Network Effects in Everything

A network effect is when a product or service becomes more valuable as more people use it. The original example is the telecommunications network, which becomes more valuable to its users as more people are connected to it.

The idea is intuitive. It’s no use being the only person able to make a telephone call. When your friend joins the network, the ability to make a call is now valuable. As more people join the network, the ability to make a call becomes more and more valuable.

The value of the network is proportional to the square of the number of users (Metacalfe’s law)

The telephone example is known as a same sided network effect, as users increase the value of the network for other users. So, what’s a two-sided network effect? This is where demand creates a network effect that increases the value of being a supplier delivering the product/service, or vice versa.

Uber is a good example of a two-sided network effect. As more riders join Uber, being an Uber driver becomes more attractive, as they will have more customers to serve, and they will make more money. When more drivers join Uber, being a rider becomes more attractive, as it becomes easier to get a driver, and waiting times are reduced. Similarly, Spotify becomes more valuable for the listener as more artists join and stream their music on the platform. Spotify becomes more valuable for the artist when more listeners sign up, as now there is a bigger audience for their work.

The perennial question when it comes to network effects is how do you get them started. Why would you sign up for Spotify as a listener if there are no artists on the platform? Why would you put your music on Spotify if there are no listeners? Similarly, a driver won’t be interested in signing up to Uber if there are no riders, and riders get no value from the service if there are no drivers. It is kind of a chicken and the egg problem.

One way to overcome this problem is to subsidise those who you want to join the network. Uber did this by offering free rides to those who got their friends to sign up. Spotify do it by offering a free version of their product. This is a way to get the ‘flywheel’ going. Once it gets going, it begins to pick up momentum, as every time someone joins the network, joining becomes more attractive for those currently outside the network. If the network reaches critical mass, the subsidies used to entice users can be removed, as the value of the network alone is now enough to attract new participants.

The extraordinary thing about network effects is how powerful they can become. Uber launched in 2010 and reached their first billion trips on the service five years later in 2015. It only took them six months to record another billion trips and reach 2 billion overall. In 2017, the following year, 3 billion trips were facilitated by the platform, bringing the overall total to 5 billion. Only one year later they had doubled this figure again, hitting 10 billion trips overall in 2018.

It’s easier to appreciate these numbers with a graph, so I put a (very) rough one together. Up and to the right as they say..

Rough data taken from here — only landmark figures are provided so I had to estimate other data points.

However, while network effects can fuel extraordinary growth, they can also unravel just as quickly. Think of the early social media platforms, such as Bebo and MySpace. Both had very strong network effects and grew rapidly, but suddenly there was a mass exodus of users, and things rapidly fell apart. This is always a lurking threat for products/services fueled by network effects.

I mentioned Stripe on a previous post. In a nutshell, they help companies sell their products/services online — their mission is to increase the GDP of the internet (geeky and great!).

When a company using Stripe’s payment platform (the Global Payment Treasury Network — GPTN) accepts a payment from a customer online, Stripe take a % of that transaction. What is great about this business model is how it aligns Stripe’s incentives with their users incentives. Said another way, when Stripe’s users do well, Stripe does well. This leads Stripe to come up with some really creative and inventive ways to help their users succeed. It’s a virtuous cycle!

One said creative and inventive way Stripe help their users succeed is by optimising the online payment process. The GPTN reduces the number of online card transactions that are declined, without increasing the number of fraudulent transactions. This can drastically increase revenues for businesses powered by Stripe. For example, Stripe just announced that Postmates saw a $70 million uplift in revenue after adopting their payment products. There was a similar story with Twilio earlier this year.

How do they do this? Well, Stripe handle a lot of transactions (250 million every day, roughly 13,000 every second..!!). They don’t disclose the exact cumulative value of these transactions, as they’re a private company, but the most up to date publicly available ballpark figure is hundreds of billions of dollars every year. Here is where the network effect comes in. After handling so many transactions, Stripe’s GPTN, powered by machine learning, is really, really good at knowing what causes a transaction to be declined at a granular level at a particular bank. This is important because different details matter at different banks. For example, different banks prefer different date formats on payment requests. Using dd/mm/yyyy will work better at some banks than dd/mm/yy, and vice versa. This is just one example, but in an antiquated payment system with so many different players, there are so many more quirks, and Stripe’s GPTN understands them better than anyone else. Then, Stripe leverages that knowledge to adapt details on payment requests, something they call ‘Stripe Adaptive Acceptance’, to give the requests the best chance of authorisation. As Will Gaybrick, Stripe’s Chief Product Officer explains, Stripe updates their GPTN 16 times (!) every single day. Each update is based on data generated by their users payment transactions, improving the payment authorisation rate, and thus revenue, for all Stripe users and Stripe. This is the data network effect in action.

So, what inspired the title of this post? Well, what I love about the phenomenon of network effects is how you begin to see them everywhere once you learn about them. The video below is an unconventional example, but it fits the theory pretty well — have a look!

Joining the guy to dance wasn’t too attractive when he was alone, but as more people joined, the group gained momentum, and joining to dance became a more and more attractive proposition. When the group reached critical mass, it began to grow rapidly!

This example is slightly tongue in cheek, and I’m sure many would dispute labeling this as a network effect.. I just thought it was a fun way to end the post.

Thanks for reading!


Some words on what I’m thinking about. Interested in technology, literature, music, football, reclaiming our cities for the pedestrian. Writing to learn.