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

13 min readJan 8, 2022


Part II: The Cold Start Problem

Chapter 4 — Tiny Speck (Slack)

When you start a new product with network effects the first step is to build a single tiny network that’s self-sustaining. You just need one to get started.

Tiny speck was building a multiplayer game called Glitch. It was a multiplayer game in the browser. The product however had poor retention — 97% who signed up would be out of there within 5 minutes. It was a leaky bucket. The only way Glitch would become fun was when a lot of people were playing it but the product never scaled to unlock that experience.

Years later TinySpeck relaunched with a second product, called Slack

To help the team build Glitch in 2009 they needed collaborative tools to communicate. It was all text. It was based on Internet Relay Chat otherwise known as IRC. IRC predated the web and was very old thus missed standard features. IRC didn’t have search. It didn’t store old messages. It was extremely confusing to find the right channels and right people to talk to. You had to download one of many IRC apps, connect to the right IRC servers and join the right channel.

The Tiny Speck team built a chat tool on top of IRC to store old messages, photos and make conversations easily searchable. This was the chat tool that glued work together. The service was hosted internally. This enabled team-wide collaboration and became part of the workflow at TinySpeck.

When it was clear that Glitch was not going to work they needed to pivot. So they moved to focus on the chat tool. It was rearchitected to allow any company to use it. Rebuilt with its own backend and solved all the issues IRC had. The second step was a private beta testing period. Stewart Butterfield — Co-founder of Slack said he wasn’t careful in targeting the initial base of customers to use the beta version of the app. He had friends at other companies and tried to convince them to use Slack. He managed to sign up 45 companies. The teams that adopted tended to be Startups. Fewer than 10 people, Just like Slack.

The minimum number of people to be defined as a team is 3. The slack team learned more and more as the team shared the product with larger and larger networks. Larger teams proved that channel discovery and coworker discovery was a problem. Each increase in team size required a rethinking of the design to form stable atomic networks that would grow.

A larger company might have 10’s of thousands of workspaces each with its own sets of channels. Each workspace might be a business unit. In later years enterprise sales became a huge accelerator. Companies like Slack, Zoom and Dropbox implemented bottom-up growth.

Introducing the Cold Start —

New products die when they flood their initial entry into the market. Every network product starts with just a single network. Small subscale networks naturally want to self-destruct. When people go to a product and they don’t see anyone they know using it they’ll leave. What solves this? The atomic network. The smallest network where there are enough people where everyone will stick around. These networks often have sides whether they are buyers and sellers or content creators and consumers. Generally, one side of the network will be easier to attract. The most important part of any early network is attracting and retaining the hard side of the network which is the small percentage of people that typically do most of the work in the community. For example, most of Wikipedia was written by a small number of editors, 5% of Uber’s users carry most of the load in the rideshare marketplace.

To attract the hard side you need to solve a hard problem. Design a product that is compelling to the key subset of your network. Tinder did this for the most attractive users in its network. The most successful network effects-driven apps are also sometimes dead simple. Zoom did an amazing job of this.

Chapter 5- Anti- Network Effects

For the first user on Slack within a workplace, the experience won't be great if there are 0 other users. How many users do you really need? 3 people, is the minimum amount to become a customer but it’s not enough to sign up they also need to be chatting over time. Eventually, they reach a threshold (around 2k messages) where they’ll stay on the product. After 2k messages, 97% of Slack users are still using slack today. How many users does your network need for your product experience to become good? Do analysis on the size of their networks x-axis and engagement metrics on the y-axis. For Uber — more drivers meant lower waiting time and thus more users. Facebook’s Famous Growth maximum — 10 friends in 7 days is a similar idea. Users that come in with more friends have higher retention. You want to maximize that. Every product has this threshold. Zoom has a low threshold required to form a stable network. Airbnb and Uber are two-sided which requires one or more parties involved given the hyperlocal constraints.

For the new products, it’s important to have a hypothesis for the size of your network even before you begin. Communication apps can be one-to-one. Contrast that to products that are highly asymmetrical with content creators and viewers or marketplaces with buyers and sellers. They require a much bigger number to hit the threshold and a bigger effort to get started. The size of the network helps determine a launch strategy. Solving the cold start problem requires a team to launch a network and quickly create enough destiny and breadth. You need the right people on the network. 10 ppl using slack all from the same team is better than 10 random people in the same company. The key is the atomic Network. Smallest stable network from which all other networks can be built.

Chapter 6- The Atomic Network- Credit Cards

Network effects normally start small. In a single city, college campus, or small beta tests. Once you nail the small network. You can build an adjacent network. You can copy and paste many times. Some companies grow team by team, city by city, or college by college.

