How do you build a marketplace website? MVP scope, the chicken-and-egg problem, and payments

Below are three working tools: an MVP scope table, a cold-start checklist and a payment decision table. Every platform rule is linked to the vendor or regulator, as of September 30, 2026.

Short answer: five steps

A marketplace is two products glued together by a transaction. Buyers need to find something. Sellers need to get paid. The platform sits in the middle and holds the risk. Here is the order that keeps that risk small.

1. Choose the wedge. One category, one geography or community, one side to recruit first.
2. Scope the MVP around a single transaction. Listing, search, order, payment, payout, dispute. Everything else waits.
3. Seed supply manually. Onboard the first sellers yourself, on the phone if needed.
4. Wire payments through a platform provider. Stripe Connect, PayPal's multiparty solution or Adyen for Platforms. Do not hold other people's money in your own bank account.
5. Measure liquidity weekly. Track the share of listings that get an order within a set window. Grow the next category only when that number is healthy.

The rest of this article expands each step. Steps 1 and 3 decide whether the business works. Steps 2 and 4 decide how much the build costs.

Pick the wedge: niche, side, and liquidity metric

Thin inventory kills a young marketplace faster than any missing feature. A buyer who searches and finds three listings leaves. A seller who lists and gets no orders for a month leaves too. So the first decision is how small to start.

Andrew Chen, in his book The Cold Start Problem, calls the smallest self-sustaining group an "atomic network." In the chapter he published on andrewchen.com, he names sellers as the "hard side" of a new marketplace. They do more work and are harder to acquire and keep. His advice is to build for that side first, and the other side follows.

Use this checklist before any design work starts.

Cold-start checklist (9 points)

• [ ] Hard side named. Write down which side is scarce. For most goods and services marketplaces it is the seller.
• [ ] One category. "Wedding photographers," not "creative services."
• [ ] One city or one community. A local service marketplace needs density in one metro before it needs a second.
• [ ] Minimum inventory target. Decide how many active listings a buyer must see on the first search page. Do not open to buyers below that number.
• [ ] Manual supply plan. A named list of sellers you will contact personally, with a script and a follow-up date.
• [ ] Single-player value for sellers. A reason to join before buyers arrive: a free booking calendar, a clean portfolio page, invoicing.
• [ ] Demand channel for week one. One channel you control: a newsletter, a community, a partner's customer list.
• [ ] Liquidity metric defined. For example, the percentage of listings with at least one order within N days. Pick the window that fits your purchase cycle.
• [ ] Kill or expand rule. The metric value at which you add a second category, and the value at which you rethink the wedge.

The liquidity metric matters most. Traffic and sign-ups look good in a pitch deck. They say nothing about whether transactions happen. Track orders per listing from day one.

MVP scope: what to build first

A marketplace MVP is judged by one question. Can a stranger pay another stranger, and can both trust the result? Anything that does not serve that path goes to "later."

The table below sorts 17 common features. "Skip" means you do it by hand or with an off-the-shelf tool until volume forces a build.

Two rows cause most scope fights. Messaging tends to grow into a full chat product. Keep it attached to an order in v1. The mobile app is the other one. A responsive website covers both sides at launch, and an app makes sense once you know which side opens it daily.

For timing, the marketplace service page lists an MVP in 4–6 weeks. That scope covers onboarding, payments, listings, basic analytics and admin tools (as listed on toimi.pro on September 30, 2026). Anything beyond that list is estimated after a brief.

Budget follows scope. A tight v1 with a fixed list suits a fixed quote. A product still hunting for its wedge changes weekly, and that changes the contract math. Read the comparison of fixed price or time and materials before you sign either.

The chicken-and-egg problem

Buyers come for sellers. Sellers come for buyers. Nobody comes first unless you give them a reason. Three tactics are worth copying.

Build a tool for one side. Chris Dixon described this in 2015 as "come for the tool, stay for the network": attract users with a single-player tool, then pull them into the network. For a marketplace, the tool goes to sellers. A booking page, a quote builder, a simple storefront. They get value on day one even with zero buyers.

