The Cold Start Problem: How to Start and Scale Network Effects by Andrew Chen Part IV — Escape Velocity (Book Notes)

9 min readJan 23, 2022


Chapter 17 — Dropbox

When networked products start to work. They can really work. By the time of dropbox’s IPO in 2018, cofounders of Dropbox had built a Saas startup that was the fastest in the category to reach a billion dollars in annual recurring revenue. Many of their key metrics looked like a hockey stick. to over 500 million users in 8 years.

Dropbox solved the cold start problem with the classic come for the tool strategy with files syncing across personal computers as the initial tool, followed by a network of shared folders by colleges, friends, and family. They innovated with a referral program where users can give and get storage by inviting friends. User growth was explosive. By 2012 they had 100 Million registered users. High expectations pushed up the valuation to 4 billion dollars. The time arrived for Dropbox to focus on making money at 200 employees. Dropbox wanted to focus on consumers and photos but by a few years in they started hiring people to run marketing and sales and they let self-serve happen on its own. This allowed users to upgrade through their credit cards. Just from having this upgrade page they had 10 of millions of dollars in recurring revenue without monetization being a top priority. The bottom-up approach was working.

Dropbox’s insights on their users were profound. Some users joined Dropbox as part of the come for the tool strategy but stayed with the tool without increasing their engagement by sharing documents with others. In contrast, the ones who used Dropbox for collaboration and sharing, the network features. Because of significantly more valuable over time. Dropbox’s users could be divided into two categories, LVA low-value actives and HVA high-value actives which were useful as a quality indicator. To make sure HVA’s would be acquired not LVA’s.

Dropbox’s focus on businesses came from multiple directions. First, the team surveyed users and realized that many HVA were upgrading their Dropbox accounts for use at work. It was much easier to sell to businesses, especially when Dropbox built features that aligned with businesses like administrative controls, extra security, and integrations into other products like Microsoft Office. When they analyzed the type of files that were at the center of user engagement. For example, which types of files were edited and moved over and over again? Which files are shared, collaboratively edited, and interacted with? They found documents/spreadsheets and presentations. Dropbox in years before its IPO began to orient itself in a new direction. Focus on the Highest value users in its highest value networks interacting with its highest value files.

Introducing Escape Velocity

When new products see success and start to scale it’s often called hitting escape velocity. In this phase, the problem quickly becomes maintaining a fast growth rate and amplifying successful product’s network effects. In Dropbox’s case, the product went through multiple phases. The cold start phase began as a replacement for USB drives as a tool, while upselling users to use shared folders. As consumer and business use cases multiplied the product reached the tipping point leading to hundreds of millions of active users. In the Escape Velocity phase, the company needed to continue scaling the userbase. The key to this phase was to understand the high and low-value users and then focusing in on the workplace. Every new product has to achieve and sustain escape velocity.

Chapter 18- The Trio Of Forces

3 Systems underlying the network effect

The Network effect is not one effect. It can be broken down into a trio of forces.

The Acquisition Network effect

The Engagement Network effect

The Economic Network effect

The Acquisition Network effect

Is the ability for a product to tap into its network to acquire new customers. Only network products can tap into viral growth. The ability for users in its network to tell others In their own personal networks. Features that amplify network effects are oriented around viral growth. Referral features. Tapping into contacts to create suggestions. Improve conversion along key moments in the invitation experience. All these bring down the CAC.

The Engagement Network effect

How a denser network creates higher stickiness and usage from its users. Use cases and loops that define how users derive value while engaging with a product. Twitter is a lot more interesting to use with media outlets celebrities and politicians on it rather than in the early days when there were only a few people on the platform. A more diverse set of use cases now arrive like tracking political news, keeping abreast of what’s happing in your industries, and keeping up with celebrities. These elevated use cases drive key metrics of the product.

The Economic Network effect

The ability for a network product to accelerate its monetization, reduce its costs, and improve its business model as the network grows. Workplace models for example often convert to higher tiers of pricing as more knowledge workers grow within a company. The more workers that adopt a product the more advanced features they’ll want to upgrade to. Especially if the features are collaborative in nature. An example of this is Slack charging for the ability to search through all messages in a company.

Chapter 19- The Engagement Effect Scurvy

Users are often divided into separate groups called cohorts. Which then allows them to be monitored separately. How many are still around the day after they sign up against 7 days or 30 days? Are new users having a better experience in the first few weeks vs an older cohort that was using a bigger version of the product? Cohort retention curves give us insight into whether a product is working or not.

