“Every successful networked product starts with an atomic network — the smallest set of users or participants that delivers enough value to sustain itself.” — Andrew Chen
The instinct when building a networked product is to think big. Launch everywhere. Get everyone. Build the global network. This instinct is almost always wrong, and following it is how promising products die.
The correct instinct is the opposite: start with the smallest possible network that can sustain itself. This is the atomic network — the indivisible unit from which all larger networks are built. Finding it and filling it is the solution to the cold start problem.
Chen illustrates this with a deceptively simple question: what is the minimum number of users for your product to deliver real value? For Zoom, the answer is two — any two people who want to video call. For a marketplace, it might be enough buyers and sellers in a single city. For a dating app, it might be enough single people in a single neighborhood.
The atomic network isn’t just the minimum — it’s the right minimum. It’s the smallest configuration where every user in the network experiences enough value to stay and invite others.
Not all networks have the same structure. Chen identifies several patterns:
Products like Uber, Airbnb, and eBay connect two distinct groups — buyers and sellers, riders and drivers, hosts and guests. For these products, the atomic network must include both sides.
An Uber with no drivers is worthless. An Uber with no riders is worthless. The atomic network requires enough of both, in close enough geographic proximity, to create value. This is why Uber launched city by city rather than nationwide — a nationally thin network would be an everywhere-useless one.
The key insight: In two-sided markets, you must solve both sides of the cold start problem simultaneously, which is why they’re especially hard to launch.
Products like messaging apps, email, and video calling have a deceptively small atomic network — in theory, just two people. But in practice, the relevant atomic network is the social graph of a specific community.
WhatsApp didn’t grow by getting random pairs of users. It grew by embedding itself in specific communities — first among people who knew each other in India and Brazil — and spreading through existing social ties. The atomic network for a messaging app is a dense social cluster, not a random pair of strangers.
For platforms like YouTube, Reddit, or TikTok, the atomic network includes both creators and consumers. But there’s an interesting asymmetry: you can fake the consumer side in the beginning with archived content.
Reddit’s founding team created thousands of fake accounts and posted content themselves until the community was real enough to sustain itself. YouTube accepted uploads from anyone, building a content library before a true community existed. The atomic network for a content platform is dense enough content to keep readers engaged — and readers engaged enough to attract creators.
Chen uses Uber as a master class in atomic network thinking. The cold start strategy was built around a single insight: geography is everything for a ride-sharing app.
The atomic network for Uber in any city was roughly 30 drivers. With 30 drivers concentrated in a dense urban area, wait times dropped below five minutes — the magic threshold where passengers experience the “this is amazing” moment. Below 30 drivers, wait times were too long, drivers sat idle, and the network felt broken.
Uber’s launch in San Francisco in 2010 was a masterclass in atomic network construction:
The result was that Uber’s earliest users experienced a product that worked remarkably well. The 5-minute wait time wasn’t an accident — it was engineered by managing the size and geography of the atomic network.
Understanding the atomic network also means understanding what happens when you get the size wrong.
Many products launch nationwide or globally before establishing dense local networks. The result is a product that technically has millions of users but delivers almost no value to any of them — a thin network.
Foursquare had millions of users worldwide but not enough in any given restaurant or venue to make check-ins feel meaningful. The social experience was sparse everywhere because the network was spread too thin.
The opposite mistake is staying too small — either through exclusivity or slow growth — such that the network never develops self-sustaining momentum.
Some products try to be too curated, restricting growth so aggressively that the network never reaches critical mass. The art is finding the size where the network is dense enough to deliver value but open enough to grow.
For anyone building a networked product, the practical question is: what is my atomic network?
Chen suggests several diagnostic questions:
The answers define the target for your cold start strategy. Your job is not to build the global network — your job is to build one great atomic network, then repeat.