The Cold Start Problem
Free tiers used as the tool for reaching network critical mass.
Summary
Chen, a partner at Andreessen Horowitz who previously worked at Uber, addresses the specific problem of launching a network product that has no value until other people are using it.
His framework has five stages. The cold start problem itself, where the product is worthless until a minimum viable network exists, and the solution is to build the smallest network that provides value to its members, which he calls an atomic network, and to build many of these rather than trying to grow one large one. The tipping point, where the product acquires networks faster than it loses them, achieved through mechanisms including invitations, paid acquisition against known lifetime value, and what he calls the come for the tool stay for the network approach. Escape velocity, where growth becomes self-sustaining through the acquisition, engagement and economic effects operating together. The ceiling, where growth stalls because of market saturation, spam, degraded quality or overcrowding of the network. And the moat, where network effects become defensive against competitors.
The specific content on atomic networks is the most useful part. Chen's argument that Uber launched city by city and even neighbourhood by neighbourhood, and that a network product's minimum viable unit is far smaller than founders assume, is directly applicable.
The book draws on Chen's own experience and on interviews with operators at network companies. It is practitioner synthesis rather than research, the failure cases receive less attention than the successes, and the venture capital vantage point shapes which examples appear.
Key takeaways 7
- A network product must first build the smallest network that provides value to its own members rather than attempting to grow one large network.
- The minimum viable network is far smaller than founders assume, which is why launching city by city or campus by campus works.
- Building many small networks separately and then connecting them is more reliable than growing a single network from the centre.
- Offering a useful single-player tool first, with the network arriving later, solves the cold start problem for products that permit it.
- Growth stalls at a ceiling caused by saturation, spam, quality degradation or network overcrowding, and each requires a different response.
- Network effects become defensive only after the network is dense enough that a competitor cannot assemble an equivalent one.
- The economic effect of a network, meaning improving unit economics as it grows, is distinct from the acquisition and engagement effects and arrives last.
Who should read it
Founders and investors building or evaluating network products who need a staged framework for the launch and growth problems specific to them.
Who recommended it 1
Similar books 6
Chosen by shared topics, the entries that hold them, and the people who recommended both.
Economics for information goods: pricing, versioning, lock-in and network effects.
Recommended by: Mohnish Pabrai, Tim O’Reilly
How platforms create value by matching groups, unlike linear pipeline businesses.
Recommended by: Reid Hoffman
Digital costs near zero mean goods can be given away, which changes whole industries.
The arithmetic of conversion rates, whales and lifetime value.
How multi-sided platforms such as card networks get both sides on board at once.
Practitioner account of how platform businesses capture a transaction.
Where it sits in the reference 2
Topics