Every marketplace brief eventually reaches the same question: which side do you launch first. Get it wrong and the product is technically finished and commercially empty.
Why this is an engineering question too
The answer changes what gets built first. If sellers are onboarded before buyers exist, the priority is a listing experience good enough that sellers stay engaged through an empty period — notifications, a sense of progress, low friction to add stock. If buyers arrive first, the priority is search and browse working convincingly on a thin catalogue, which usually means curation tooling for the team, not just the algorithm.
What we built for this brief
The client had committed sellers before launch, so we optimised for the seller side first: a listing flow that could be completed from a phone in under two minutes, and a dashboard showing views and enquiries even before there were sales to show. That dashboard existed purely to keep sellers checking in during the quiet weeks.
- Instrument the side you launch second more heavily than the side you launch first, because that's where you'll be diagnosing "why isn't this working" in month one.
- Fake liquidity carefully. We showed sellers realistic buyer search terms pulled from a waiting list, rather than any invented activity — the line between encouraging and dishonest is exactly where a made-up sale would sit.
- Decide the fee model before the fee code. Marketplaces revisit take rate constantly in year one; we built it as a configuration the client could change without a deploy.
The quiet failure mode
A marketplace can look successful on every dashboard — signups, listings, sessions — while the actual transaction rate stays flat. We agreed with the client upfront that transactions per active seller was the one number that mattered, so a good-looking vanity metric couldn't quietly stand in for it.
A marketplace with plenty of activity and no transactions is a content site with extra steps.
What we would scope differently next time
We'd push the take-rate and payout logic even earlier. It arrived in week six of a four-month build and every seller conversation before that point had to hedge on money, which cost us credibility we didn't need to spend.