Send the same brief to five marketplace developers and you can get numbers ranging from three thousand to ninety thousand. That spread is not a sign that four of them are wrong. It is a sign that the question has more variables in it than anyone tells you. Here is what actually sits behind the numbers.
Why the Quotes Are So Far Apart
I have watched a founder receive a three thousand dollar quote and a ninety thousand dollar quote for a brief she considered identical, from two firms she respected. She assumed one of them was trying it on. Neither was.
“Build me a marketplace” is a category rather than a specification. Each of those developers read the brief, pictured a product, and priced the one they pictured. One imagined a configured template with a logo on it. The other imagined split payments, a booking calendar, an admin console and a dispute flow. Both were reasonable readings of the same paragraph.
The market really is that wide, too, and you can see it publicly. Individual developers on the Sharetribe Experts directory list rates from fifteen to twenty-eight dollars an hour, with no-code setups from around two hundred dollars. North American boutique agencies list a hundred an hour with MVPs from five thousand. Offshore agencies advertise custom builds from around three thousand. Industry research puts MVP development between thirty and fifty thousand and full builds past a hundred.
Every one of those is a real price for a real thing. They are describing different products and, more usefully, different amounts of risk transfer. The two hundred dollar setup leaves all the risk with you. The ninety thousand dollar build moves most of it to the agency, and that transfer is a genuine part of what you are buying.
The Four Approaches
No-code setup. Two hundred to five thousand.Configure an existing platform, connect Stripe, go live. Sharetribe, Prometora at around ninety-nine a month, Kreezalid from a couple of hundred euros, Arcadier, Softr. You get a working marketplace running on someone else's transaction logic. You do not get to change that logic later.
More founders should use this than do. If you have not yet proven both sides will show up, this is a cheap way to find out.
Platform build with customisation. Five to thirty thousand. Start on a marketplace platform and extend it. A custom transaction flow, a different payout structure, a booking calendar that matches how your category actually behaves. Most funded and semi-funded founders land here.
What drives the cost is not the length of your feature list. It is the distance between your transaction and the platform's default. A rental marketplace with a standard booking flow is inexpensive. The same platform handling staged escrow, multi-party payouts and a bidding process gets expensive quickly, because the work is spent going against the tool rather than with it.
Custom build. Fifty thousand and up. Everything from scratch. The right answer when your transaction genuinely has no analogue, when regulation or data residency rules out hosted platforms, or when you have scaled past what a platform can carry.
AI-assisted build. The one that is still settling. Lovable, v0 and Cursor have made producing a working interface dramatically faster, and we use them every day. What they are very good at is getting to something that looks finished. What they are not yet dependable at is money movement, permissions, dispute states and the edge cases that decide how your marketplace behaves on a bad day. The scaffolding got cheap. The judgment did not.
If You Already Built One the Traditional Way
Worth saying plainly, because a fair number of people reading this have a completed build behind them and a number in their head they are not enjoying looking at.
That was the right decision. Until recently, a marketplace with any real transaction complexity needed a team of engineers for several months, and the going rate reflected genuine effort by people doing skilled work. Nobody overpaid by choosing that path. The cost structure shifted underneath the whole category in the last couple of years, and it shifted forward, not backward. Your build was priced correctly for the year you commissioned it.
What is worth revisiting is not the money already spent. It is whether the platform you have still fits the business you have turned out to be, which is a completely different question and usually has a much smaller answer than “rebuild everything.”
“The tools got cheap and the decisions did not. That is a change in what building costs, not a verdict on anyone who built before it happened.”Darren Cody, Marketplace Studio
The Costs That Are Not in the Quote
None of these are anyone's fault. They are simply not development line items, so they do not appear on a development quote.
Scope that turns out not to be needed. Marketplace briefs tend to grow in the writing, because a founder specifying a product they have not run yet has to guess, and guessing generously feels safer than guessing tightly. In our own engagements, a two to three week pre-development sprint usually removes twenty to forty percent of the planned scope before development starts. Not because the original brief was careless, but because a prototype in front of five real users answers questions that no amount of thinking answers.
The integration tax. Payments, identity verification, messaging, search, mapping, notifications, tax handling. Each looks like one line and each carries setup, testing and edge-case work past the initial connection. Payment processing alone runs roughly two to three and a half percent plus a fixed fee per transaction, which lands closer to two to five percent of gross merchandise value once payouts and gateway costs are included. That is an ongoing cost against your take rate.
The running cost. Hosting, subscriptions, third-party tools, monitoring, and someone available when something breaks. Somewhere between five hundred and three thousand a month depending on approach and scale, from launch, whether or not transactions are happening yet.
Trust and safety. Refunds, evidence trails, dispute policy. Almost nobody builds this at MVP because it is hard to picture needing it. Adding it after the first real dispute costs more than a simple version would have.
Supply acquisition. Not a development cost at all, and the most commonly missing line in the plan. Your first twenty listings get sourced by hand, by you, over weeks.
A Realistic Total for a First Marketplace
What we actually see, all in, from idea to first real transactions.
| Approach | Build | First 12 months running | All-in year one |
|---|---|---|---|
| No-code, self-configured | $200 to $5,000 | $1,200 to $6,000 | $2,000 to $11,000 |
| Platform build with customisation | $5,000 to $30,000 | $6,000 to $18,000 | $11,000 to $48,000 |
| Managed strategy and build | $15,000 to $45,000 | $9,000 to $24,000 | $24,000 to $69,000 |
| Fully custom | $50,000 to $150,000+ | $18,000 to $60,000 | $68,000 to $210,000+ |
Two things stand out. Running costs are a meaningful fraction of build costs in every tier and rarely make it into the plan. And the difference between tiers is less about quality than about how much of the thinking comes included, and how cheap it is to change direction afterwards.
What Actually Determines Your Number
Four things, roughly in order of impact.
How far your transaction sits from a single payment to a single provider. That is the cheapest thing in marketplace software. Deposits, split payouts, staged payments, conditional refunds, subscriptions layered on transactions, quotes and negotiation, multi-vendor carts. Each one adds real work. If you are in B2B, rental or services, those pages set out exactly which of these your category drags in.
How many sides and roles exist. Two-sided is standard. An agency tier, admins acting for suppliers, or a third party approving transactions multiplies permission logic and testing well beyond what the feature list suggests.
How specific your wedge is.Being able to say who the first hundred suppliers are, in which city, doing what, keeps scope still. “We will start broad and narrow later” is what expanding scope sounds like before it happens.
Whether demand is validated. This one decides whether the rest of the budget was worth spending, and it is the cheapest thing on the list to check.
The Question Underneath the Question
Most people asking what a marketplace costs are asking something slightly different: how much do I need to risk to find out whether this works.
That version has a better answer. The cheapest route to knowing is not the cheapest build. It is a couple of weeks of validation, a clickable prototype, five real conversations with people who would actually supply, and an honest look at the unit economics. That work costs a fraction of any tier above, and it either saves the build budget or tells you exactly where to point it.
We publish our own starting price of fifteen thousand because a number on a page saves everyone a call that was never going to go anywhere. If that is well outside your range, there are good options above that are not us, and we are happy to point at them. What we would suggest to anyone, at any budget, is spending a few weeks on what you are building before spending anything on building it.
Working out what a marketplace should cost is most of what happens in pre-development, and the number usually comes down during it rather than up. If you already know what you are building, Build and Launch runs eight to twelve weeks from fifteen thousand. And if you have a platform already and are wondering whether it still fits, that is a go-to-market conversation rather than a rebuild one.