Bank of America invented the credit card and picked Fresno California as an initial test site. Fresno was chosen partly because of its size. Fresno offered the critical mass the bank thought was necessary to make credit cards work and 45% of Fresno families did some business with BOA. Bank of America sent 60k Fresno residents' credit cards via mail ready to use. No application process. Credit card fees for merchants were set at 6%. Consumers received $300- $500 of instant credit. BOA initially focused on small fast-moving merchants not giants like Sears. Within a year they added SF / Sacramento / Los Angeles.

The Atomic Network

In Slack's case, the atomic network turned out to be pretty small. The threshold was the team of people around you and under 10 might be enough. Contrast that to the credit card that needs to launch to an entire city to make it work.

The network needs to have a dead-simple value proposition. The target should be on building a tiny atomic network focusing on density and ignoring the objection of market size. The attitude should be to do whatever it takes. Slack used an invite-only launch. The $5 referral fee used by Pay Pal. Dropbox’s demo video. Uber ice cream promotions. Each of these growth hacks gave a quick lift in helping to kick off future growth. For Slack, an early adopter team might start to use the product regularly until it starts to grow organically inside the company eventually the entire company upgrades to become paying customers. Rise and repeat.

Your product atomic network is smaller and more specific than you think.

Chapter 7- The Hard Side- Wikipedia

Even at the start of an atomic network, there is an important dynamic at play. There is a proportion of users that create a disproportionate value and as a result, have disproportionate power. This is the hard side of your network. They do more work and contribute more to your network but are that much hard to acquire and attain. For social networks, it's the content creators that generate the media that everyone consumes. For app stores, these are the developers that actually create the products. For workplace apps, these are the managers that author and create documents and projects who invite coworkers to participate. For marketplaces, these are usually the sellers and providers who usually spend their time attracting people to their products and services.

The Volunteers Who Built Wikipedia

With over 18 Billion page views and 55 million articles, Wikipedia was written by a small group of users, even though there are hundreds of millions of users. There are only about 100,000 active contributors per month. A small group of writers makes about 100+ edits in a month tt’s about 4,000 people. Meaning the active contributors only make up 0.02% of the total pool.

Across user-generated products, this is the norm, not the exception. There are nearly 100 million riders on Uber but just a few million drivers. There are 2 billion active users on Youtube but just a few million upload videos. This relationship exists everywhere.

The Easy Side Vs The Hard Side

Why is there a hard side at all? Because there is a side that simply requires more work. Users on the hard side have complex workflows expect status benefits as well as financial outcomes and will try competitive products to compare. Their expectations are higher. The hard side of the network creates a ton more value. Consumers are typically on the easy side of the network and are typically cheaper to attract and retain.

The hard side of a network should be able to answer detailed questions.

Who is the hard side of our network and how will they use the product? what is the unique value prop to the hard side and in turn the easy side? How did they first hear about the app and in what context? For users on the hard side as the network grows why will they come back more frequently and become more engaged? What makes them sticky to the network such that when a new network emerges they will stay on your product? These questions will give you a deep understanding of the motivations of our users.

The hard side of a network is important to understand for a product looking to launch its atomic network. You might consider that they are the most important group of users to start with. You need to know why your product will appeal to them starting on day one.

Chapter 8- Solve A Hard Problem Tinder

The hardest problem to solve in creating the first atomic network is attracting the hard side. For example, focusing on attracting content creators to a new video platform. Sellers on a new marketplace. PMs in a company to a new workplace app and then the other side of the network will follow.

The answer is solving an important need for the hard side — online dating

Dating apps are network effects-driven products that grow city by city and the more folks that join the network the more chances are that people will find matches.

The problem of too many love letters

Attractive people, particularly women are the hard side of the online dating network. The hard side needs to know how to sift through all the replies on dating apps. The hard side of the network needed more signals to help sort through all their matches. Tinder did this by integrating with FB so they could show the number of mutual friends they had which built trust. Tinder also made it so you can only be matched by people who were around you using the GPS on your phone (which was new at the time), so people with mutual friends living around you are the sort of people you might meet in real life. Built-in messaging creates a layer of trust. You can just unmatch without worrying about getting harassed.

Solving one of the important obstacles in the cold start problem

The hard side for marketplaces is usually the supply side. Marketplaces tend to revolve around their sellers. Uber's so-called power drives constitute 20% of the supply but create 60% of the trips. Most valuable users on the platform. The hard side is usually the supply side which is usually the workers and small businesses who provide the time, products, and effort and are trying to generate income on the platform. Sold as collectibles or coaching sessions. They often do this as alternatives to hourly jobs.