The sad truth about the stinkiness of new products

Retention is the most critical metric in understanding a product. Most of the time the data is not pretty. When you look at the engagement data for the entire industry it’s told the story over and over. Users don’t stick to their apps. Nearly 1 in 4 people abandon mobile apps after only 1 use. Of the users who install an app 70% of them arent active the next day. By the first 3 months, 96% of users are no longer active.

How new use cases drive more engagement

When a small team in a company adopts Slack they may only use a few use cases. As they bring more employees onto the product new use cases are unlocked. There might be a pool party channel or channels for each office to announce location-specific events for SF/NY. At a16z they have hashtag books or hashtag 2030. Each of these new channels has a new use case. The more people on the Slack network the more that these cases will develop.

LinkedIn’s userbase was tiered on frequent usage. At Linkedin, they segmented the users as active last 7 days out of the last week. Active the last 6 days. Active the last 5 days and so on. It let them dig into each segment separately and helped them understand their needs/motivations and what it would take to move them up in engagement.

Depending on the type of users/motivations and intent, different approaches will work. Early users might just need a few more connections to colleges. Power users might need to discover advanced features on search, recruiting, and creating groups to have new and more powerful ways to connect with people. Segmenting users gives them the right tools to educate users at different times to impact their usage.

Engagement Loops

The Engagement Network Effect makes products stickier over time. But how? This process can be modeled as an engagement loop that shows how users derive value from other users in a network in a step-by-step process. For social or communication products. It often starts with a creator posting or creating content. The content is then sent to everyone they're connected to. Depending on the size of the network they get a nice string of likes and comments back. That’s the payoff that keeps them going. Marketplace products have a similar loop where sellers list their goods which are then seen by buyers browsing through the listings. The larger the network of customers the more likely an interested buyer will see it, which is more likely that a transaction will happen. Collaboration tools work in a similar way. A member of the hard side of the network initiates by sharing a product or document. Then coworkers engage to close the loop.

If the network is too sparse the loop is broken. Not enough users will see a photo to reply with likes and not enough users are on a platform to purchase a product. If a loop is broken when a user churns which further cascades the problem. If the network is too small and too Inactive and a loop breaks, they’ll be less likely to use it in the future. However, if the product scales and the loop gets denser then the loop gets tighter.

Chapter 20- The Acquisition Effect- PayPal

The ability for a network to attract new networks as it scales. This is one of the most magical and explosive forces in the technology world. Viral growth. The PayPal Mafia made growing products into a science. Early Youtube, provided an embeddable player that could be added to any blog or myspace profile. Linkedin’s use of email contacts that connect with your work colleagues. Or Eventbrite’s email invitations to an audience of potential attendees.

PayPal started with a product called FeildLink which allowed people to send and receive money with palm pilots and other PDA’s. For this product to work you needed the sender and the receiver of payments to both have PDA’s for this to make sense. It wasn’t going to work. That’s when PayPal was born.

The idea evolved. Let people send money across the internet no handheld is required at all. This became PayPal. It was as easy as clicking on a link, signing up, then you can send and receive money over the internet. In fact, you had to sign up to receive money and they might send money to others as well which would make other people sign up as well. Even with this they still had slow growth. The product hadn’t found its killer use case. David Sacks was the leading product at PayPal. They didn’t have a clear picture of the ideal users they didn’t know who to target. The We Use PayPal button. There were hundreds of listings that were using “We use PayPal” on their eBay listings and it was up to the team to supercharge that growth.

Growth started to pick up after they gave 10 dollars to every PayPal user who invited a friend. Also dropped $10 into the account of the person as soon as they signed up. This supercharged viral growth to insane amounts.

Product Driven Viral Growth

Network products are unique because they can embed their viral growth Into the product experience itself. When a product like dropbox has built-in features like folder sharing It can spread on its own. Pay Pal’s badges and core user-to-user payments accomplish the same. This is the product and network duo at work again. Where the product attracts people to the network and the network brings more value to the product.

Amplifying the Viral Factor one step at a time

Uber’s viral loop involved a referral program that was exposed during the onboarding process. There were a dozen or so screens that a driver moved through in a signup process. Then drivers would be shown how to refer their friends and what kind of bonus they’d get for doing so. At each step, you can optimize for speed and streaming.

Measuring the Acquisition Network Effect

Pay attention to the ratios between each cohort of users. From 1000 to 500 to 250. This is often called the viral factor. Calculated at 0.5 because each cohort generates users at the next cohort at that rate. In this example, things are looking good. As you optimize the acquisition process your viral factor will continue to approach 1. After all at a viral factor of 0.95. 1k users show up and bring 950 of their friends. This is the mathematical expression of when a product actually goes viral. This may feel like a spreadsheet product. This is not copywriting user psychology and product design.