relatable product example is Instagram. What drives most people to use Instagram is all the pictures and stories that their friends post. This is all user-generated value that Instagram benefits from. Through network effects, Instagram has millions of people working for them to make its product more valuable.
The challenge is often to build enough initial stand-alone value to attract early adopters until critical mass is achieved.
6. Value creation
We have discussed the properties of network effects. But from an economic perspective, what makes them so valuable?
The answer boils down to the mathematical relationship between network size and value.
There are three general “laws” when it comes to prescribing value to a network:
Unlike the law of gravity, these are not concrete scientific “laws”. Rather, they are attempts to directionally capture the relationship between network effect size and value.
In my opinion, Metcalfe’s Law has the broadest application and is the easiest to learn first. 
Metcalfe’s Law states that the value of a network is roughly equal to the square of the number of nodes in the network. (Value = N^2-N)
The formula is based on the idea that each node in a network can establish a link with all the other participants. For example, if a network has 5 nodes, these nodes can make 20 possible links. The formula is: 5^2-5=20
If the network doubles in size to 10, the number of links doesn’t double; it more than quadruples. 10^2-10=90. 
Again, these are not scientific laws and their details can be debated academically. But fixating on that misses the point. The utility of these laws is that they provide a way to understand how powerful network effects are.
For part two we will look at the different types of network effects and how to build them.
Notes, Inspirations & Additional Readings
Thanks to Kerri and Kaleigh for their review and feedback.
 A study done by Zhang, Liu and Xu in 2015 tested Metcalfe’s Law based on Facebook and Tencent data. Using revenue as a proxy for value, their study strengthen the argument that Metcalfe’s Law is the best theory for understanding the value of networks. MAU was used as the criteria for determining the number of nodes.
 This model makes a few major assumptions (e.g., that all links and nodes are equally valuable, the network is not unidirectional, etc.), but it captures the essence of how powerful network effects can be.