Building a marketplace: the problem is liquidity, not software

A marketplace has a chicken-and-egg problem that no amount of engineering solves. Buyers do not come to an empty catalogue. Sellers do not list where there are no buyers. Every marketplace that worked has a specific answer to which side it filled first and how.

The software is the easy half

Listings, categories, structured attributes, search, filtering, seller profiles, messaging, moderation, media. It is real work and it is well understood. If a vendor tells you the build is the hard part, they have not launched one.

The supply plan is the business

You have three options and only three.

Buy it. Pay for listings or pay a sales team. Predictable, expensive, and it stops when you stop.

Import it. Find where supply already exists publicly and bring it in with permission and attribution. In this region, that is Telegram, WhatsApp and Facebook groups, which is where trade actually happens today.

Own it. Be the supply yourself at the start. Slow, but the quality is yours to control.

We used the second for our own platforms — pipelines that read public listings, extract structure from unstructured Arabic, deduplicate, and publish. It is the difference between launching with a catalogue and launching with an empty page.

Narrow beats broad

A general marketplace competes with everyone. A vertical one — cars, property, jobs — can model its category properly. Cars need forty structured attributes; jobs need none of them but need applications. Depth is what makes search useful, and useful search is what brings buyers back.

Measure liquidity, not traffic

The number that matters is the share of listings that receive a genuine enquiry. Traffic with no liquidity is a vanity metric, and it is the one that hides a dying marketplace for six months.

We have built five. See how.