Solve a hard problem for the hard side. Chen's Tinder chapter makes the case that the product must work for the scarce side first. On a services marketplace that often means fast payouts and protection from no-shows. On a goods marketplace it means low listing effort.

Fake the density you do not have yet. Launch one category in one area and hide the empty ones. An empty state that says "12 photographers in Austin" beats a national search that returns nothing in Denver. Design those empty states on purpose.

What does not work is paid acquisition on both sides at once. You pay twice and still get a thin market. Spend on the side you named in the checklist and recruit the other side only when inventory passes your minimum.

Payments, payouts and seller verification

This is where a marketplace differs from a store. Money moves from a buyer to a seller through you. Someone must verify the seller, hold funds, handle refunds and report income to the IRS. Platform payment providers take most of that on, and each splits the work differently.

Stripe's documentation lists three charge types for Connect. Direct charges go straight to the seller's account. Destination charges and separate charges and transfers go to the platform first (Stripe, Charges in a Connect integration, as of September 30, 2026). Separate charges and transfers is the type for a shared cart with several sellers.

One terminology change matters for any spec. Stripe now labels Standard, Express and Custom as legacy account types. It recommends controller properties or the Accounts v2 API for new platforms (Stripe, Connected account types, as of September 30, 2026). A spec that still says "Express accounts" is out of date.

On taxes, the IRS threshold is back to a clear number. Payment apps and online marketplaces must file Form 1099-K when a payee's goods or services payments exceed $20,000 in more than 200 transactions (IRS, Understanding your Form 1099-K, as of September 30, 2026). Who files depends on setup. Stripe issues 1099-K forms only where Stripe controls pricing or the connected account pays Stripe's fees directly. In other setups the platform is responsible (Stripe, Tax reporting for Connect, as of September 30, 2026).

Payment decision table

Sources for the table: Stripe identity verification, PayPal multiparty overview, PayPal seller onboarding, PayPal multi-seller payments, Adyen for Platforms, marketplaces, all as of September 30, 2026.

Physical goods add one more law. Under the INFORM Consumers Act, online marketplaces must collect, verify and disclose data on "high-volume third party sellers." That means 200 or more sales of new or unused consumer products and $5,000 or more in revenue in a continuous 12-month period. Once a seller crosses that line, you have 10 days to collect bank, tax ID and contact details (FTC, Informing Businesses about the INFORM Consumers Act, as of September 30, 2026). This article is not legal or tax advice. Check both rules with your accountant before launch.

Trust: reviews, disputes, moderation

Strangers pay strangers only when the rules are visible. Three pieces belong in v1.

Reviews tied to real orders. Allow a review only after a completed transaction. That blocks most fake reviews without extra tooling. Do not hide negative reviews. The FTC's Consumer Reviews and Testimonials Rule took effect on October 21, 2024, and it covers review suppression (FTC, Q&A on the rule, as of September 30, 2026). Write your moderation criteria down and apply them to all reviews.

A written dispute path. List the dispute types: item not delivered, not as described, buyer changed their mind. For each, state the deadline, the evidence you need and who decides. Log each decision against the order. Without this, support improvises, and the rules drift.

A moderation queue. Every new listing passes an admin check at launch. It is slow. It is also cheap, and it teaches you what automated rules to write later.

Off-the-shelf marketplace software or custom

You do not always need a custom build on day one. Hosted marketplace software can test a wedge fast. Sharetribe, for example, lists plans from $99 to $299 per month billed yearly. Extra transactions beyond the plan quota cost $0.19 or less each (Sharetribe pricing, as listed on sharetribe.com on September 30, 2026). Payments run through Stripe.

That route suits a founder still testing the category. It gets harder when the transaction itself is your edge: custom booking logic, quotes with deposits, multi-party orders, ERP or inventory sync. Ask three questions before you choose.

• Does your core flow fit the tool's transaction model? If you fight it on day one, you will fight it every week.
• Who owns the payment accounts? If sellers onboard to Stripe under your platform account, moving them later is a project. Ask your provider how connected accounts migrate.
• Can you export users, listings and order history? Test the export before launch.

A custom build costs more up front and gives you control over flows, data and fees. The right order for many founders is to validate the wedge cheaply, then rebuild once liquidity is proven and the tool limits growth.