Interactive essay / marketplace simulator
This essay deconstructs marketplace theory through live simulations. The goal is to sharpen the intuition needed to build a liquid, defensible, and durable business.
All models are wrong, some are useful. The numbers are indicative ranges drawn from operator frameworks and public case studies. Use these figures to visualize trade-offs and not to forecast a specific company's future.
Mental model
The smallest dense network that produces a happy transaction.
Supply and demand only matter where they can actually match.
Pricing, trust, matching, and tooling reshape the same system.
Uber-like, Faire-like, and Care-like marketplaces need different defaults.
The Seven Moves
Growth only compounds once you achieve a critical mass of early matches.
Ignore aggregate supply; the only metric that matters is local fill rate.
Your "rake" is more than revenue. It dictates how users behave on the platform.
Trust, frequency, and variety determine which growth levers actually work.
Different structures (B2B vs. C2C) shift the friction from the buyer to the seller.
High friction and high frequency create a massive incentive for users to take transactions off-platform.
Strategy is seasonal; your tactics must shift as you move from seeding to defending.
Module 1
The Cold Start is a momentum problem. The same dollar of spend can either build a flywheel or vanish into a void, depending on how quickly you generate early matches.
The cold-start problem is usually described as a paradox: buyers will not come without suppliers, and suppliers will not list without buyers. The operator version is harsher. Every failed search or empty calendar is a "tax" on your retention. The "Atomic Network" strategy works by shrinking the map: dominate a single neighborhood before trying to win a nation.
Scenario and levers
Start by choosing what kind of marketplace you are trying to seed. The sliders set the initial imbalance and the amount of paid help each side receives before the chart runs.
52-week cohort retention run
This is a cohort retention chart for what happens to a single wave of new customers, not a cumulative signup chart. The buyer and supplier lines show active people still willing to use the marketplace. They fall when too many people fail to match and churn faster than acquisition replaces them.
Why can the graph go down?
In the simulation, suppliers who get no transactions are less likely to return next week, and buyers who search but cannot find a match also leave. The cold-start problem is visible when active users shrink even though you are still spending on acquisition or subsidies.
Subsidy-dependent
54The marketplace survives, but liquidity is still fragile and subsidy-dependent.
Module 2
A healthy average can hide a dying neighborhood. This map reveals why the location of your supply is as critical as the volume of your supply.
Liquidity is the probability that a buyer can find what they want, and that a supplier can get used. The "Average Trap" is the quickest way to miss a failing market. For hyperlocal services like delivery or childcare, supply two miles away might as well be on the moon. The map turns fill rate into a geographic object so you can see why total counts can lie.
Controls
Red means demand cannot find usable nearby supply. Green means the relevant local radius is dense enough to transact.
Right now, 67% of demand lives in cells above the liquidity threshold. Radius changes who can match across nearby cells; distribution changes where supply starts.
Demand-weighted local liquidity
Module 3
The take-rate problem is nonlinear: extracting more can reduce supplier retention, buyer conversion, and ultimately platform revenue.
Take rate is not a passive accounting choice. A higher rake raises platform revenue per completed transaction, but it can also push suppliers to multi-home, raise buyer-facing prices, or move repeat customers off-platform. Gurley's 'rake too far' argument is that the best take rate is often below the maximum rate the platform can technically impose.
Product-nature sliders
Platform revenue curve
At this take rate, GMV is 72% of the maximum simulated GMV. Raising take rate by 5 points changes revenue by $18k and supplier churn is estimated at 21%.
Module 4
Product nature predicts which marketplace patterns transfer. A services marketplace can look more like Care.com than Airbnb even if both are consumer brands.
The essay's long catalogue of levers only becomes useful after you know what kind of product is being intermediated. Uber and DoorDash are high-frequency, local, time-sensitive, relatively standardized systems. Airbnb is lower-frequency, high-trust, and more heterogeneous. Faire is B2B, financed, and procurement-like. The radar makes those shapes comparable.
Select or configure
Frequent, local, perishable, standardized dispatch.
Implications
Biggest differences versus Faire
Shape explorer
Module 5
Business model is the transaction architecture: who brings supply, who owns demand, who pays, who has power, and which side is hardest to move.
Product nature tells you what is being exchanged. Business model tells you who is exchanging it and where power sits. A C2C resale marketplace, a B2B procurement marketplace, and a B2B2C delivery network can all be 'marketplaces,' but they have different hard sides, sales motions, trust burdens, take-rate headroom, and leakage risks.
Choose the architecture
C2C marketplaces coordinate many individuals on both sides. Supply is often unique or lightly professionalized, so search, reputation, and trust do much of the work.
Operator read
Hard side
Usually supply quality
Monetization
Seller fee, buyer protection fee, listing fee, or ads
Core question
Can enough trustworthy long-tail supply be made discoverable?
Biggest difference versus B2B
Sales cycle length differs by 62 points. This is where copied playbooks are most likely to break.
C2C value flow
Supply
Platform
Demand
Compare C2C vs B2B
Module 6
Leakage pressure rises when absolute fees are large, relationships repeat, and parties need to meet or communicate off-platform.
Leakage is the moment the platform creates a match but loses the transaction. Hagiu and Wright's framework is useful because it does not rely on vibes. If the fee dollars are meaningful, the same buyer and supplier are likely to repeat, and the work requires off-platform contact, the incentive to leave becomes structural. Babysitting, tutoring, home services, and care work all sit in the danger zone.
Risk inputs
Estimated leakage pressure
Moderate risk: invest in carrots
54Score is based on Hagiu and Wright's three leakage drivers. It should be read as directional.
Module 7
A marketplace stage is a pattern across fill rate, retention, growth, and multi-homing. Misdiagnosing the stage usually leads to the wrong lever.
The same metric means different things at different maturity stages. In cold start, low fill rate is expected and the task is finding a small network that works. At tipping, the task is removing subsidy crutches and strengthening retention. At the ceiling, adding more supply may not help because the bottleneck has moved to product quality, trust, category expansion, or defensibility.
Self-assessment inputs
Diagnosis
Estimated stage
Source backbone
The simulations translate these sources into adjustable mental models. The source links are included so the simplifications remain inspectable.
Atomic networks, hard side, cold start stages.
Minimum viable happiness, tipping, and domination.
Take-rate extraction versus marketplace health.
Fill rate, bookings growth, supply growth, GMV growth.
Fee size, repeat-purchase likelihood, and communication needs.
Network-effect types and defensibility patterns.
Thickness, congestion, safety, and simple preference revelation.
Multi-homing, competitive bottlenecks, and two-sided pricing.
Optimization, ETA prediction, and batching tradeoffs.