Why Build a Marketplace, and Which Metrics Prove It Worked?

A project with no definition of success has no condition for being finished. A marketplace without one fares worse: it gets "tested" for three years, never receives the full investment, and is then shut down as a failed experiment. It never got the conditions it needed to succeed.

The good news: there are only a handful of intentions, and after a single conversation you know which one is yours. The bad news: that same conversation almost always turns up two different ones running side by side inside the organization, and each of them leads to a different project.

This article breaks down:

• Which 6 reasons do companies build a marketplace for?
• When does a marketplace start paying back?
• Which 2 families of metrics do you need?
• What does GMV hide?

Key insights

• The reason you are building the marketplace settles the scope, the team, and the metric.
• Growth metrics tell you whether the marketplace is growing. Operational metrics tell you whether it works. GMV alone hides trouble the longest.
• Credible public payback benchmarks barely exist: they come from interested parties, they measure the survivors, and they leave the costs out.
• Set phase gates before launch, after the first quarter, and after the first year, each with a sentence saying what you do if the gate is missed.

Which 6 reasons do companies build a marketplace for?

On a slide, they all look the same: "we are launching a marketplace." In practice, each one builds a different system, needs a different team, and is measured by something different.

1. Expand the assortment without your own working capital. You want more to sell without buying stock into a warehouse.
2. Improve the margin. A commission on someone else's sale is cheaper to service than your own turnover. That holds once the volume reaches a level where it covers the cost of the team.
3. Collect demand data. You want to know what your customers are looking for before you invest in buying it yourself.
4. Defend yourself against a third-party platform. Your customers are starting to buy elsewhere, and you want to keep them with you.
5. Enter a new category or a new market. A test without the risk of stocking up.
6. Monetize the traffic you already have. Commission is one thing, but sellers also pay for visibility.

These intentions do not rule one another out. The problem is that they produce conflicting priorities, and a project only has so many hands.

How does your reason change the scope of the project?

If you want to expand the assortment, supply is the bottleneck. Recruiting sellers becomes the priority, along with shortening the path from signature to first offer. You invest in self-service, in file imports, and in integrations with the tools sellers already use. The metric you care about is the number of active sellers and the rate at which offers are added.

If you want to collect demand data, the bottleneck is analytics and what you do with it afterwards. You invest in reporting, in consistent definitions, and in a process that turns an observation into a buying decision. Recruitment can be slower as long as it is well targeted. One warning, though.

This intention carries a price in your relationships, and it is worth saying out loud. If you eventually intend to sell for yourself what you discovered through your sellers, you are competing with the people you just recruited. They will notice. Better to settle that deliberately and write it into policy than to discover it in year three, at the first conflict.

If you are defending yourself against a third-party platform, what counts is speed and breadth of offer rather than margin. A low commission at the start is an investment then.

That is why the question "why" is not a preamble to the project. It is the first project decision.

When does a marketplace start paying back?

This is the most important thing to say to the board at the outset rather than in year three.

Marketplace implementations at large organizations usually take a year or more, and contracts with platform vendors often provide that the fee starts running from signature rather than from the launch of sales. That means you spend several quarters paying for a system that earns nothing.

The industry calls this the pay-before-launch trap, and most projects fall into it, because the schedule is fixed before anyone has counted the integrations.

On the other side sits the question you would like a ready number for: when does this pay back? And here is the thing no vendor will tell you.

Credible public benchmarks for marketplace payback barely exist. That is not a gap in your research.

It is a property of this market. The material circulating in the industry has three flaws at once, and any one of them on its own is reason enough not to build a business plan on it:

• It comes from interested parties. It is published by platform vendors and by consultancies that work with those vendors. None of them has an interest in publishing a payback period or a cost of ownership.
• It measures the survivors rather than the market. The sample takes in platforms that ran for the whole study period. The ones that stalled or were shut down drop off the count, and those are exactly the answer to "what is my risk."
• It has no costs inside it. You will find turnover growth, seller counts, and product counts. You will not find the implementation spend, the license cost, the cost of the team, or the break-even point.

The practical conclusion is simple and worth putting to the board plainly: you model a marketplace business case from scratch on your own numbers, and anyone who hands you a ready payback period is quoting sales material. If someone throws such a number into a meeting, one question checks it: from what sample, and does that sample include implementations that failed?

