Unit Economics: GMV, Take Rate, CAC, LTV, and What They Actually Mean | Marketplace Dynamics Tutorial

The Core Set of Metrics

A marketplace lives or dies on a short list of numbers. Here they are, with rough definitions:

GMV            Gross Merchandise Value. Total value of transactions in a period
Net revenue    What the platform actually kept: GMV × take rate
Gross margin   Net revenue minus COGS (payment processing, refunds, some fraud loss)
Contribution   Revenue per user minus the variable cost of serving that user
CAC            Customer Acquisition Cost, usually per side
LTV            Lifetime Value, usually per side
Payback        Months of contribution needed to recover CAC
Retention      Percentage of a cohort still active N weeks or months after acquisition

Every one of these has sub-variants. Every one can be gamed. Understanding what each actually measures (and fails to measure) is most of the skill.

GMV: The Headline Number

GMV is easy to understand and easy to over-weight.

GMV = sum of gross transaction values through the platform

An Uber ride is GMV. An Airbnb booking is GMV. A £10 Etsy sale is GMV. Hot coffee? No, because Starbucks isn't a marketplace.

GMV grows when:

Users transact more (real growth)
Transactions get bigger (can be good, can signal customers migrating up-market)
More users show up (good, if they transact)

GMV can also grow while the underlying business weakens:

Discounting drives more transactions but lower margin per transaction
The platform starts transacting lower-quality items, which pad GMV but damage reputation
The platform enters adjacent categories that dilute focus

A marketplace that only reports GMV is flattering itself. A marketplace that reports GMV and per-user transaction frequency and retention is being honest.

Take Rate, Revisited

Covered in chapter 5. Here in the context of financials:

net revenue = GMV × take rate

If GMV is £100M and take rate is 15%, net revenue is £15M. That is the real top-line.

Watch for:

Blended take rate vs category take rate: a marketplace's blended rate might be 15% but vary from 5% in a commodity category to 30% in a premium one
Effective take rate vs stated take rate: stated rate is the sticker number; effective rate includes discounts, promotions, and subsidies. Eg Uber's 25% stated rate was closer to 20% effective after promos

Annual reports of public marketplace companies are instructive here.

Gross Margin

gross margin = net revenue - COGS

What counts as COGS?

Payment processing fees: typically 2 to 3% of GMV
Refunds and chargebacks: 1 to 3% of GMV depending on category
Direct fraud losses: a cost of doing business; higher in high-trust categories
Insurance claims paid out (Airbnb AirCover, Uber's liability)
Customer support costs: arguably COGS for a transaction; reasonable to include

A marketplace with high gross margin (70%+) has room for growth spending. A marketplace with low gross margin (30% or less) is squeezed; every point of growth costs a lot of what it brings in.

Contribution Margin

contribution = revenue per transaction - variable cost per transaction

Variable costs include everything that scales with volume: payment fees, fraud, ops per-transaction cost, customer support per dispute, etc.

Contribution margin tells you what each transaction actually gives back to the business. Fixed costs (engineering, rent, brand) come out of total contribution.

A marketplace can have a great contribution margin and still be unprofitable if fixed costs are too high. It can also have a thin contribution margin and be profitable once volume is enormous. Either way, contribution margin is the per-transaction health check.

Customer Acquisition Cost (CAC)

CAC = marketing spend / new users acquired

Simple formula, fraught application. Fights over CAC:

Per side: marketplaces report CAC separately for supply and demand. A £200 CAC for a driver (Uber) is different from a £5 CAC for a rider (Uber)
Fully loaded vs paid: "paid CAC" is just ad spend; "fully loaded" includes BD, referral programs, promotions, sales salaries. Fully loaded is honest
New vs reactivated: a returning dormant user is cheaper to acquire than a brand-new one, but platforms sometimes count both as "new" to flatter the number

CAC trends reveal a lot. If CAC is falling year-over-year, organic acquisition is improving. If CAC is rising, either the category is saturating or the platform is out-competing its own efficiency.

Lifetime Value (LTV)

LTV = contribution margin per period × expected periods active

Or, more informally: "how much money will we make from this user before they churn?"

