Marketplaces look simple from the outside, two sides, a transaction, a take rate. From the inside, they are one of the hardest products to design. Most fail at the shape stage, long before any code gets written.
01. Start with the matching, not the listing.
The moment you make supply scrollable, you're competing with Google. And you lose, because Google ships a marketplace every morning by accident. Design the request → curated match flow first; the listing comes after, as a side effect.
What this looks like in practice: a request form with the four or five questions only your marketplace can answer for, a 24-hour response window, and a small handful of curated suggestions, not an algorithmic feed.
"If your homepage is a search bar, you've already lost the differentiation argument."
On Dreilokale, the Swiss event-location matching platform Morvion built end-to-end, the homepage isn't a search bar. It's a single request the organiser fills in under two minutes. Tailored venue proposals come back inside 48 hours. The matching engine, CRM, CMS, and email automation sit behind it.
02. The operator dashboard ships at v0.1.
If you can't run the marketplace as a single human from a dashboard, you can't run it with code. The operator dashboard is the actual product on day one. Everything you eventually automate, you should be willing to do by hand first.
The questions to answer in the dashboard: which requests are open? Which suppliers have I notified? Which haven't responded? What do I need to nudge today?
03. Two dashboards, not one.
The requester / organizer dashboard and the supplier / operator dashboard are different products. They have different success metrics, different cadences, different vocabularies. Trying to unify them is a politeness that costs you both sides.
A single "account" page that tries to serve both roles. It always reads as a compromise. Build two.
04. Trust is a design problem.
Trust signals are the reason people click “request” instead of searching Google one more time. They have to land in the first viewport. Reviews are the obvious one; less obvious are response-rate badges, named-operator pages, and the small detail that you reply in under an hour.
- Show response rate next to every supplier card.
- Name the operator who curates the matches.
- Display the actual reply time, not a vague promise.
05. Lead credits beat percentage take rates (early).
Until you have real liquidity, charging a percentage of GMV creates the wrong incentive on both sides. The supplier hides the deal off your platform; the buyer feels they overpaid. Lead credits, small, predictable, paid by the side with budget, keep everyone aligned while you build trust.
Move to percentage take rate later, when both sides would lose something real by leaving.
06. Empty state ≠ failure.
Design for the “small, curated” state, not for the “1,000 venues, infinite scroll” state. Year one of every marketplace is the empty state. Make it look intentional: a short list, named suppliers, a personal note from the operator.
"A small marketplace that looks curated outperforms a large one that looks scraped."
What to skip.
You don't need: messaging at v1 (use email + reply-to magic), ratings before you have 50 transactions, AI-powered matching before you have humans matching, mobile apps before you have desktop conversion. Every one of those is a place we've seen a founder spend three months and regret it.