What you do know for certain, because it follows from arithmetic rather than from a study: if payback is measured in years, the sponsor has to be the board and not the e-commerce department. Someone assessed quarterly will not keep that investment alive.

And a second thing: every fixed cost calculated on turnover pushes the payback point further out. At €100 million of GMV, one percentage point is €1 million a year, whether or not you were in the black that year.

Do that arithmetic on your own numbers before you sign a multi-year contract.

How to build your own expectation instead of a borrowed one. Three figures you already hold, and they are enough for a first model: current online sales in the categories you want to open; what it costs you today to service a single order; and the realistic time it takes to bring a new supplier into your systems.

From those, build three scenarios: cautious, base, and ambitious. Then come back to them after two quarters with real data.

A model built on your own numbers and corrected after six months is worth more than someone else's table from a slide deck.

To calibrate your ambition, one figure that everyone in the industry knows is enough: at mature European retailers, sales from external sellers run into the tens of percent of online turnover. At the largest global players, they pass half.

That is a level reached after years rather than in the first year, and it tells you where the ceiling sits rather than what to expect at the start.

Which marketplace metrics do you need to measure?

Most organizations measure a marketplace with a single number. That number is GMV, and it is exactly the one that hides trouble for the longest.

1. Growth metrics

Growth metrics tell you whether the marketplace is growing:

• GMV and its share of online sales
• the number of active sellers (active, not registered)
• the number of offers and category coverage
• average cart value and how it moves
• organic traffic on the new categories
• the share of customers who buy both from you and from sellers

2. Operational metrics

Operational metrics tell you whether the marketplace works:

• the share of orders fulfilled within the promised time
• the share of orders with a problem: cancellations, returns, complaints
• the share of shipments with no tracking number
• the time from signing a contract with a seller to that seller's first sale
• the share of sellers who leave in the first year
• the completeness of product data

The second family usually gets left out of board presentations because it does not grow nicely. And it is the one that predicts what happens to the first family two quarters from now.

One metric deserves separate attention: the time from signing the contract to the seller's first sale. It looks like an operational metric, and it is a growth metric.

Shortening onboarding translates directly into how many sellers you can take on with the same team. The order of magnitude worth measuring yourself against: a well-organized onboarding is measured in days, an average one in weeks, and a broken one in months, and ends with a seller who stopped replying.

If you measure only one thing in the first year, measure this one.

Which phase gates should you set?

A definition of success without gates is a wish. A gate is a sentence in the form: "we move to the next phase if X, and if not, we do Y."

The minimum set that works:

1. Before launch

How many sellers and how many offers have to be ready on day zero. This is the most frequently skipped gate and the most expensive one. A marketplace that starts with a dozen or so sellers looks to the buyer like an empty shop, and an empty shop does not generate the data you were going to base your next decisions on.

2. After the first quarter

Whether sellers are selling or merely registered. Whether orders arrive on time. Whether the team is keeping up with the workload.

3. After the first year

Whether your share of sales is growing or flat. Whether the cost of servicing one order is falling. Whether there is a category where the marketplace beats buying the goods yourself.

And now the most important call in this whole article, because the advice you will hear on it is flatly contradictory. One school says: start as small as possible, validate, and only then invest.

Otherwise, you burn years and millions before you find out whether it works. The other school says: the marketplaces that died almost always started as a limited experiment run by one department.

They never reached the mass at which anything happens.

Both are right, because they are talking about two different things. The resolution worth remembering:

You may narrow the functional scope and the number of categories. You may not narrow the organizational commitment or the mass of supply at launch.

The narrowing that kills does not read "one category." It reads: one department, a pilot budget, a dozen or so sellers, and no sponsor on the board.

What does marketplace GMV hide?

Three things worth warning the board about before they turn up on their own.

Service quality can dip for a while. When part of your orders are fulfilled by strangers out of warehouses you do not control, satisfaction and conversion react.

We have no hard market statistics for this. What we have is the account of a practitioner who watched it happen at three consecutive launches of a large marketplace, and it is exactly the kind of thing vendor material does not cover.

Better to allow for it in the plan than to explain it away in month three.

