Marketplace Validation
How to validate a marketplace idea before you build anything
Marketplace ideas are uniquely expensive to be wrong about, because you cannot half-build one. A normal product can launch small and grow. A marketplace has to convince two groups of strangers to show up at the same time, which is why marketplace validation matters more here than almost anywhere else in software, and why the generic market validation advice written for single-sided startups keeps failing founders in this category: it tests one audience, and you have two.
Skipping the step is tempting for an honest reason, too: the reliable methods all involve talking to people before you have anything to show them.
Here are the three validation methods that actually tell you something, in the order I would run them. I have used all three, and I have also skipped all three at one point or another, which is how I know what each skipped step costs.
First, What Is Not Validation
Four signals founders mistake for proof
The "no competitor" argument
Finding no direct competitors is not validation. It might mean the market does not exist, not that you are first.
Survey enthusiasm
People are generous in surveys and stingy with behaviour. Stated intent to use a marketplace that does not exist yet costs the respondent nothing.
Waitlist signups
An email address is the cheapest commitment there is. A waitlist tells you a headline worked, not that a transaction will.
Single-sided excitement
Buyer enthusiasm is easy to find and nearly worthless alone. Supply is where marketplaces die.
Method One
Interview both sides, separately
Not surveys. Conversations. Five potential providers and five potential buyers, asked about their current behaviour, not your idea. What do providers do today to find customers, and what does it cost them? What do buyers do today to solve the problem, and where does it fail them?
The trap this protects you from is single-sided validation. Buyer enthusiasm is easy to find and nearly worthless on its own, because buyers are agreeing to a better option that costs them nothing to praise. Supply is where marketplaces die. If providers have an existing channel that works, your marketplace is asking them to add a take rate to a problem they have already solved, and no amount of buyer excitement fixes that. That is the same asymmetry behind the cold start problem.
What this method cannot tell you: whether anyone will actually complete the loop. People are generous in interviews and stingy with behaviour. Which is why the next two methods exist.
Method Two
Run the loop manually, the concierge test
Before building anything, be the marketplace yourself. Take requests from buyers over a form or a group chat, match them to providers by hand, and broker a few real transactions with a spreadsheet as your platform.
It is unglamorous and it is the single highest-signal test available, because real money changed hands and you learned the take rate conversation firsthand. If you cannot broker five transactions manually, software was not the missing piece.
What it cannot tell you: whether the product experience works, whether people transact without you personally greasing every step, and whether the model holds at any scale beyond your own hours.
Method Three
Put a working prototype in front of both sides
This is the step that used to be a six-figure decision and is now a days-long one, which changes where it sits in the sequence. A working prototype of your core loop, seeded with realistic data, shown to the same providers and buyers you interviewed, tells you the thing the first two methods cannot: does the loop make sense to strangers without you in the room narrating.
I wrote a full guide to building a marketplace prototype with AI tooling if you want to do it yourself; the spec-writing discipline matters more than the tool. And if you would rather bring the question to someone who has built marketplace products since 2016, that is the service I run: a working prototype that answers your riskiest question, with the scope boundaries stated up front.
What a prototype cannot tell you: whether people will pay at scale, whether acquisition costs work, whether liquidity holds. Those answers only exist in a real market. The prototype's job is to make sure the expensive attempt at real answers is pointed at something coherent.
The Honest Sequence
Interviews, concierge, prototype
Interviews first, because they are free and they kill the ideas that deserve to die early. The concierge test second, because real transactions beat stated intentions. The prototype third, when the question left standing is about the product loop itself.
And a note on what validation is for, because the word gets used as if the goal were a green light. The goal is a decision. "This idea, revised this way, is worth a real build" and "this idea is not it, and I found out for the price of some conversations and a prototype" are both successful validations. The failure mode is not a dead idea. It is a live one that never got tested and ate a year.