The Power of Network Effects and how they work
This blog has been written based on the book “The Cold Start Problem” by Andrew Chen. It is an exceptional book written from his experience at Andresen Horowitz - a very successful and well know VC fund. We have struggled to find a concise resource that unpacks network effects and this was a great book to develop our understanding.
A network effect describes what happens when products get more valuable as more people use them. The network effect was first described by AT&T President Theodore Vail who described the telephone as useless without someone on the other end of the connection. As highlighted on the cover photo, AT&T advertised the value of connecting with others, not the telephone hardware. In the case of the telephone, there are two things at play; a) the physical telephone and b) the network of connections that you can call. A successful network requires both a product and its network. For example, Uber has an application (product) and active users to request a ride (network). In today’s world, many network effect products are made of software and the network is the people that are connected to each other.
How to tell if a business has a network effect?
1- Does the product have a network of users? Does it connect people with each other whether for commerce, collaboration, communication, or something else?
2- Does the ability to attract new users become stronger as the network grows?
Often the answers to the above are not binary and have different degrees of strength.
The History of Network effects
There was tremendous excitement about network effects in history and the 1999 Dot Com boom was amongst the greatest. From 1995 to 1999 the NASDAQ rose over 400%. In 1996 only 20M users had access to the internet and this connection was 24kbps dial-up speed. Today there are 5B internet users with an average speed of 80Mbps (3,000x greater). Despite how early we were in internet adoption the market became excited, at its peak, AOL was worth $224B and was one of the most valuable companies in the world. The early excitement stemmed from a principle called Metcalfe’s law which is defined as:
The systemic value of compatibility communicating devices grows at the number of connections squared.
This is arguably why valuations of internet and technology businesses are initially so confronting - the growth is non-linear and therefore difficult to value on a static valuation. As a company doubles its users the network's value is 4x not 2x. Metcalfe’s law however has some drawbacks because it is solely quantitative. It fails to take into consideration the quality of engagement, the differences in network types, and the crowding effect to mention a few.
Network effects don’t only exist in PC networks they exist in biology where some animals live longer when in a large group (safer from predators, finding food, etc). This is knowns as the Allee effect. There is a point known as the Allee threshold were below this point growth turns negative, These groups tend to go toward zero. It goes without saying these principles apply to technology products, until you hit density people don’t have a lot of incentive to connect, beyond this tipping point the benefits rapidly accelerate.
Networks can become overcrowded though for example when a communication product starts sending too many irrelevant messages or a social product filling up with irrelevant content or too many listings for a marketplace. This overcrowding among other contributors can lead to a network collapse, and this relationship works in a similar way, when a few nodes disconnect the value reduces slightly but when they reduce beyond the tipping point the value disappears rapidly.
Self-sustaining small networks and Slack
Given the power of network effects, it obviously begs the question of how to get beyond the tipping point to allow the network to create the perpetuating value. Chen outlines in the Cold Start problem that to start a new product with network effects the first step is to build a single, tiny network that’s self-sustaining on its own. This sounds straightforward but it is not.
An extreme example of building a small network was Tiny Speck. Tiny Speck was a company that raised $17M from top investors and was led by star founders who had done it all before, they went on to hire 45 developers and build a multi-player game called Glitch. It was a complete flop and within 4years it was shut down. It was observed during prior analysis that 97% of the players that signed up left within 5 minutes and given the benefit of the game was derived from the interaction with other players it seems clear in hindsight that the rapid churn was the issue. It simply never got the scale to produce any benefits for players. The great irony from this story was that the derivative effect of developing Glitch was that Tiny Speck needed a way for the 45 engineers to communicate with each other (The Tiny Speck Network). At the time (2009) communication tools supporting remote work were near non-existent. The team used an old tool called Internet Relay Chat (IRC) which was clunky, but it did the job, to improve it the team built some additional functionality to make their work easier. When Glitch eventually failed Tiny Speck focused on this IRC product which they renamed to “Slack”! it ended up being taken over for $28B by Salesforce.