Fiddly to calculate. Requires:

Good retention curves (so you can estimate expected periods)
A view on contribution margin that doesn't change too much over time
Discount rate for future revenue

Most early marketplaces estimate LTV with big error bars. That is fine; LTV is a tool for reasoning about CAC, not a precise accounting number.

LTV per side

Both sides have LTV. A driver's LTV is the platform's share of every ride that driver gives over their active life. A rider's LTV is the platform's share of every ride the rider takes.

A well-balanced marketplace values both sides in its unit economics. A marketplace obsessed with only one side tends to under-invest in the other.

Payback Period

payback = CAC / (contribution per period)

If you spend £100 to acquire a user and they deliver £10/month of contribution, payback is 10 months.

Payback matters because it's a cash-flow metric. A 24-month payback with a 3-year LTV is mathematically profitable but requires a lot of capital to carry the users between month 0 and month 24.

Typical targets:

<6 months: exceptional; rare, usually indicates viral acquisition
6 to 18 months: healthy, especially for high-LTV marketplaces
18 to 36 months: workable if you can fund it and retention is strong
>36 months: you're running on hope and investor patience

Cohort Retention

A cohort is the set of users who first did something (signed up, bought, listed) in the same period. Track the cohort over time.

Retention curves take three common shapes:

Smiling (good): cohorts retain well, and later cohorts retain better than earlier ones. The platform is compounding
Frowning (bad): users leave fast, and the drop accelerates. Often a symptom of bad matching, bad trust, or bad early experience
Flatlining (neutral-to-bad): after an initial drop, retention stabilises. Workable if the initial drop isn't catastrophic

Marketplaces lean hard on cohort retention because the cross-side loop depends on users staying. A cohort that churns in month 2 took the platform's acquisition spend without contributing much.

The "three curves" practice

Look at three cohort curves:

New supply: how quickly do new sellers/drivers earn a first transaction? Good ones do so in days; bad ones in weeks
New demand: how quickly do new buyers/riders transact? Same question, other side
Re-activation: how many dormant users come back? A marketplace with strong re-activation has better fundamentals than one that only grows by acquiring new users

Put these on one chart per month. Healthy trajectories look like all three improving over time. Declining marketplaces see at least one of them degrading, often first.

Where Metrics Mislead

GMV includes cancelled/refunded transactions if you're not careful. Gross GMV ≠ net bookings
CAC excludes the cost of the free stuff. First-ride-free is a subsidy, not just a marketing cost. It should show up somewhere
LTV extrapolates. You observe 6 months of data and project 3 years of lifetime. The error bar is huge; people treat the number as precise
Average obscures distribution. Average user spend is £20 per year; the median is £4, because a small fraction of users spend thousands. Decisions based on "average" miss the mass
Blended metrics hide failing markets. Cities X and Y are great; city Z is a disaster. Blended CAC looks fine. Fix the city Z problem by reporting by city
Retention curves that exclude churned users. "Retention of active users" is not the same as "retention of the cohort". Read carefully

A Rough Marketplace Health Check

Can you answer these for your marketplace?

What is GMV by quarter for the last 8 quarters?
What is take rate by category? Is it rising or falling?
What is contribution margin on a transaction today?
What is CAC for each side, paid and fully loaded?
What is average LTV for each side, with your confidence interval?
What does the retention curve look like for the last four cohorts?
How does any of the above differ between your best market and your worst?

If you can't, the marketplace is not being managed, it's being stared at. If you can, you have the diagnostic surface to fix what's broken.

Common Pitfalls

"We grew GMV 3x this year." What happened to take rate? To contribution margin? To retention? GMV growth without the others is smoke

"Our LTV/CAC is 4:1." Over what horizon? With what assumptions about retention? With what inclusion rules on CAC? 4:1 reported at an investor pitch and 4:1 calculated for internal planning are often very different numbers

"We'll fix unit economics with scale." Sometimes true. Often cope. If contribution margin is negative, scaling just burns more money. The curve has to bend at some real volume, and you should be able to show why

"We have great cohort retention." Of which cohorts? First cohorts are often your most enthusiastic users; later cohorts may behave differently. Show the trend

"CAC is going down." Good. Is the new acquisition channel making up the volume difference? Is organic share growing? Context matters

Next Steps

Continue to 09-supply-demand-imbalance.md for the three states a marketplace can be in and how to get out of each.