Sean Ellis vs Andrew Chen: Growth Experiments or Network Effects?
Comparing Sean Ellis's growth hacking and experimentation with Andrew Chen's theory of network effects in The Cold Start Problem: two approaches to growth, and when each applies.
Two of the most cited writers on startup growth approach it from different sides. Sean Ellis gave the field the term growth hacking and a repeatable method of experimentation. Andrew Chen, a former head of growth at Uber, wrote about the dynamics of products whose value rises with users. People compare them to decide whether growth is a process or a property of the product.
What is the core difference?
Ellis assumes that growth comes from finding and exploiting levers across the funnel, through many small experiments, run quickly, scored by expected impact. The method applies to most digital products.
Chen assumes that for some products, growth is shaped by how the network behaves. The challenge is to get the first small network working, then reach a tipping point, rather than to tweak a funnel.
One offers a method for improving growth. The other explains why some products grow differently.
What does Ellis say?
Sean Ellis coined the term growth hacking in 2010 after working at companies such as Dropbox and LogMeIn. Hacking Growth (2017), written with Morgan Brown, describes a process: align a cross-functional team, define a north star metric, generate ideas, prioritize experiments, run them, and analyze the results. He popularized the ICE scoring method of impact, confidence, and ease for ranking experiments.
He also proposed the 40 percent very-disappointed test for product-market fit. His approach is practical and drawn from startup work, and he has built a business around training and advice.
What does Chen say?
Andrew Chen is a general partner at the venture firm Andreessen Horowitz and was previously head of rider growth at Uber. His 2012 essay Growth Hackers Are the New VPs of Marketing helped popularize growth as a discipline. The Cold Start Problem (2021) offers a theory of network effects and the stages by which a network product grows.
Those stages begin with a small atomic network, followed by a tipping point, escape velocity, hitting the ceiling, and building a moat. He draws on products such as Uber, Tinder, and Slack. The framework applies to products whose value depends on other users.
How do they compare?
| Dimension | Chen | Ellis |
|---|---|---|
| Core idea | Network effects and the stages of a network | A repeatable process of growth experiments |
| Scope | Products whose value rises with users | Most digital products |
| Key concepts | Atomic network, tipping point, escape velocity, moat | North star metric, ICE scoring, experiment cadence |
| Typical question | How do we get the first network working? | Which experiment should we run next? |
| Origin | Experience at Uber and venture investing | Early-stage startups and growth teams |
| Main risk | Applying network logic to a product without network effects | Optimizing small levers in a product that does not retain |
| Best known for | The Cold Start Problem (2021) | Hacking Growth (2017, with Morgan Brown) |
When does each one fit?
Ellis's method fits when you have a product with some traction and need a disciplined way to find what drives growth. It is especially useful for improving activation and retention, where small changes accumulate.
Chen's theory fits when your product's value depends on other users: marketplaces, social products, collaboration tools. It helps explain why early efforts should concentrate on a tiny network, rather than spreading thin.
For a product with weak retention, neither growth tactics nor network logic will help until the core experience works. Ellis's product-market fit test is a first check.
What does this look like in practice?
A new marketplace for local tutors struggles to attract both sides. A Chen reading says to focus on one neighborhood and one subject, reach enough tutors and students that a booking is likely, and expand from there. The atomic network is the unit of work.
An Ellis reading adds the experimentation layer. Within that neighborhood, test messages to tutors, onboarding steps, and referral prompts, scoring each experiment by impact, confidence, and ease. The theory chooses where to concentrate, and the method finds what works inside it.
"My product is [description] and it is [a marketplace, social product, or other type]. First, using Chen's cold start ideas, tell me whether it has network effects, and if so what the smallest atomic network would be. Then, using Ellis's method, list ten growth experiments for that network, score each for impact, confidence, and ease, and pick the first three."
Why it works: the theory narrows where to focus, and the scoring picks what to try first.
Can you use both together?
Yes. For a network product, Chen's stages tell you what you are trying to achieve at each point, and Ellis's experiments are how you get there. For a product without network effects, Ellis's method stands on its own.