Once companies started communicating with Slack it became self-sustaining as a communication utility, a slightly lower churn rate than Glitch!
The key to getting a small self-sustaining network cranked up is to capture the “hard side” of the network – the small percentage of users who do most of the work on the network. Once you have captured the network the point at which they become self-sustaining depends on the product, some examples are:
Slack: Requires 3 people to really work but according to founder Stewart Butterfield after 2,000 messages 93% of people are still using Slack today.
Zoom: You just need 2 people – one who wants to call someone else.
Airbnb: According to Co-Founder Nate Blecharczyk 300 listings with 100 reviewed listings was the magic number to see growth takeoff in a market.
Uber: ETA’s down below 3minutes on average and get 15-20 online cars on the road at the same time.
The overall key takeaway is to add a small group of the right people at the same time using the product in the right way. A common thread among the big network effect companies is that they all started small and built these small networks into large groups of networks.
The 1% - Hard Side of the network
As mentioned earlier a key contributor to building the first network is winning the key network contributors. Nearly every network has them, the starkest example is Wikipedia.
Wikipedia is one of the top 10 visited websites in the world and generates 18 billion views and 500 million visitors a month. It is a networked product with editors contributing articles and visitors reading them. Since it was founded in 2001, more than 55 million articles have been written and are over 90 times larger than the Encyclopedia Britannica. The amazing story of Wikipedia is that this enormous amount of content has been created by a minuscule population of only around 4,000 active editors. One extremely hyperactive editor is Steven Pruitt; he is a US customs officer during the day and in his spare time voluntarily edits for Wikipedia. He has made nearly 3 million edits and written 35,000 articles. He was named one of the most influential people on the internet in TIME magazine for his contribution.
When you look across user-generated products this is the norm, not the exception, some more examples are outlined below:
Uber: 100 million riders, 2 million drivers, 20% of these drivers create 60% of the trips.
YouTube: Two billion active users but 2 million users that upload videos.
Consumers are typically easier to capture and retain than contributors and this is mainly due to the 1/10/100 rule:
1% of the user population might start the group or thread within a group
10% of the user population might participate actively by authoring content or responding
100% of the user population benefits from the activities of all the above groups (Lurkers).
The 1% of users becomes extremely valuable to a network and this is because of what’s known as the “Power Law”. Instagram and YouTube both refer to this power law as the top 20% of influencers or content creators capturing most of the engagement, attracting millions of followers and tens of millions of views. It’s important to understand the motivations of these key contributors and focus on aligning product functionality and the business model towards them, some examples are below:
Social Media: Status – Social feedback loop. Would the creator be disappointed if no one saw it?
Communications Apps: Utility - Deepening relationships with friends.
Marketplaces: Economics – making money
Multiplayer Games: Fun and status – enjoyment.
Marketplaces tend to revolve around sellers, these are the users that are difficult to get. It’s referred to as “supply” and is the workers and small businesses who provide the time, products, and effort to try and generate some income on the platform. The first move for a marketplace is to build a critical mass of supply onto the marketplace. Once you collect the supply, typically you need to bring on-demand but then follow up with more supply quickly. The normal order of events is supply, demand, supply, supply, supply!
Sometimes the easiest supply to capture is assets or time that are underutilized:
Airbnb: Underutilization of guest bedrooms and second homes
Craigslist and eBay: Selling the stuff you don’t want anymore
Uber: Using your car to earn some income
Killer product – sometimes doesn’t need to be new or complicated
Zoom is the best example of a killer product unseating long-standing providers with the best killer feature of all - the “it works” feature. Zoom was founded in 2011 and grew from 10 million daily meeting participants at the end of 2019 to over 300 million just a few months later. The Zoom product that seemed too simple allowed a frictionless experience of inviting and having a meeting that ultimately enabled and accelerated the viral expansion which continued to reinforce the network effects and kept engagement high as its user base grew.
