Why Standard LTV and CAC Formulas Break for Marketplaces

Every week I see marketplace founders plug their numbers into the same SaaS formulas. Monthly revenue times gross margin divided by churn. Customer acquisition cost from total spend divided by new customers acquired. The LTV:CAC ratio comes out at 4x or 5x. The deck looks clean. The business runs out of cash eighteen months later. I've lived these dynamics from the inside — I co-founded PizzaPortal, a food delivery marketplace we sold to Delivery Hero — and much of what follows comes from that experience.

The formulas aren't wrong. They're wrong for marketplaces.

A SaaS company has one customer type. A marketplace has two. A SaaS company's revenue is what the customer pays. A marketplace's revenue is a fraction of what the customer pays. A SaaS company's retention curve applies to a single user. A marketplace's retention depends on both sides showing up.

These aren't minor adjustments. They change every number in the model.

You Have Two Businesses, Not One

Uber has riders and drivers. Airbnb has guests and hosts. Etsy has buyers and sellers. DoorDash has eaters and restaurants (and dashers — actually three sides, which makes it worse).

Each side has its own:

When a founder tells me their CAC is $35, my first question is: which side? Because a marketplace that spends $35 to acquire a rider and $350 to onboard a driver has a blended CAC that means nothing. It obscures the actual economics of both sides.

I've reviewed hundreds of pitch decks where the founder presents a single LTV:CAC ratio for a two-sided marketplace. That number is useless. Worse than useless — it's actively misleading, because it hides the side that's bleeding money.

Airbnb's cost to acquire a guest through paid channels is completely different from the cost to onboard and verify a new host. The host requires identity verification, listing photography (Airbnb used to send professional photographers for free), and often months of hand-holding before the first booking. The guest needs a Facebook ad and a $20 coupon.

If you blend those two CACs, you get a number that describes neither reality.

As I cover in The Two Numbers That Build or Break Every Business and in my essays for marketplace founders, the power of LTV and CAC comes from their specificity. The moment you blur the boundaries of who you're measuring, the numbers lose their diagnostic value. In marketplaces, that blurring happens by default unless you actively prevent it.

Revenue Is Not GMV

This is the single most common mistake in marketplace financial models.

In SaaS, when a customer pays you $100/month, your revenue is $100/month. Simple.

In a marketplace, when a customer pays $100, your revenue might be $15. The other $85 goes to the supply side. Your actual revenue is the take rate — the commission or fee you keep.

Founders who calculate LTV based on GMV (gross merchandise value) instead of net revenue inflate their unit economics by 5x to 15x. I've seen decks where the "LTV" is $2,400 and the actual lifetime revenue to the company is $240. That's not a rounding error. That's a different business.

Here's what take rates actually look like across major marketplaces:

An Etsy seller who generates $10,000 in GMV per year produces $650 in revenue for Etsy. A DoorDash customer who orders $5,000 in food per year generates $1,250-$1,500 for DoorDash.

The LTV formula for a marketplace must start with net revenue:

If you're using GMV in your LTV calculation, stop. Recalculate everything. Your unit economics are probably negative, and you need to know that before your investors figure it out.

Supply and Demand Die at Different Rates

Here's something that makes marketplace LTV genuinely hard: the two sides don't churn at the same rate.

An Airbnb host who lists a spare bedroom might stay on the platform for 5+ years. They've done the work of setting up the listing, getting reviews, optimizing their pricing. Switching costs are real. A guest, on the other hand, might book 2-3 times over two years and then disappear — they moved, they stopped traveling, they switched to Vrbo because it was $30 cheaper on their last search.

Uber drivers in most markets churn within 12-18 months. Riders in those same markets might stick around for 3-5 years. The supply side burns out. The demand side has low friction.

This asymmetry has big implications for how you allocate costs.

When supply lives much longer than demand, each supply-side user serves multiple generations of demand-side users. A host who stays five years might serve 200 different guests. The $500 you spent to onboard that host gets amortized across all of those transactions. Only a fraction of the supply CAC should be allocated to any single transaction cycle.

When supply churns faster than demand — like with Uber drivers — you're constantly re-acquiring the hardest, most expensive side. Your steady-state CAC never drops to zero because you're always refilling the supply pool. The business looks like it's growing, but the acquisition machine never turns off.

The formula most founders use assumes a single churn rate. Marketplaces need two:

These are linked — the revenue generated by one side depends on the other side existing — but the retention dynamics are independent. You need to model them separately.

Network Effects Are Supposed to Fix This

The promise of a marketplace is that network effects make everything better over time. More drivers mean shorter wait times mean more riders mean more drivers. The flywheel spins. LTV goes up. CAC goes down. Margins expand.

That's the theory.

In practice, network effects are often weaker than founders claim, slower to materialize than models project, and geographically constrained in ways that prevent them from scaling.

If network effects are genuinely working in your marketplace, you should see specific things in your cohort data:

If LTV is flat or declining despite top-line growth, your network effects aren't doing what you think they're doing. You're probably growing by spending more, not by becoming a better marketplace.

