Retail Media Network on a Marketplace: Build It, Buy It, or Hand It to an Agency

A retail media network on a marketplace sells sellers the visibility you already promised them once you took the commission, which makes the share of paid space the number that decides how much of your own search result you are willing to break.

Selling your own advertising space is what the trade calls retail media. You sell your sellers visibility you already promised them in part when you took the commission.

If you do not settle how much of that space may be sold, the team carrying an advertising revenue target will settle it for you, and you will find out from your own search results.

This article breaks down:

• Which 4 surfaces can you sell?
• What is the highest bid a seller can afford?
• How much traffic does it take?
• Which 3 routes lead there?

Key insights

• The four surfaces are search, the product page, category and home pages, and channels outside your site, and the product page is the one you leave for last, because it advertises a competitor beside the winner your own rule picked.
• The highest bid comes out of the seller's margin rather than your rate card: €880 after commission, a 20% margin, a quarter of it spent on acquisition, and 50 clicks per order give a ceiling of €0.88 a click.
• The traffic threshold is 6.25 million list views a month: a three-person team at €44,000 needs 50,000 paid clicks, which at a 0.4% click-through rate is 12.5 million impressions across two paid slots.
• The three routes are your own module, an off-the-shelf product, and an agency, and none of them takes the inventory off your hands: the slots, the labelling, the deduction from the balance, and the creative review stay with you.

Why does retail media sell the same access twice?

The canon of this series: a €1,000 cart, a 12% commission, €120 for you, €880 left for the seller. The commission is the price of access to demand.

The seller pays you because their offer stands in your search results and on your product page. Retail media sells them that same access a second time, only higher up the queue.

That is a parameter to set rather than an accusation. On a list of twenty results, every paid position takes the place of one the organic rule picked, so you are building a product whose quality is measured by how much it broke.

One number describes it: the share of paid space. Two slots out of twenty is 10%, six out of twenty is 30%.

That share grows on its own. A team with an advertising revenue target has one lever that works immediately and needs nobody's approval: add one more slot.

Revenue goes up in the same month. Result quality degrades over a quarter, and under a different owner.

That is why the share of paid space gets set once, as a number in the configuration, and changes by decision rather than by iteration.

Who wins the product page organically is settled by the rule that picks the winning offer. The ordering between products is settled by search and filtering.

Retail media replaces neither rule. It adds a second layer on top of them.

Which 4 surfaces can you sell?

1. Search results

The most expensive inventory, because the buyer has said out loud what they want.

For the same reason, it is the most sensitive to a swapped position.

2. The product page

The most explosive one, because you advertise a competitor next to the page winner your own rule selected.

One seller pays you to undercut your own decision, and the winner pays a commission on the same page. Leave this surface for last.

3. Category pages and the home page

The safest, because the intent is broad and a slot is easy to sell for a fixed amount, with no auction.

This is usually where the first real revenue appears.

4. Email and surfaces outside your own site

A separate consent regime and a separate question about data. See who controls the personal data.

What all four share, and what decides the cost of implementation, is this: paid positions have to be resolved by the same system response that returns the organic results. Three requirements follow, and nobody thinks about them at signing.

Duplicates have to be filtered out, so the same product does not stand on the list twice: once as an ad and once organically. The ad server needs a hard time limit.

And every position has to be logged: what stood where and why. How to label such a position for the buyer is covered by price history and discount messages.

What you owe your sellers on the parameters that drive the ordering is covered by the rules on decisions about sellers.

What is the highest bid a seller can afford?

There are three billing models.

1. Per click. What you pay with is the seller's conversion risk, and you need an auction and volume.
2. Per impression. You sell the space rather than the outcome. It works at small scale and with fixed slots, but it has to be sold by hand.
3. Per sale. This is a raised commission rather than an advertising model, and the seller will work that out in ten minutes and compare it against your grid (how much commission you can charge).

The ceiling on a bid comes out of the seller's arithmetic rather than out of your rate card. After commission, they keep €880.

Assume a 20% margin, which is €176 on the order. If they give a quarter of that margin to acquisition, they have €44 per order.

At a 2% conversion rate, meaning one order per 50 clicks, the highest bid that still pays for them is €0.88 per click.

That is a ceiling rather than a price. The real bid comes out of the auction and is often lower.

But the ceiling says something sales decks do not: your advertiser is a reseller rather than a brand. A brand funds advertising from a marketing budget and accepts a different return.

A reseller funds it from the margin you have already cut by 12%. That is why your commission is the upper bound on your own advertising revenue.

