It looks like a catalogue with more sellers.

That’s the reading that gets marketplaces quoted like e-commerce. Underneath the product cards, though, there’s money that isn’t yours, commission rules that change how sellers behave, and selling mechanisms a catalogue doesn’t have.

It’s also the reading that makes them collapse halfway through.

The two sides

Two users on the same platform, who want opposite things.

Whoever sells

Wants to be found first, pay lower commissions, and talk to the customer directly. Everything you give them takes something away from the other side, or from you.

Whoever buys

Wants choice, price, and the certainty that if something goes wrong, someone answers for it. They’ll pay the commission only as long as that certainty is worth the price.

A marketplace isn’t software with two account types: it’s a balance held up by rules, and the rules are code.

Where it gets complicated

Six things a catalogue doesn’t have to do.

The money

It isn’t yours, it passes through you

Tracking what comes in, who it goes to, when, and how much you keep. An e-commerce site just takes payment: here money moves between third parties and you’re responsible for the handoff. It’s the part where a mistake doesn’t stay hidden — it shows up on a statement.

The rules

Commissions change behaviour

How much you keep, on what, and with which exceptions isn’t a percentage: it’s the lever that decides who stays and who leaves. Changing it means recalculating things that already happened, and that has to be planned for in advance.

The mechanism

Selling isn’t always «buy now»

Auctions with different formulas, order books, offers with a deadline, prices that move. Every mechanism has its own rules of fairness and its own edge cases, and the edge cases show up once there are a lot of users at once.

The people

Two populations, two journeys, one system

Sellers and buyers have different sign-ups, different permissions, different things they can and can’t see. And sometimes the same person is both.

What happens after

Shipping, refunds, returns

The least interesting part is the one that generates the most support tickets. It has to integrate with whoever actually carries it out, and it has to hold up when something doesn’t arrive.

Keeping order

Moderating what users upload

Content to check, behaviour to recognise, and attempts to take the conversation off the platform — which is the normal way a marketplace loses its margin.

Where we’ve done it

Sailsquare — 5 years on the platform, alongside their team.

Before

A platform where people organizing boat trips meet the people booking them, with strong seasonality: months where nothing happens, and weeks where everything happens at once.

After

5 years inside the product, alongside their CTO: a faster platform, automated internal processes and a system ready to grow.

Memorabid — an auction platform that holds up at the peak.

Before

Online auctions and charity evenings with hundreds of people bidding together. Twenty-seven different auction mechanisms, each with its own rules.

After

Infrastructure rebuilt and software deeply optimised: the platform holds up under the evening's peak. Two years without the founder having to think about it.

Why it matters

The low quote is the first symptom.

Whoever quoted a catalogue finds out halfway through that they’re building something else. And by then it’s rarely renegotiated: it gets cut. Infrastructure that holds up test traffic but not the real thing, edge cases left out, moderation pushed to after launch.

None of these cuts show up on day one. They all show up together on the day the platform is actually needed — auction night, launch week, the moment both sides are online at the same time.

If a marketplace quote looks low to you, it probably is.

How it starts

It depends where you are. If the marketplace doesn’t exist yet, you start by understanding what it actually costs to build — Build Plan. If it exists and it’s causing problems, the door is Rescue.

If instead you already know what needs doing and want to see how we work before committing, there’s the Dry Run: two weeks, and it ends there.

Astrorei · When software matters

When software becomes
a business matter.