It is easy to think that Zoom’s simplicity was its competitive advantage, but this kind of simplicity is very hard to implement in practice. It is a distinctive quality of networked products - they do one thing well.
Networked products such as Zoom and Twitter facilitate an experience of interacting with other users. This is why they appear to feature not products. Contrast this to traditional products which attempt to build out functionality improvements which improve the engagement with the product.
Eric Yuan, Zoom founder, said when studying the network experience “...I studied the Dropbox pricing strategy and wondered why did they start charging at 2GB instead of 1GB and I realized the more you use a good product the more likely you are to pay – I started Zoom pricing at 40mins so people could get the full experience”. Networked products love to be free.
In summary, the ideal product to drive network effects combines both factors: simple to use and understand at the same time. It needs to bring together a network of users that is hard to copy by competitors.
User Experiences to build the network
One of the most important parts of building a good experience is avoiding what Chen refers to as “Zeroes”. These are events where the experience of the user is a zero. When building a network effect it’s crucial to avoid these user experiences, some examples are below:
Uber: Rider opens the app to hail a ride and there are no drivers available
Monday: User logs in and missing documentation needed for work that no one has filled out
Slack: Log into contact another user and that user hasn’t downloaded yet
Social Network: Sign up and no friends or favorite content is available
To avoid these zeroes is not as simple as adding a user, it needs constant activity to be built substantially. The real cost of a zero experience has lingering destructive effects because users churn and don’t believe in the reliability of the service so can be lost forever.
The Tipping Point
There are many ways to get to the tipping point but ultimately it is the point at which the virality, stickiness, and monetization start to scale on their own, “when a product can grow to take over a whole market”. There are many ways to nurture the scalability of growth such as:
Invite Only strategy: LinkedIn, Facebook, and Gmail used this approach. The main reason it works so well is that it naturally provides a good experience where your known connections and friends are already connected because they invited you – ensuring you get a good first experience.
Come for the Tool, stay for the network: Instagram is probably the best example. It unseated Hipstamatic because they attached a network to a photo filtering app. The tool attracts users, but the network makes them sticky. Some more examples are:
Instagram: Create photos + Share
Asana: Organize + Collaborate
GitHub: System of Record + keep up to date with others
Glassdoor: Lookup + Contribute with others
Paying up for Launch: It's sometimes smart to get the market quickly by paying up for density. Coca-Cola did this best with coupons - handing out free coupons to everyone to redeem at the store (resulting in lots of stores stocking Coca-Cola). Visa also did this by mailing out credit cards to attract merchants (resulting in the new Visa card being accepted at hundreds of stores).
Partnering: Piggybacking off an install base can leapfrog adoption and be a smart way to hit scale quickly. The best example in history of this is Microsoft/IBM. MS-DOS was custom built for IBM and got the immediate scale to attract the most developers, which then went on to attract users and then PC makers. MSFT went on to capture 80% market share for operating systems.
Competitive advantage of Network Effect businesses
In summary, the competitive advantage of a network effect business model comes from its users, not the product, and this is difficult to replicate. This has been highlighted several times with the best-case study being Airbnb vs Wimdu.
Wimdu vs Airbnb: The Samwer brothers in Germany almost exactly copied Airbnb’s website design and business model and scaled extremely fast raising $90M (The largest investment in a European start-up ever) and hiring 400 staff in less than 100days. Wimdu went on to attract 50,000 listings and $130M of revenue within a year. However, within 2 years Wimdu went to zero. All the scale they rapidly attracted was poor quality and the users had a poor experience. One of the early employees at Airbnb quoted:
“Not all supply is created equal, Wimdu’s top 10% of inventory was our bottom 10%. Customers were unhappy with their experience, and you need this high NPS to get viral growth”.
I would encourage anyone who enjoyed this blog to buy the book as it covers much more detail than I have highlighted here. I hope you enjoyed the read.
Shaun Trewin CA