The assortment forecast is usually overstated. An experienced seller decides deliberately which products go on which channel.

They weigh the commission, the competition, and the effort of writing descriptions, and they keep part of the assortment to themselves. That is the norm.

Counting "number of sellers × their catalog" therefore gives you a figure you will never see. Recruiting a seller is not the same thing as getting access to their assortment.

That is a separate conversation and a separate piece of work, and someone has to be assigned to it.

The team is bigger than it looks at the start. A planning heuristic to start from, before you measure it against your own process: one recruiter can bring on the order of a hundred sellers a year and make them active.

That is an order of magnitude rather than a measurement. It is still enough to show that three hundred sellers after a year means three full-time people rather than "someone from e-commerce at half capacity." On top of that come onboarding and account care, which are further roles.

How do the 6 reasons compare on scope and metric?

What does your reason for building a marketplace decide?

The sponsor has to sit at a level that outlasts the horizon. If payback is measured in years, the owner of the business case cannot be someone assessed quarterly.

The definition of success feeds straight into the fee model. If you are measured on margin, every cost calculated on turnover is your problem.

If you are measured on speed of entry, implementation time matters more than the rate.

Operational metrics have to be built together with the system. The share of orders with a problem, or the time to a seller's first sale, can only be calculated if the right events are recorded from day one.

Cannibalization is a conversation that takes several passes. Even the people who consider it a myth admit that one meeting will not settle it.

How do you set up marketplace metrics in 5 steps?

1. Write down one main intention and two secondary ones. One, rather than three of equal weight.
2. Pick one metric from each family. One growth metric and one operational metric become the pair the board judges the project by.
3. Set the launch mass: how many sellers and how many offers on day zero. That is a number.
4. Write the gates as conditional sentences. "If fewer than X sellers are selling after two quarters, we do Y."
5. Name the sponsor and the person who runs the metrics day to day. These are usually not the same person, and both have to exist.

Which mistakes do companies make when measuring a marketplace?

1. A pilot as a way of avoiding the investment decision

A narrow functional scope is sensible. A narrow organizational commitment is the best documented way to get the project closed after a year.

2. Measuring only GMV

It grows the longest, and it breaks the most abruptly. Without operational metrics, you find out about the problem when it comes back to you as a drop in repeat purchases.

3. An assortment forecast built by multiplication

The number of sellers times their catalog is a figure you will never see.

4. Headcount sized for launch day

The launch team is temporary by definition. Plan for growth.

5. No sentence saying what you do if it does not work

Without it, there is no decision to scale, because there is no alternative to compare it against.

What do you have to model on your own numbers?

It contains no forecast for your company. A marketplace business case has to be modeled from scratch on your own numbers.

If someone hands you a ready payback period, ask what sample it comes from and whether that sample includes implementations that failed.

Nor does it give commission benchmarks per category or a target cost structure. That is the subject of a separate chapter.

The figures given here are orders of magnitude and planning heuristics, not a measurement of your case. Treat them as a starting point for a conversation with your own finance people: good for calibrating ambition and for catching assumptions that are obviously unrealistic.

Bad for entering into a spreadsheet as a forecast.

Summary: What should you measure in your marketplace?

Name the reason first, because it decides which number counts. Then track both families: growth to see whether the marketplace is getting bigger, operations to see whether it is working.

Model the business case on your own numbers rather than on a borrowed payback period. Talk to a marketplace expert if you want to pressure-test your definition of success before the board sees it.

Frequently asked questions

Which metrics matter most for a marketplace?

Two families, and you need both. Growth metrics cover GMV and its share of online sales, active sellers, offers, and category coverage. Operational metrics cover the share of orders fulfilled on time, returns, disputes, and seller response times. GMV on its own hides trouble the longest.

How long does a marketplace take to pay back?

There is no credible public benchmark, and that is a property of the market rather than a gap in your research. Published figures come from platform vendors and their consultancies; they measure the platforms that survived the study period, and they exclude implementation and licence costs. Model it on your own numbers.

Why is GMV a poor single metric?

GMV grows while service quality, margin, and seller health are getting worse. It says nothing about orders fulfilled on time, about disputes, or about whether your margin after servicing costs is positive. Pair it with operational metrics from the start, because those have to be built into the system rather than added later.