The Cold Start Problem: How to Start and Scale Network Effects by Andrew Chen (Book Notes)

9 min readJan 1, 2022


Part I Network Effects

Chapter 1 — What’s A Network Effect Anyway

A Network effect is what happens when products get more valuable when more people use them

Uber — More users join the app the more users would quickly find someone to take them from point A to point B. Also, this meant it would be easier for drivers to fill their time with trips increasing their earnings.

Classic Example: Telephones

  • Its value depends on the connection with the other telephone and increases with more connections

Fundamental duality

  1. The physical product, the telephone
  2. The network of people and physical wiring that serve to interconnect the phones

A successful network effect requires both a product and its network

For Uber, the product is the app that people run on their phones and the network refers to all the active users at any given time who are connecting with Uber to drive or ride. In this example, the product is made up of software and the network is made up of people.

The Billion Users Club

Companies like eBay /Open Table / Uber / Airbnb are examples of marketplace networks comprising buyers and sellers.

Dropbox, Slack, and Google Suite are workplace collaboration tools built for the network of your team members.

Instagram, Reddit, Tik Tok, Youtube, and Twitter are networks of content creators and consumers, and advertisers.

Apple — 1.6 Billion iOS devices

Google — 3 billion devices

Facebook — 2.85 Billion users across their social network and messaging apps

Microsoft- 1.5 devices running windows + 1 billion running office

They all have network effects so that their product becomes more useful the more people use them.

The Network

The Network is defined by people who use the product to interact with each other. For Youtube the network is defined by software, it is the content uploaded by creators and viewers that watch them.

They connect people but they don’t own the underlying assets. The value is in connecting the guests with their hosts with Airbnb. Apple doesn’t own the devs that publish apps to the App Store. Youtube doesn’t own creators or its videos. The entire ecosystem stays on because the value is in bringing everyone together.

… and the Effect

The effect part of the Network Effect describes how the value increases as more people start using the product.

People stay on the network and use it more because other people are also using it more.

How do you tell if a product has a network effect and how strong is it?

Important questions are

Does the product have a network? Does it connect people whether commerce collaboration communication or something else at the core of the experience?

Does the ability to attract new users to become stickier or to monetize become stronger as its network grows longer? Does the user face a cold start problem where retention is low when there are no other users?

It’s critical to understand how they launch. How they grow and scale and how they compete.

New products are difficult because we are in a zero-sum time of attention.

Currently with constrained attention from users / Fierce competition / limited marketing channels to access new users/network-based competitors / unclear future application platforms creates intense pressure on the industry. when a new product leverages Network Effects adjacent industries can be disrupted quickly.

Chapter 2 — A Brief History

There are weak advantages to being first since the winning startup is usually a later entrant.

The winner usually doesn’t take all but has to battle other networks in different geographies and different customer segments.

One key theory popularized in the dot com era presented a highly flawed view of network effects that theory is Metcalfe’s law.

Metcalf’s Law: The systemic value of compatibly communicating devices grows as a square of their number. Said plainly, each time a user joins an app with a network behind it the value increases to n squared.

This theory originated in the ’80s.

In the late ‘90s, it became popular to promote Metcalf’s law to mark up valuations and first movers.

Anyone who’s built a Network from scratch will tell you that’s it’s irrelevant.

It leaves out important aspects of building a network, like what you should do right at the beginning when no one is using your product. It doesn’t encompass the quality of user engagement or the multi-sidedness of many networks (i.e. buyers and sellers). Nor the difference between active users vs people who just signed up. OR the degraded version of a product when too many ppl sign up.

Metcalf’s law is a simple academic model that fails when faced with real-life messiness.

If Metcalf’s law is broken then what’s better?

Meerkat’s Law

Social animals: Study of Meerkats, Sardines, Bees, Penguins

These benefits by being together, more nodes in these networks are better whereas if the population were to decrease the benefits can quickly go away making them more susceptible to collapse. If the population grows too quickly then overpopulation negates these advantages which cause the population to plateau. Social animals have network effects too.

The Math of Meerkats

  • Meerkats are hyper-social
  • Live together in groups of 30–50
  • Referred to as gang or mob
  • One will stand on its legs and alert the group of an intruder and its urgency level

Goldfish grow more rapidly and can resist water toxicity when they’re in groups. Flocks of birds can confuse and resist predators. Meerkats warn eachother of danger. This concept is called the Ali Threshold — which is a point at which animals would be safer and ultimately grow faster as a population. When there are not enough meerkats in the mob to warn either of the predators it’s more likely that one will get picked off by a predator. After that, it’s a circular dynamic because with even fewer meerkats they’re less likely to protect themselves leading to a smaller and smaller population.

The biological to product comparison is simple here. If there are not enough people on a messaging app most people will delete it. As the userbase shrinks it will be more likely that each user will leave.

What happens when there’s a healthy mob of Meerkats? More and more mobs get formed. If you come in above the Ali Thresholdthe population will grow because they can be healthy and protected. Even if predators pick one or two off as long as the overall population stays high it will keep growing. But it can’t last forever because of limited resources based on the environment. Often called the carrying compacity. The Network Effect version of this happens when there’s overcrowding from too many users. For communication apps — too many messages social apps to many contents and feeds for marketplaces to many listings so finding the right thing is hard. Add the right features to aid discovery/ combat spam and increase relevance in the UI you can increase the carrying capacity for users.