At a 15% commission rate, the seller keeps €850, the margin drops to €170, the acquisition budget to €42.50, and the bid ceiling to €0.85. Two revenue lines drawing from one pocket.

How much traffic does a retail media team need?

The minimum team is three people: someone who looks after advertisers, someone who runs campaigns and reviews creative, and someone on the product and reporting side. Assume a fully loaded cost of €44,000 a month.

That figure carries no ad server and no vendor's cut.

At €0.88 a click, you need 50,000 paid clicks a month just to break even on the team. At a 0.4% click-through rate on a paid position, that is 12.5 million ad impressions.

With two paid slots on a list of twenty, that is 6.25 million views of search and category listings a month. Below that threshold, the conversation is about a project rather than a revenue line.

Now the other side of the same calculation. Those 50,000 clicks at a 2% conversion rate give 1,000 orders of €1,000 each, so €1 million in GMV and €120,000 of commission for you.

Advertising revenue from that same flow is €44,000, about one-third of the commission. And if half of those orders would have happened without advertising, the incremental part is €500,000 of GMV and €60,000 of commission.

Across that half, the €44,000 is money moved from the seller's margin into yours, without a single extra unit sold.

This is the warning of this article, and no volume calculation contains it: advertising revenue scales with the number of advertisers rather than with traffic competing for the same query. An auction where one seller bids on a given query pays the floor price, and nothing but the floor price.

At a floor price of €0.20, those same 50,000 clicks bring in €10,000 instead of €44,000: the same traffic, revenue smaller by a factor of 4.4. The cure is more sellers in the same category rather than more traffic or a bigger share of paid space, which means recruitment (finding your first sellers).

Raising the share from two slots to six would give 37.5 million impressions in this calculation and €132,000, but only on the assumption that the click-through rate does not change, and it does, because positions further down get clicked less often. Tripled inventory does not give tripled revenue, and you will not learn the difference without a test on your own traffic.

Which 2 conditions decide whether sellers buy ads?

Advertising settled by a separate transfer will not be bought. The spend has to come off the seller's balance in the same cycle in which you settle sales: they have €880 due, they spent €300 on a campaign, they receive €580, and the statement shows both lines separately (the ledger behind a balance).

Settle three things up front.

A payment profile per seller. Balance or invoice, and what the limit is inside a cycle.

A limit set once usually stays forever, because changing it mid-cycle breaks the settlement, so treat it as a decision rather than as a field.

A negative balance. Advertising spend does not disappear when the sales dry up.

It rolls into the next cycle and joins the same queue as other deductions. That is the moment an advertising program starts producing collections work.

A return does not hand the advertising budget back. Commission on a returned order usually goes back to the seller (the commission on a refunded order), and the money spent on the click does not, because the click happened.

This is your only revenue line immune to returns, and your sellers will notice that faster than your own finance team.

Creative review is cheap in cost and expensive in timing. A product ad has no separate creative: it is the product page you already reviewed (the approval queue).

A banner does have one, and it has to be checked before it runs. Two hundred creatives a month at eight minutes each is just under 27 hours, a fraction of a full-time role.

But the campaign starts on Monday, and the file lands on Friday after hours, so the cost is the window and the person on duty rather than the work. You check four things: that the price on the creative matches the price of the offer, that there is no comparison against another seller, trademarks, and product claims.

A price on the banner that differs from the price in the cart is not a graphics problem (showing the price and price history and discount messages).

Which 3 routes lead to a retail media network?

None of the three routes takes the inventory off your hands. The slots on the list, the labeling of paid positions, the deduction from the balance, and the review of creative sit on your side in every variant.

An agency sells space it cannot render itself. So "we are handing this to an agency" is a decision about who sells rather than a decision about building and who ends up holding the data on your buyers.

On the middle route, one commercial condition matters most: the vendor's cut of advertiser budgets. Practitioners report that it is often counted in double-digit percentages, negotiated case by case, and highest for retailers that are small and new.

Those are exactly the ones with the least to share. The same practitioners report that advertising modules bundled with commerce platforms are bought rarely, and that operators reach instead for a specialized vendor or an agency.

What does retail media change about the rest of your build?

1. Advertising changes the arithmetic of the commission grid

Since the advertising budget comes out of the same margin as the commission, raising the commission rate lowers the ceiling on the bid. The two revenue lines have to be planned together.

2. The organic rule has to be able to say what it would have shown without the advertising

Without a record of the organic result, you cannot calculate how much of the sales a campaign moved and how much it created. That is the only question your board will ask you about retail media.

3. An advertising program changes what is in the seller contract

Deductions from the balance, limits, the rules for rejecting a creative, and the way terms get changed are an obligation rather than a promotional rulebook (the rules on decisions about sellers and the documents you sign with a seller).

