CompanyGraph
A platform earns a share of transactions while its rules decide whether buyers and sellers can keep using the market.
A platform sells coordination, not just access
A marketplace can look simple in a revenue table: transaction value passes between a buyer and a seller, and the operator records a fee. The physical and commercial service is larger. Search, ranking, identity checks, payment, fraud controls, dispute handling, delivery tools, and customer support determine whether a transaction can be completed and trusted. A platform may also supply demand, but the seller still carries inventory, labour, returns, taxes, and the risk that a sale is cancelled.
The take rate is normally described as platform revenue divided by a defined transaction base. That definition is useful only when the numerator, denominator, refunds, advertising, fulfilment, and payment services are specified. A marketplace can report a stable commission while the seller’s total cost of participation changes through new advertising requirements, fulfilment charges, or ranking rules. The number is an observation of a contract and an accounting boundary; it is not a complete description of the economic relationship.
Price is allocated across sides
In a two-sided market, buyers and sellers respond to one another. A platform may charge buyers little and charge sellers more, or subsidize one side with advertising or subscriptions. Rochet and Tirole’s two-sided-market model shows why the allocation of prices between sides can matter as much as the total price: the side that is more responsive to participation may need a lower charge for transactions to happen at all.
There is therefore no universal “correct” take rate. A seller compares the platform’s incremental demand and protections with its commission, advertising, fulfilment, service, and customer-acquisition costs. A buyer compares price, selection, delivery, reliability, and the ability to obtain redress. A higher fee can be sustainable when the platform adds a service that participants cannot cheaply reproduce. It can also be unsustainable even when the platform has a large network, if suppliers can move repeat customers elsewhere or if the rules make their investment unsafe.
Governance changes the transaction itself
Governance is not a soft layer added after pricing. Eligibility rules decide which sellers can enter. Search and recommendation systems decide which offers are visible. Quality standards, identity checks, payment holds, refunds, and dispute procedures decide whose mistakes are absorbed and who bears the loss. A rule can improve trust while raising operating cost; it can also increase short-term conversion by shifting risk to a participant who has less bargaining power.
Because these rules select which participants remain, governance affects the supply available to buyers. A clear suspension process may remove fraudulent inventory and make honest sellers more willing to invest. An opaque ranking change may make a seller’s prior spending, stock, or staff suddenly unproductive. The platform can describe a policy in a help page, but the economically relevant observation is how the rule changes exposure, payout, appeal, and the ability to continue trading.
The posted commission is not the seller’s whole bill
Etsy’s 2024 annual filing illustrates why the boundary matters. It describes marketplace revenue from transaction, listing, and payments-processing activities, and seller services such as on-site advertising and shipping labels; it also describes a seller set-up fee introduced in 2024. Those categories do not establish what any particular seller pays, but they show why a single commission percentage cannot stand for the full cost of reaching a buyer.
For an investor, the useful calculation is cohort-specific: total platform charges and required spending divided by the seller’s relevant sales, after refunds and fulfilment costs, compared with the margin the seller could earn through a realistic alternative. A seller with a differentiated brand and a direct customer list may leave after a modest increase. A small seller with no discovery channel may tolerate a much larger burden. Reported take rate, seller payout, repeat purchase, seller concentration, and off-platform leakage should be read together.
CompanyGraph tracks the margin print live: companies whose gross, operating, and net margins all sit elevated, the gross and net legs benchmarked against industry peers.
Three Margin Ratios Elevated Across Gross, Operating, And Net Levels
Industry-benchmarked gross margin, operating margin (mapped against own scale), and industry-benchmarked net margin are all in elevated ranges
Margin level is the recorded outcome. The screen cannot separate pricing power from mix, cost timing, or one favorable year, and it says nothing about durability.
Amazon shows how fees and visibility interact
In its 2023 complaint against Amazon, the Federal Trade Commission alleged that Amazon charges sellers selling fees and advertising fees, and that it can make sellers offering lower prices elsewhere less visible in search. These are allegations in litigation, not established findings. They nevertheless provide a concrete test of the concept: the seller’s economic burden is shaped not only by the fee deducted from a sale, but also by whether keeping visibility requires additional spending and whether an outside price can be used without losing access to demand.
The case also shows why governance should not be reduced to a moral label. Ranking, pricing, and advertising rules can be designed to prevent bad experiences, protect a platform’s investment, or increase extraction; the same rule can have different effects depending on enforcement and alternatives. The investor’s task is to identify the mechanism and then check the observed response: seller entry and exit, inventory quality, advertising dependence, price dispersion, complaints, refunds, and repeat transactions.
Where take-rate economics can fail
- Gross share mistaken for profit. Platform revenue still has to fund payment losses, fraud prevention, support, infrastructure, refunds, and compliance. A high take rate is not evidence of high economic profit.
- Average burden hiding exposure. A blended rate can conceal that a few large sellers receive concessions while small sellers pay more, or that advertising is necessary for visibility.
- Temporary extraction mistaken for pricing power. Revenue per order can rise while order volume, seller investment, or service quality deteriorates. The lag may be long when participants have inventory or contracts to work through.
- Governance costs omitted. Trust and safety, appeals, moderation, and fraud controls consume staff and money. Cutting them can improve a current margin while damaging the service that makes the platform useful.
- Network effects treated as permanent. Multi-homing, direct customer relationships, open standards, regulation, and a rival’s better terms can reduce switching costs. A large network creates an advantage, not an immunity.
What investors can actually test
- Reconcile reported platform revenue with the transaction base, refunds, advertising, fulfilment, and payment services used to calculate it.
- Track seller or supplier contribution margins, concentration, retention, repeat purchasing, and the share of demand that can move to another channel.
- Read policy changes beside operating data. A fee increase accompanied by better conversion or lower fraud is different from a fee increase accompanied by declining supply quality and rising complaints.
- Examine who controls ranking, data access, customer contact, appeals, and payment timing. These determine whether a participant can correct a problem or merely absorb it.
- Separate allegations, platform self-descriptions, accounting measures, and observed participant behaviour. None alone proves that governance is healthy or extractive.
Platform governance is durable when participants can earn enough from the relationship to keep supplying, buyers continue to trust the result, and the operator can finance the controls that make transactions possible. The concept does not predict the right fee from a single percentage. It asks whether the platform’s pricing and rules preserve the activity from which that percentage is taken.