Marketplace Metrics: GMV, Liquidity and Take Rate

By Brian Tr · Updated Oct 7, 2026

A marketplace doesn't sell a product the way a store does. It sells matches between two groups of people, and takes a cut when a match turns into a transaction. That structure changes what you have to measure. Revenue alone is too late and too narrow. Traffic alone says nothing about whether anyone actually transacted. And a single blended number can hide the fact that one side of the market is quietly leaving.

This article is the hub for marketplace measurement. It defines the core metric stack, shows how the numbers connect, separates the ones that flatter from the ones that warn, and explains how the stack shifts as a marketplace matures. Where a metric already has its own deep reference, we summarize it and link out rather than repeat it.

The basic arithmetic: GMV, take rate, net revenue

Three numbers sit at the top of almost every marketplace dashboard, and they're linked by one equation.

GMV (gross merchandise value, sometimes called gross bookings or GTV for total volume) is the total value of the transactions that flowed through the platform. Take rate is the share of that volume the marketplace keeps. Net revenue is what lands on the income statement.

GMV × take rate = net revenue

Public companies define these terms slightly differently, and the differences matter when you compare them. Etsy's FY2025 10-K defines GMS as the dollar value of items sold in its marketplaces, excluding shipping fees and net of refunds, and says plainly that it doesn't represent revenue earned. It defines its revenue take rate as revenue divided by GMS. Uber's FY2025 10-K defines Gross Bookings as the total dollar value, including applicable taxes, tolls and fees, of rides, delivery orders and freight revenue, without adjusting for consumer discounts and refunds, driver and merchant earnings, or driver incentives. It describes its Revenue Margin as revenue as a percentage of Gross Bookings.

Notice the choices buried in those definitions. One includes taxes and fees, the other excludes shipping and nets out refunds. Neither is wrong. But a "GMV" figure from one company isn't directly comparable with another's until you know what's inside it. When you define your own GMV, write down what's included (taxes, shipping, tips, cancelled and refunded orders) and don't change it.

The equation also explains why GMV and revenue can move in different directions. Etsy's own numbers show it. According to the FY2025 10-K, GMS fell 5.3% to about $11.9 billion in 2025, while revenue rose 2.7% to about $2.9 billion, because the revenue take rate rose from 22.3% to 24.2%. Revenue held up even though the volume shrank, because the share kept went up. That's a useful reminder that growth in GMV and growth in revenue are separate questions. For the mechanics of setting and changing the percentage, see marketplace take rate.

The full metric stack

The a16z essay 13 Metrics for Marketplace Companies, by Jeff Jordan, Li Jin, D'Arcy Coolican and Andrew Chen (February 2020), is the most widely used checklist. It groups the numbers into matching efficiency, structure, economics, competitive position, and retention. The table below follows that spirit but organizes the stack by the question each metric answers.

Each deserves a short explanation.

Liquidity measures

The a16z list opens with match rate, market depth and time to match. Match rate asks how successfully the two sides find each other. Market depth asks whether enough supply exists that a user opening the app can find what they need. Time to match measures the gap between supply being created and finding demand, which the essay also frames as inventory turnover. These are the buyer-side and supply-side readings of one idea: the chance that showing up leads to a transaction. We cover definitions, denominators and the density distinction in marketplace liquidity, and where the structural cause lives in marketplace density.

The practical point for a metric stack: pick one buyer-side and one supply-side liquidity metric, define the denominator carefully, and cut it by category and geography. A blended number is an average of strong and weak pockets.

Repeat rate and cohort retention

Volume is easy to buy and hard to keep. The a16z essay includes user retention cohorts, core action retention cohorts, and retention by location. The idea is to compare groups of users by when they joined and watch whether newer cohorts retain better than older ones. Improving cohorts are a sign the network is getting more valuable. Flat or worsening ones suggest growth is being bought, not earned.

The marketplace-specific version is cohort GMV retention: take everyone who first transacted in a given month and plot the GMV they generate in each following month. If the curve flattens above zero, you have a repeat business. If it decays toward nothing, every dollar of GMV had to be acquired fresh. Core action retention matters too, because a logged-in user who never transacts isn't retained in any way that earns money. The repeat-behavior side is covered in marketplace repeat usage, and the supply-side equivalent in supply retention.

Unit economics and contribution margin

The a16z essay lists unit economics as its own metric and notes that improved network effects often show up in improved unit economics over time. For a marketplace, the useful unit is usually one transaction: net revenue from the take rate, minus the variable costs that come with it (payment processing, support, insurance or guarantee costs, refunds, incentives). What's left is contribution margin per transaction.

Two cautions. First, subsidies to either side belong in this calculation, not above it. A marketplace that pays drivers or sellers bonuses to keep supply present may show a healthy take rate and a negative contribution margin. Second, contribution margin per transaction is only half the picture. Multiply it by how many transactions a user makes, and compare the result with what it cost to acquire that user. The general concepts are in unit economics and gross margin.