Both assume a product worth growing. If retention is low, the first job is the product, not the funnel.
"Here is our retention and usage data: [paste]. Tell me whether we have evidence of product-market fit using Ellis's very-disappointed survey as one measure and retention curves as another. If not, suggest what to fix before running growth experiments, and which network or funnel stage we are in."
Why it works: confirming fit first prevents optimizing the growth of something people do not keep using.
Where to go next
Sean Ellis and Andrew Chen both sit in the Marketing & Sales category. For the originals, read Hacking Growth by Ellis and Morgan Brown and The Cold Start Problem by Chen. To keep a weekly operating rhythm as a founder, Founder Weekly Review is a $129 tool built for it.
Frequently asked questions
Who coined growth hacking?
Sean Ellis introduced the term in 2010 to describe someone whose true north is growth, a person who tests ideas across the product and marketing to find what scales. Andrew Chen helped popularize the idea in his 2012 essay Growth Hackers Are the New VPs of Marketing. Hacking Growth (2017), by Ellis and Morgan Brown, set out the method more fully. Both men were part of the early wave of Silicon Valley growth practitioners, which is why their names are often linked.
What is the 40 percent rule for product-market fit?
Ellis proposed asking users how they would feel if they could no longer use a product, and treating a result of 40 percent or more answering very disappointed as a sign of product-market fit. It comes from his work with startups and is a heuristic, not a law. It works best with a meaningful sample of active users. Ellis suggests asking the question of people who have used the product at least a few times, and treating the result as one signal alongside retention.
What is the cold start problem?
In Chen's The Cold Start Problem (2021), it is the challenge of getting a product that depends on network effects off the ground, because it is worthless without users on both sides. He describes stages: starting with an atomic network, reaching a tipping point, achieving escape velocity, hitting the ceiling, and building a moat. The idea is that a network product can be empty and useless at first, so the early work is to make a very small network valuable on its own terms.
Do these approaches conflict?
No. Ellis's method is a repeatable process of experiments for finding growth levers in most products. Chen's theory explains how products with network effects grow and what is distinctive about their early stages. A network product can still use Ellis's experiments, and Chen's stages help decide which experiments matter at each stage. For example, Chen's tipping point tells you when to stop concentrating on one city and begin to expand, which then shapes which experiments to run.
Which is better for a non-network product?
The documented work does not rank them. If your product's value does not depend on other users, Chen's network framework applies less, and Ellis's experimentation approach is more directly useful. Chen's book is a theory of a particular class of products, so check that yours is one before applying it. Ellis's own sequence puts product-market fit first, so a product that fails that test needs work on the product before growth tactics.
Can AI run growth experiments?
It can brainstorm hypotheses, score them, draft tests, and summarize results. It cannot run the tests or see your users. Use it for structure, and make decisions from real data with enough volume to be reliable. It is good at generating and scoring ideas, and weak at knowing which of them your particular users will respond to without testing. Keep a log of what each test predicted and what happened.
Written by Gareth Hoyle. Last updated 8 October 2026. Part of the authority.md guides library.
More guides.
Daron Acemoglu vs Ha-Joon Chang: Do Good Institutions Cause Growth, or Follow It?
Comparing Daron Acemoglu's institutions-first account of national prosperity in Why Nations Fail with Ha-Joon Chang's historical critique of the institutions literature: two views of development, with the main point of disagreement. General education.
Bill Ackman vs Carl Icahn: Two Styles of Activist Investing. Not Investment Advice
Comparing Bill Ackman's public, thesis-driven activism with Carl Icahn's decades of corporate pressure campaigns: how each approaches companies, boards, and the public. Not investment advice.
Adam Smith vs Ha-Joon Chang: The Founder of Market Economics and One of Its Historical Critics
Comparing Adam Smith's Wealth of Nations and Theory of Moral Sentiments with Ha-Joon Chang's history-based critique in Kicking Away the Ladder: what Smith actually argued, what Chang adds, and where they differ. General education.