Two-Sided Markets: The Model Underneath Every Marketplace
The Two-Sided Model
A two-sided market has two groups of users who need each other. The platform's value to group A depends on how many people from group B are also on it, and vice versa. Without one, the other finds the platform useless.
A few examples:
Platform Side A (demand) Side B (supply)
Uber riders drivers
Airbnb guests hosts
Etsy buyers sellers
OpenTable diners restaurants
App Store app users app developers
Tinder people dating other people dating
Stock market buyers sellers
The last entry matters: two-sided markets are not new. Financial exchanges, classifieds, auction houses, and bazaars are all two-sided. The internet just lowered the friction enough to build them at much larger scale and for much narrower use cases.
Cross-Side Network Effects
Cross-side network effects are the defining property of a marketplace. More drivers make Uber more useful to riders. More riders make Uber more useful to drivers. The effect goes both ways, and it compounds.
More supply → better choice for demand → more demand → better pull for supply → more supply → ...
This loop is the whole reason marketplaces are valuable. Get it spinning and you have a compounding business. Fail to start it and you have a website.
The strength of the effect varies:
- Strong cross-side effects: riders strongly prefer to use the app with more drivers; drivers strongly prefer the app with more riders (Uber, Lyft)
- Weaker cross-side effects: buyers on Etsy do not deeply care whether there are 50,000 or 500,000 sellers, as long as there are enough in the category they want. The effect has a plateau
When planning a marketplace, the question to ask is: how much does side A care about side B's size? If the answer is "a lot", you have a high-compounding business and a hard cold start. If it's "some, up to a point", you have an easier start and a weaker moat.
Same-Side Network Effects (Often Negative)
Same-side effects describe what happens when more users on your side of the market affect your experience. For supply, they are usually negative: more sellers means more competition for the same buyers, meaning lower prices and more churn.
More sellers → more competition → lower margins → supply churn
This is why well-run marketplaces manage supply quality carefully. Letting anyone list anything floods the platform with low-quality supply, which degrades the experience for the demand side too (matching becomes harder) and for the supply side (they can't make money).
Demand-side same-side effects can also be negative. Rush hour at Uber means riders compete for the same drivers. Airbnb during peak travel season means guests compete for the same listings. The platform's response (dynamic pricing) is partly a mechanism for smoothing same-side congestion.
Some marketplaces have positive same-side effects on the demand side: the more buyers, the more reviews, the richer the trust infrastructure. That is a data network effect layered on top of the two-sided model (more on this in chapter 7).
The Chicken and Egg Problem
On day one, no one is on the platform. A seller who lists finds no buyers. A buyer who searches finds no sellers. Neither comes back.
This is the coordination trap. It is not an engineering problem. It is not a marketing problem. It is a sequencing problem.
Four strategies have worked, roughly in order of frequency:
1. Subsidise one side
Pay users on the constrained side to show up. Uber gave drivers guaranteed hourly rates when the platform was thin. DoorDash paid restaurants' fees to get them to participate. Airbnb offered free professional photography to early hosts.
Money solves the coordination problem directly. It is expensive and has to be stopped eventually, but it works when the unit economics eventually justify it.
2. Single-side-first (tools, then marketplace)
Give one side a useful product that works without the other side being present. Once that side is locked in, add the marketplace.
OpenTable started as reservation software for restaurants. Restaurants paid for it because managing reservations is annoying. Once many restaurants used OpenTable, they added the consumer-facing marketplace (diners searching for tables).
Houzz started as a home design community (demand side). Professionals joined because their clients were on Houzz.
Honeybook started as a CRM for freelancers. Once freelancers were on it, their clients became a pool worth monetising.
This strategy has worked repeatedly. The tools side is a foothold. The marketplace comes later.
3. Piggyback on another marketplace
Airbnb famously scraped Craigslist (then and now) to seed their supply. Early PayPal grew on eBay. Instacart listed on its own app but got supply by using the aisles of existing grocery stores (which were themselves a kind of physical marketplace).
If another platform has the users you need, the cheapest cold start is to hijack it (within the bounds of its terms of service, or just outside them if you are willing to accept the risk). Most marketplaces that did this early have since been cut off by the platform they piggybacked on.
4. Pick a very small seed market
A global marketplace is impossible on day one. A local one is merely very hard. Airbnb's first market was one conference in San Francisco. Uber's first market was drivers and riders in the Mission District. DoorDash's first market was four restaurants near Stanford.
The principle: density beats scale. A marketplace that works in one neighbourhood is a much better business than one that sort-of-works in a dozen cities. You build out from dense success, not across thin potential.
The Cold Start: A Sequencing Framework
For any new marketplace, ask:
- Which side is harder to get? Usually supply. Supply often has lower marginal willingness to show up for speculative demand
- What can I give the harder side before the other exists? Tools, guaranteed revenue, cheap acquisition to the first side, photographer visits
- What is the seed market small enough to fill? One conference, one city, one university, one niche
- What signal tells me it's working before I scale? Almost always some form of repeat transactions per user in the seed market
Most failed marketplaces skip step 3 and try to launch globally. The one that works is local, dense, and boring for the first year or two.
Why This Matters for Everything Else
The two-sided model is the base on which every other marketplace concept sits. Liquidity (chapter 3) is the measurable form of the cross-side effect. Matching (chapter 4) is the platform's attempt to compensate when the effect is weak. Pricing (chapter 5) is partly a mechanism for balancing supply and demand when they drift. Network effects (chapter 7) are what happens once the cross-side loop is spinning hard.
If you internalise only one thing from this tutorial, make it this: marketplaces compound when both sides grow together, and collapse when they don't.
Common Pitfalls
"We have a strong cross-side effect, so we just need to get started." Strong effects are double-edged. They punish you in the cold start as hard as they reward you once liquid. Plan the subsidy budget before you raise
"The supply side will appear once demand is there." Sometimes. Often not. Most marketplaces get supply first because supply has a clearer reason to show up (money) than demand (curiosity)
"We can manage a two-sided market the same way we managed our SaaS product." Different metrics, different acquisition funnels, different churn drivers, different pricing. Everything is two products
"We will solve cold start with marketing." Marketing accelerates a flywheel that already turns. It does not start one. If the underlying cross-side loop is not working, ads make the problem bigger by bringing in users who churn
Next Steps
Continue to 03-liquidity.md to learn the single metric that decides whether a marketplace is working.