How do you calculate whether retail media pays?

Five steps on your own numbers, doable in one day.

1. The bid ceiling. The amount left after commission × the seller's margin × the share of that margin they will give to acquisition, divided by the number of clicks per order. In our example: 880 × 20% × 25% ÷ 50 = €0.88.
2. The clicks needed to cover the team. The monthly cost of the team divided by the bid ceiling: 44,000 ÷ 0.88 = 50,000 clicks.
3. The traffic that is required. The clicks divided by the click-through rate, then by the number of paid slots on the list: 50,000 ÷ 0.4% = 12.5 million impressions, ÷ 2 = 6.25 million list views.
4. Competition rather than traffic. Take your hundred most frequent queries and count how many sellers have an offer for each one. Below three, the auction is a rate card, and your revenue will sit on the floor price.
5. Set it against the commission. Advertising revenue as a percent of the commission from that same flow. Industry analyses put mature retailers on the order of a fraction of a percent up to about one and a half percent of GMV, but those retailers have surfaces outside their own site, and they have brand budgets. What travels is the method of calculation rather than somebody else's table.

Which mistakes do operators make about retail media?

1. An auction built before there are any advertisers

The consequence: an expensive mechanism resolving competition that does not exist, and revenue frozen at the floor price.

2. The share of paid space set by the team that sells the advertising

The consequence: the number grows by one slot a quarter, and nobody has a date from which it was too much.

3. Advertising outside the seller's balance

The consequence: adoption drops to a handful of the largest sellers, because for everyone else the accounting cost is higher than the campaign.

4. The product page sold as the first surface

The consequence: a conflict with the rule that picks the winner, in the one place your sellers watch most closely.

5. Cannibalization measured by return on ad spend

ROAS is the revenue attributed to a campaign divided by its cost. It counts sales that follow a click rather than sales that would not have happened without the advertising.

The consequence: the report shows success even when the buyer would have bought the same product one position lower anyway.

What do you still have to settle about your own ad inventory?

Confirm three things with a specialist, because they touch the law: how to label positions that stand higher because somebody paid (price history and discount messages cover the mechanism), what you have to disclose to your sellers, and how (the rules on decisions about sellers), and whether you may sell space to sellers you compete with using your own offer.

Families of regulation to name to your lawyer, without settling which of them apply to you: those on digital services, in the part covering the labeling of advertising and recommender systems; those on transparency in the relationship between a platform and its sellers; those on the protection of personal data, where targeting is involved; and competition law, if your position in the category is strong.

Separately, settle with a tax advisor how you account for advertising revenue alongside commission ("Marketplace Revenue Recognition: Do You Book the Commission or the GMV?").

Our claim is narrower and independent of those answers: an operator who cannot state the share of paid space as a number is selling it anyway. They just do not know how much.

Summary: What does selling your own shelf cost?

It costs a share of the result you already sold once, and the number nobody writes down is how big that share is. Every paid position takes the place of one the organic rule picked, so the share of paid space is the parameter that decides how much of your own search result you are willing to break.

Left to the team carrying an advertising target, it grows by a slot a quarter, and the quality falls under somebody else's ownership. The economics are bounded by your own grid: your advertiser is a reseller funding ads out of the margin you already cut by 12%, so a higher commission rate lowers the bid they can afford.

And the revenue does not scale with traffic at all. It scales with how many sellers compete for the same query, which turns a retail media plan into a recruitment plan.

Ask a vendor where the advertising spend is deducted from, and what the organic result would have been without the ads. Building a marketplace where the share of paid space is a number in the configuration rather than a quarterly drift?

Talk to us about the build.

Frequently asked questions on retail media networks

What is a retail media network?

A retail media network is advertising space a retailer or marketplace sells on its own surfaces: search results, product pages, category pages, and channels beyond the site. On a marketplace, it sells sellers the visibility they already pay a commission for, one position higher up the queue.

How much can a marketplace charge for an ad click?

A marketplace can charge for an ad click no more than the seller's margin allows, which is a ceiling your own commission sets. After a 12% commission, the seller keeps €880; at a 20% margin and a quarter of it given to acquisition, 50 clicks per order put the ceiling at €0.88.

Raise the commission to 15%, and the ceiling falls to €0.85.

Should a marketplace build or buy a retail media network?

A marketplace should buy or hire rather than build first, because that reaches revenue in months rather than years, and none of the three routes removes the work that sits on your side. The slots, the labelling of paid positions, the deduction from the seller's balance, and the creative review stay with you whichever route you pick.