To solve the cold start problem for marketplaces. Often the first move is to bring a critical mass of supply onto the marketplace. eBay you start with the sellers of collectibles, for Airbnb you might start with people who have a few extra rooms in their place. For social platforms like youtube, it might be video creators, for Github, it’s helpful to bring key developers and open source products. After that, it’s time to bring demand. Once that’s working it’s all about supply again thus the order of operations is supply, demand, supply, supply, supply. While supply might be easy to get onto the network early on eventually supply might become a bottleneck.

Nights and Weekends

How do you find where the hard side of a network is engaged but their needs are unaddressed?

Looks a hobbies and side hustles.

There are millions of content creators, app developers, marketplace sellers, and part-time drivers that power the hard side of networks. They’re smart early adopters. Photo sharing and messaging products like IG and Youtube stems from countless amateur photographers and videographers that like to record travel / special occasions/architecture/beautiful people and everything else. What people are doing on their nights and weekends represents all the underutilized time and energy of the world. If put to good use this can be the hard side of an atomic network.

Chapter 9- The Killer Product — Zoom

The product idea matters. What makes an idea for a network product good or bad? Why was Zoom’s initial idea so non-obvious? Zoom seemed too simple, literally. Especially when WebX / Skype had already conquered the market. Zoom didn’t have more features but it had the most important feature of all. The “it works” feature. Zoom’s value proposition reinforced the network effects within a team and between companies by enabling frictionless meetings. Allowed attendees to enter within a click of a link. Frictionless usages created stronger network effects. More easily able to acquire users onto its network and frictionless enough to keep user usage high. Network products often do one thing well.

Network products everything else

Network products facilitate experiences that users have with each other. Whereas traditional products facilitate how users interact with the software itself. They grow and succeed by adding more users which cerate network effects. Whereas traditional products grow by building better features and supporting more use cases. Network products need to satisfy multiple needs of a network not just buyers but sellers too. Content creators and viewers. The most important features around network products are around how users find and connect with each other. Examples (Photo tagging/sharing permissions / or people you may know). The richness of this depends on who's on the network rather than the feature set.

The ideal product to drive network effects combines the product idea itself should be as simple as possible easily understandable by anyone as soon as they encounter it. At the same time, it should bring together a rich, complex infinite network of users that is impossible to compy by competitors. Zoom of course is a fantastic example of this idea.

Why Network Products Love To Be Free

Being free affects how killer products can spread rapidly. At one point Dropbox charged at 2 gigs instead of one because it gave users time to use it and the more you use it the more likely you’ll hit the cap and start paying. Zoom wanted to be the same way so they set the limit to 40 minutes per meeting. They wanted to make sure that users got the full experience of the product and if you liked it you’ll eventually pay. Even in the early days, these key product decisions made a huge impact on Zoom’s trajectory. The product combined an easy-to-use app that allows meetings to form quickly, had a simple value proposition, and was free to use initially. During beta release, the product was just a download button on a webpage. People at Standford tried it and wanted to pay. After the first customer, the viral growth continued. It was so viral he didn't have a marketing team for the first 4 years. Making zoom a freemium business means that it was easier for the network to grow. Offering a free tier is a recurring theme for many network products. Some are ad-supported, others have premium by subscription, others are powered by microtransactions.

Chapter 10- Magic Moments- Clubhouse

A product that hasn’t solved the cold start problem will fail to have any magic even in its early days. Often the network will seem empty. Once the network forms the magic moments happen all the time. Early on clubhouse (2020) hadn’t solved the cold start problem. When you opened the app it was often empty. The product also lacked key features. No profiles and you couldn’t follow other users. There were little bursts of magic, dropping in talking to friends he (Andrew Chen) hadn’t seen in months. Less than a year after its launch, clubhouse added millions of users a month after its launch. A large diverse group of networks began to form in every geography. Multiple clubhouse rooms were being formed, even some with celebrities and political giants.

The Clubhouse Story

Years before they worked on Phone-A-Friend, an app to connect groups of friends over audio. Then they worked on Uncalendar to fill spare moments with quick catch-up phone calls, later and more substantially TalkShow. which made it easier to produce podcasts. People could quickly start a podcast record with hosts, edit and publish all in one app. However, Talkshow was too heavyweight for the creators. It took a lot of setups and a lot of work for the hard side of the market. It simply took too long to get a magical moment. Early on the initial networks formed amongst tech early adopters which numbered in the thousands. They created magic moments consistently. The next 50K people brought Clubhouse to the mainstream culture. The Black creative community centered on entertainment and media hotspots like Atlanta Chicago NY and la started to join the network (mid 2020). This was propelled by musicians comedians influencers and creators hosting shows on a regular basis. This unlocked the next set of users growing to millions of people globally. The product started to have skyrocketing retention and engagement numbers. Audio content filled in an important void when we needed it in the pandemic.