The honest definition of "MVP" for a marketplace
A marketplace is two businesses in a trenchcoat: supply (sellers, providers, listers) and demand (buyers, seekers, bookers), and each half needs enough of the other to care. That interdependency is why marketplace MVPs fail on scope before they fail on market: the "minimum" that still functions is smaller than founders expect and more deliberate than a feature list. The honest MVP has exactly one transaction type, one vertical, and one side curated by hand. Everything else, open signups, reviews, messaging, mobile apps, multi-city, is version two or later, and the architecture should know that even if the launch does not show it. The service side of this lives at marketplace development; this page is the decision manual.
The cold-start problem is a business problem wearing engineering clothes
No code solves cold start. The playbook that works is sequence, not symmetry: curate supply by hand first (dozens, not thousands, chosen for the demand you can already name), then seed demand with the one channel you already have, then automate what manual effort proves people want. Marketplaces that open with open signups on both sides launch into a ghost town and read the silence as "no market", when the actual finding was "no density". Density first, openness later, and the MVP is scoped to make manual curation cheap, which is a genuinely different build.
Split payments: the part everyone underestimates
The moment money routes from buyer to seller through you, the build changes class. Split payments (buyer pays, platform takes a cut, seller gets the rest) run on managed infrastructure, Stripe Connect-class products exist precisely so you do not become a payment facilitator, but they carry onboarding (KYC per seller), payout schedules, dispute flows, and tax plumbing that flat commerce never meets. The regulatory edges (when your entity structure triggers facilitator status) are counsel questions; the engineering job is architecting so that conversation is boring. This is the single most common under-scope line in marketplace quotes, if a quote prices your marketplace like a shop with extra logins, the payouts are the fantasy in it.
What the MVP actually contains
- One transaction type, end to end, booked, paid, fulfilled, disputed. Depth beats breadth; the second transaction type is v2.
- Curated supply onboarding, an admin panel for you, not a self-serve funnel. Your team is the moderation engine at launch.
- Demand-side search that works, findability is the product for buyers; filters beat cleverness.
- Trust surfaces, minimum viable, real profiles, honest photos, a dispute path with a human on your end. Full review systems wait.
- Payments with real payouts, split infrastructure wired, tested with small real transactions, payouts verified to a real bank account.
- Analytics from day one, marketplace metrics are two-sided (supply utilization, demand conversion) and the dashboard exists before launch, not after the first argument about what is working.
Build on a platform or go custom?
Marketplace-in-a-box platforms (Sharetribe-class) and WordPress+plugin stacks cover standard shapes, products, bookings, simple commissions, at a fraction of custom cost, and for validating demand they are often the honest answer. Custom earns its cost when the marketplace's mechanics are the differentiation: unusual matching, specialized workflows, niche trust models the box cannot express. The brutal middle path, start on the platform, rebuild custom after traction, is a real strategy with a real cost (migration debt), and the audit weighs it per mechanics rather than by fashion. Context for either: what builds cost and the cost calculator.
The staging plan that keeps this affordable
Version one (the MVP): curated supply, one transaction type, manual everything-you-can. Version two: automate what manual work proved demand for, self-serve onboarding, reviews, messaging. Version three: the widening, more verticals, more transaction types, the app. Each stage funds the next or kills the project cheaply. The failure pattern is launching as version three on version-one evidence; the honest pattern is letting usage, not ambition, schedule the roadmap. The marketplace development page shows how the staging gets scoped and priced; a brief with your vertical and the transaction type starts the conversation.