When populations and networks collapse.

When the oceans are overfished. The population of sardines, tuna, and other fish can tip over. Collapsing in just a few short years. Tech product with Network effects becomes the same way. They become a bit less useful when people leave and fully collapse below the tipping point the other way.

Early 1900’s

Rows of sardine factories were built

They harvested hundreds of thousands of tons of sardines each year.

Given its a small fish around 5 billion sardines are caught each year at the peak harvesting.

All of a sudden it stopped in the 1950s

The fish didn’t come back for the next year or the following and the same the year after the Sardines were gone.

In the early years, a sardine catch was 800 million tons. Just a few decades later it collapsed to 17 tons.

Overfishing spelled the end for the Monteray (California) fishing industry

Sardines have network effects

The Ali Curve At Uber

When there are few drivers in the city it takes a long time to get a ride this is called having a high ETA.

Conversion rates are low because who has time to wait 30 mins for a ride.

Unless you have a few dozen riders the value to the user is nearly 0. They won’t really use the app and drives won’t stick around either so the network will essentially dissolve.

Once you pass a tipping point things start to work — rides can come in 15 mins which isn’t the best but still useable. Get it down to 10 or 5 then even better. The bigger the network of drivers the more convenient it gets. The rideshare network in the city starts to see the classic network effect.

Meerkat’s vs Metcalf’s

After all, humans are social animals we share, photos, collectible sneakers, projects, splitting dinner expenses.

Rather than hunting a mating out networks help us with groceries and dating apps.

The same dynamics apply to us and a mob of meerkats.

You want to come in above the Ali Threshold is what we like to call the tipping point.

Chapter 3- The Cold Start Problem

This framework is a new way to think about Network Effects split into stages.

Cold Start Theory — Most important stage in Network Effects.

Cold Start Theory Has 5 Primary Stages:

  1. The Cold Start Problem
  2. The Tipping point
  3. Escape Velocity
  4. Hitting the Ceiling
  5. The Moat

Stage 1 — The Cold Start Problem

Most new networks fail.

If a new video sharing platform launches and doesn’t have a wide variety of content ready early on users won’t stick around. The same is true for marketplaces and social networks. If users don’t find who or what they want. They’ll churn. In most cases the Network Effects that startups love so much actually hurt them. Anti-Network effects are destructive.

Solving the cold start problem requires getting all the right users and content on the same network to work at the same time. Execute and launch.

To solve the cold start problem. Build an atomic network (the smallest stable network that can grow on its own i.e. zoom video conferencing network can work with just two people vs Airbnb that requires hundreds of active listings in a market to become stable.

Stage 2 — The Tipping point

To win a market you need to build many many more networks to expand into the market. For example, Tinder’s successful initial launch at USC unlocked other colleges nearby, then into different cities (LA). Each launch makes adjacent networks easier and easier and easier to unlock until the momentum becomes unstoppable. This is why we see the most effective network effects grow city by city. Company by company, or campus by campus. Saas products normally grow inside of companies, landing, and expanding. Also jumping between companies as products gets passed over by consultants and employees.

Stage 3 — Escape Velocity

All about working to strengthen network effects and stay in growth.

  1. Acquisition Effect- Let’s product tap into the network to drive low-cost, highly efficient user acquisition via viral growth.
  2. Engagement Effect — increase interaction between users as networks fill in
  3. Economic Effect- improves monetization levels and conversion rates as the network grows

The Acquisition Effect is powered by viral growth. A positive early user experience that compels one set of users to invite others onto the network. For example. Paypal‘s viral referral programs and LinkedIn’s recommendations for connecting people are two examples that increase the power of acquisition effect.

The engagement effect Manifest’s itself by increased engagement as the network grows

  • Pulling them up the engagement ladder
  • Introducing new use cases via incentives
  • Marketing and communications
  • New product features

Uber did this by leveraging people up to airport trips /dining out /daily commutes.

Economic effect — Directly affects the products business model improving it by (increasing conversions and key monetization flows / ramping up revenue per user as the network grows)

Slack — As more teams adopt the software they’re more likely to convert into a paying customer.

Fortnite- Customized costumes and guns will monetize better as more of the users friends join the game together

Stitched together all lead to a flywheel that can monetize into billions of users

Stage 4 — Hitting the Ceiling

This is when a network hits the ceiling and growth stalls.

Driven by a variety of sources such as CAC that spike due to market saturation and as viral growth slows down. Similarly, there is a law of shitty click-throughs that drives down the performance of acquisition and engagement loops over time as users tune out of marketing channels. There are fraudsters, overcrowding, and context collapse. All aspects of a network that grows and matures. In the real world product tend to grow rapidly, then hit a ceiling, and as the team addresses the problems another growth spurt emerges. Each one gets more complex over time.

Stage 5 — The Moat

Network Effects to compete is tricky when other companies in the same product category can leverage the same effects. Every workplace collab tool can leverage network-driven viral growth to increase stickiness and strong monetization as users arrive. Same for marketplaces and messaging apps. This poses a Network-based competition about how one product ecosystem challenges another. Airbnb faced this when a European competitor (Wimdu)arrived. Airbnb had to fight them off by competing on the quality of their network and scaling its effects.