CAC per side

A marketplace has two acquisition problems, and they rarely cost the same. Supply might be recruited manually or through partnerships, while demand might come from search and paid channels. Blend them and you'll hide the expensive side. Track customer acquisition cost separately for buyers and sellers, and compare each against the contribution it generates. A cheap buyer is only cheap if there's supply for them to buy from.

Concentration and multi-tenanting

The a16z essay treats concentration or fragmentation as a metric in its own right and argues that marketplaces with greater fragmentation on the supply and demand sides are more valuable. The reason is bargaining power. If a handful of sellers account for most of GMV, they can leave, negotiate fees down, or take the relationship off-platform. A simple health check is the share of GMV from the top 10 sellers and the top 10 buyers, tracked over time.

The essay also lists the prevalence of multi-tenanting (users active on competing platforms at once) and switching or multi-homing costs. High multi-tenanting means your users compare you constantly.

Vanity metrics and health metrics

Some numbers feel good and say little. Others feel dull and predict everything. A rough way to sort them is to ask whether the number would change if the marketplace were quietly getting worse.

The pattern is the same each time. A volume number is paired with an outcome number, so you can tell whether growth in the first is real. GMV that rises while repeat rate and contribution margin fall is a marketplace buying its numbers.

One more trap sits in GMV itself: leakage. If buyers and sellers meet on the platform and then deal directly, that value never appears in GMV, and your real liquidity is higher than the measured one. The measurement gap shows up as falling repeat rates among relationships that look healthy elsewhere.

How the stack changes by stage

Nobody needs thirteen metrics on day one. What matters shifts with the stage, and trying to optimize a late-stage metric too early is a common way to waste effort. The sequence below is reasoning from how the metrics relate, not a benchmark from any study.

Seeding (before there's a real market). The question is whether any match happens at all. Watch match rate and time to match in your smallest viable segment, and count raw transactions, not growth rates. GMV is tiny and take rate is rarely the question yet.

Early traction. Liquidity needs to hold in one market or category before you expand. Add repeat rate and cohort retention, because they tell you whether early users found the matches valuable. Add per-side CAC so you know which side is expensive to recruit.

Scaling. Now GMV growth, take rate and contribution margin matter together. This is where unit economics get serious, because incentives that were cheap at small scale may not stay cheap. Segment everything by market and category, and watch concentration as your best sellers grow.

Maturity. Growth often shifts to monetization: raising take rate, selling advertising or value-added services, and defending against multi-tenanting. The metrics here are net revenue per transaction, revenue mix, and cohort GMV retention by vintage. Mature marketplaces also expand sideways into new categories or regions, which restarts the early-stage questions locally (see expanding a marketplace to new cities and categories).

A useful discipline: whatever the stage, report each metric next to the one that would expose it. GMV beside contribution margin. Match rate beside time to match. Take rate beside repeat rate. If the pair diverges, you've found the conversation worth having.

A starter dashboard

If you had to pick a short list, this is one defensible version. It covers volume, monetization, matching, stickiness and cost, with each metric paired with a counterweight.

1. GMV (with the definition written out) and net revenue.
2. Take rate, by category.
3. One buyer-side liquidity metric (search-to-fill or match rate) and one supply-side metric (sell-through or utilization).
4. Time to match.
5. Cohort GMV retention, by acquisition month.
6. Contribution margin per transaction, after incentives.
7. CAC for buyers and CAC for sellers, separately.
8. Concentration: share of GMV from top sellers and top buyers.

Every one of these should be cut by market, category and cohort before it's averaged. For how these sit inside a wider growth measurement approach, see the growth metrics hierarchy.

Key Facts: Marketplace Metrics

• The core identity is GMV × take rate = net revenue. GMV and revenue can move in opposite directions: in Etsy's FY2025 10-K, GMS fell 5.3% while revenue rose 2.7%, as the revenue take rate rose from 22.3% to 24.2%.
• GMV definitions differ by company. Etsy's GMS excludes shipping and is net of refunds. Uber's Gross Bookings include taxes, tolls and fees and make no adjustment for discounts or refunds.
• The a16z "13 Metrics for Marketplace Companies" (February 2020) covers match rate, market depth, time to match, concentration, take rate, unit economics, multi-tenanting, switching costs, and four kinds of retention analysis.
• Cohort comparison is the test of network health: newer cohorts should retain at least as well as older ones.
• Pair every volume metric with an outcome metric: GMV with contribution margin, listings with sell-through, users with cohort retention.
• Track acquisition cost per side. Buyers and sellers rarely cost the same to win.

Related Reading

• What Is Marketplace Liquidity?
• Marketplace Take Rate
• Marketplace Repeat Usage
• Why Density Beats Reach in Local Marketplaces
• Unit Economics
• Growth Metrics Hierarchy