NFX, the venture firm that literally named itself after network effects, has published extensive research on this. They identify different types — direct, two-sided, data, platform — and most marketplaces only have weak two-sided effects that plateau at local saturation. Sarah Tavel's Hierarchy of Marketplaces framework at Benchmark breaks it down further: the first level is just getting transactions to happen (GMV growth), the second is making the experience good enough that users come back (retention), and the third is when the marketplace becomes the default — when participants actually can't leave because too much of their livelihood or behavior is embedded in the platform.

Most marketplaces never get past level one. They have transaction volume but no stickiness. Users are promiscuous. The marketplace is a commodity.

Andreessen Horowitz has written about this too — their marketplace framework emphasizes that network effects must be measured, not assumed. The test is simple: plot your LTV by cohort over time. If the lines are going up and to the right, network effects are real. If they're flat, you have a distribution channel, not a network effect. If they're declining, you have a subsidy machine.

Every User's Value Depends on the Other Side

This is the part that makes marketplace unit economics fundamentally different from any other business model.

In SaaS, a customer's LTV is determined by what they pay and how long they stay. Those two variables are mostly independent of other customers. One customer churning doesn't reduce the value of another customer.

In a marketplace, a rider with no drivers has zero LTV. A host with no guests generates nothing. A freelancer on Upwork with no clients produces zero take rate. The value of each user is a function of the other side's density and quality.

This means LTV is not a fixed attribute of a user. It changes — sometimes dramatically — based on local supply-demand balance.

Uber learned this in city after city. When they launched in a new market with heavy driver subsidies and rider coupons, both sides showed up. Remove the subsidies before reaching liquidity, and both sides evaporate. The individual user's LTV was contingent on the marketplace sustaining a certain density.

The leading indicator here is fill rate (or match rate). What percentage of demand-side requests get fulfilled? What percentage of supply-side availability gets utilized?

When fill rate is high (above 80-90% depending on the category), both sides see value. LTV stabilizes. When fill rate is low, demand churns because the marketplace doesn't work, and supply churns because there's no money to be made.

Your LTV model needs to account for this. A user acquired when fill rate is 40% has a completely different expected lifetime value than a user acquired when fill rate is 90%. If you're using a single LTV number across all cohorts regardless of marketplace maturity, you're lying to yourself.

As I discuss in The Two Numbers, the biggest trap in unit economics is treating them as static. They're dynamic — and in marketplaces, they're dynamic in ways that compound. A small drop in supply density causes a small drop in fill rate causes a moderate increase in demand churn causes a further drop in supply utilization causes more supply churn. The feedback loops work in both directions.

What Investors Actually Want to See

When a marketplace founder walks into a meeting at a16z, Benchmark, or any firm that's done marketplace investing, the partners aren't looking for a single LTV:CAC slide.

Andreessen Horowitz tracks 16 metrics for marketplace businesses. Bill Gurley at Benchmark has written about 10 factors he evaluates. The specifics vary, but the pattern is consistent. Investors want:

Bill Gurley's evaluation of marketplace quality goes deeper than metrics. He looks at fragmentation of the supply side (more fragmented = better for the marketplace), whether the marketplace improves the transaction experience or just aggregates it, and whether take rate is defensible or will get compressed by competition. These factors determine whether your LTV will hold up or erode.

The founders who get funded — and more importantly, the founders who build durable businesses — can answer a question like: "What's the 12-month LTV of a demand-side user acquired through paid Instagram in Q3, in your top 5 markets, net of all incentives and variable costs?" That level of granularity is what separates a real marketplace operator from someone who copied a SaaS metrics template and changed the title.

The Formulas You Actually Need

For supply side:

For demand side:

And the marketplace-level check:

This is more complex than the SaaS version. That's because the business is more complex. Pretending otherwise doesn't simplify things — it just delays the reckoning.

As I wrote in The Two Numbers, you don't get to choose whether your unit economics are complicated. You only get to choose whether you measure the complexity or ignore it.

The Chicken-and-Egg Makes Everything Harder

There's one more thing that breaks standard formulas: the cold start.

In a SaaS business, your first customer and your thousandth customer have roughly the same experience. The product works the same way. In a marketplace, your first customer on either side has a terrible experience because the other side barely exists.

This means your early cohorts have artificially bad unit economics. High churn, low transaction frequency, heavy incentive spending. If you include those early cohorts in your LTV calculation, the number looks terrible. If you exclude them, you're cherry-picking.

The honest approach: segment your LTV by marketplace maturity. Track cohorts separately for pre-liquidity markets and post-liquidity markets. Show both numbers. Investors respect the transparency, and you'll actually understand what your business looks like at scale versus what it looks like during the painful bootstrapping phase.

DoorDash did this well in their early markets. They knew that unit economics in a market with 50 restaurants and 500 active customers looked nothing like a market with 5,000 restaurants and 200,000 active customers. They invested in the former based on evidence from the latter. That's how you use marketplace unit economics — as a tool for capital allocation, not as a vanity metric for pitch decks.

The next two articles break down exactly how to calculate CAC and LTV for your marketplace, with formulas and worked examples.

Frequently Asked Questions

Related