If you’ve spent any time researching how to build a marketplace, you’ve almost certainly come across Andrew Chen’s writing. As a general partner at Andreessen Horowitz (a16z), Chen has become one of the most influential voices on marketplace growth, network effects, and the mechanics of getting two-sided platforms off the ground.
His book, The Cold Start Problem, distilled years of research into a framework that marketplace founders around the world now use to think about their launch strategies. His essays on network effects, viral growth, and marketplace dynamics have shaped how an entire generation of operators approaches the challenge of building platforms where both sides need each other.
Here’s what marketplace founders should take away from his work — and how to apply it.
The Cold Start Problem
The central challenge Chen identifies is what he calls “the cold start problem” — the idea that networked products are fundamentally useless until they have enough users to create value. A marketplace with no sellers has nothing for buyers. A marketplace with no buyers gives sellers no reason to list.
This isn’t just a chicken-and-egg observation. Chen’s framework breaks the cold start problem into specific, solvable stages:
1. The Atomic Network. Every marketplace needs to identify the smallest possible network that can sustain itself. For Uber, that was enough drivers in one city to guarantee a pickup within a few minutes. For Airbnb, it was enough listings in one neighborhood to give travelers real choices. For a freelance marketplace, it might be enough designers in one specialty to ensure a client gets quality proposals within 24 hours.
The mistake most founders make is going too broad too early. They launch nationally (or globally) and end up with a platform that has a few users scattered everywhere — and enough density nowhere. Chen’s work reinforces what experienced marketplace operators already know: start small, build density, then expand.
2. The Tipping Point. Once you’ve built an atomic network, the goal is to hit the tipping point where the network starts growing on its own. This is where supply attracts demand, which attracts more supply, which attracts more demand. Chen argues that reaching this point requires deliberate effort — it rarely happens organically.
3. Escape Velocity. After the tipping point, the marketplace enters a phase of rapid growth driven by network effects. The product gets better as more people use it, which draws even more people in. This is the phase where marketplaces like Uber and Airbnb went from interesting startups to dominant platforms.
The “Hard Side” of the Marketplace
One of Chen’s most practical insights is the concept of the “hard side” — the side of the marketplace that is hardest to attract and retain. In most marketplaces, this is the supply side.
Uber drivers have to pass background checks, maintain their cars, and dedicate hours to driving. Airbnb hosts have to prepare their homes, deal with guests, and manage logistics. Etsy sellers have to create products, photograph them, write descriptions, and handle shipping.
The hard side does more work, takes more risk, and has more alternatives. And if the hard side leaves, the marketplace collapses — because the easy side (buyers) has nothing to purchase.
Chen’s advice: invest disproportionately in the hard side. Build tools that make their lives easier. Reduce friction in their onboarding. Create economic incentives that make the platform worth their time.
This maps directly to how the most successful marketplace startups operate. They obsess over supplier experience, supplier tools, and supplier economics. The buyer experience matters too, of course — but the supply side is usually the bottleneck.
Network Effects Are Not Created Equal
Chen’s work at a16z also distinguishes between different types of network effects, and this distinction matters for marketplace founders:
Direct network effects — The product gets better as more users join. Each new user on the same side adds value. Social networks have strong direct network effects.
Cross-side network effects — More users on one side attract more users on the other side. This is the primary network effect in marketplaces. More sellers attract more buyers, and more buyers attract more sellers.
Negative network effects — At some point, adding more users can actually decrease value. Too many sellers competing for the same buyers leads to supplier churn. Too many options overwhelm buyers and reduce conversion.
This last point is underappreciated. Many marketplace founders assume that more supply is always better. But if your marketplace has 10,000 listings and buyers only browse 20 before making a decision, adding 10,000 more listings doesn’t help buyers — it just dilutes attention from existing sellers.
Managing negative network effects requires curation, search quality, and recommendation algorithms that surface the best matches. It also requires monitoring your marketplace data to detect when supply oversaturation is hurting seller economics.
The Law of Shitty Clickthroughs
One of Chen’s earlier essays — before his a16z days — introduced “The Law of Shitty Clickthroughs.” The idea is simple: every marketing channel degrades over time as it becomes saturated. Banner ads worked great in the 1990s. Email marketing had incredible open rates in the early 2000s. Facebook ads were cheap in 2012.
For marketplace founders, the implication is that your growth strategy can’t rely on a single channel forever. The channel that gets you from 0 to 1,000 users probably won’t get you from 1,000 to 100,000. You need to continuously find new acquisition channels and, ideally, build organic growth loops (referrals, SEO, word of mouth) that aren’t subject to the same degradation.
This is why Chen emphasizes building products with inherent virality — where using the product naturally exposes new people to it. When a host shares their Airbnb listing, or a seller shares their Etsy shop, or a freelancer links to their Upwork profile, the marketplace gets free distribution.
What a16z Looks for in Marketplace Investments
Chen and the a16z team have publicly shared their framework for evaluating marketplace startups. The signals they look for include:
- High fragmentation on the supply side — many small providers rather than a few large ones
- Recurring transaction patterns — not one-time purchases but ongoing relationships
- Ability to own the transaction — the marketplace can insert itself into the payment flow and capture a take rate
- Strong unit economics — the customer lifetime value exceeds the acquisition cost by a meaningful margin
- Defensible network effects — as the marketplace grows, it becomes harder for competitors to replicate
These evaluation criteria are useful for any marketplace founder, not just those seeking venture funding. They’re a checklist for whether your marketplace model is fundamentally sound.
Applying These Ideas to Your Marketplace
Chen’s frameworks are valuable, but they’re abstractions. Applying them requires understanding your specific marketplace’s dynamics — and that requires data.
Which side is your hard side? Look at activation rates and churn by user type. Are you at your tipping point? Track liquidity over time — is the percentage of listings that result in transactions increasing? Are you experiencing negative network effects? Monitor seller revenue per seller as you add more supply.
These aren’t questions you can answer with gut feeling. They require systematic tracking of marketplace-specific metrics — GMV, take rate, buyer-seller ratio, cohort retention, and supply utilization.
How Twosided Helps You Apply These Frameworks
The gap between understanding Chen’s frameworks and actually applying them is data. You know you should track liquidity, but are you? You know you should monitor your hard side’s health, but can you see supplier-level retention cohorts?
Twosided is built to close that gap. It’s the analytics OS for marketplace startups — giving you instant access to 150+ metrics that map directly to the concepts Chen writes about. Liquidity dashboards, cohort analysis, supply-demand health, champion identification — all available within minutes of connecting your Stripe Connect or Sharetribe data.
Andrew Chen gave marketplace founders the theory. Twosided gives you the data to put it into practice. Try it free.