Retail media networks as a profit engine

# Retail media networks as a profit engine

Two euros land in a hypermarket's P&L on the same Tuesday. The first comes from selling a pack of coffee: after cost of goods, shrink, staff and the store's share of overhead, three or four cents survive. The second comes from the coffee brand paying to sit at the top of the search results for "coffee". Most of that one survives, because the page, the traffic and the checkout were already paid for. Same shopper, same second, different economics by an order of magnitude.

That gap is why a retailer stops behaving purely as a buyer and reseller of goods and starts behaving as a media owner renting access to its own shoppers. This lesson is about the second euro: how shopper data gets packaged into inventory suppliers will pay for, what the resulting P&L looks like beside trading margin, and where the reported profit is thinner than the chart suggests.

What a retail media network actually is

A retail media network (RMN) is a retailer's advertising business: ad space on its site, app, emails and in-store screens, sold to suppliers and targeted with the persistent shopper profile the loyalty lesson describes.

Renting shelf space, except digital and measurable down to the receipt.

Three assets make it sellable:

• Audience: purchase history at household level, across categories, not demographic guesswork.
• Placement: search and category pages, where the shopper has already declared what they want.
• Closed-loop measurement: the retailer owns the ad and the till, so exposure can be tied to a basket.

Most advertising ends in an inference. Retail media ends in a receipt, with caveats we come back to.

Three networks, three shapes

Amazon is the reference case. Sponsored results are an auction, and the business now runs above $50 billion a year in advertising services revenue, larger than most global media companies.

Walmart Connect launched in 2021 out of the older Walmart Media Group. It pairs sponsored search with in-store screens and audio across a store base that sees hundreds of millions of visits a week, and Walmart's 2024 purchase of TV maker Vizio bought it a screen in the living room plus the software that sells ads on it.

Carrefour shows the European shape. It packaged its data and media assets under Carrefour Links, then in 2023 created Unlimitail with Publicis, a joint venture selling retail media across Carrefour banners in Europe and Latin America and for third-party retailers too. A grocer running low single-digit net margins treating media as a structural profit line, not a side experiment.

All three sell the thing this lesson describes, so their published ROASROASReturn on Ad Spend (ROAS) measures the revenue generated for every unit of currency spent on advertising, calculated as revenue divided by ad cost.View full definition → claims are sales collateral; read them that way. The IAB publishes vendor-neutral measurement standards worth bookmarking.

The P&L next to trading margin

Take a retailer with €50bn of sales, 25% gross margingross marginGross margin is the share of revenue left after subtracting the direct cost of producing goods or services, expressed as a percentage of revenue.View full definition → and 3% net. It builds a media business doing €200m of revenue at a 70% contribution margin: €140m of operating profit. To find €140m from trading instead, at 3% net, you would need roughly €4.7bn of extra sales, about nine points of growth. That is the whole argument in one line, and it is why boards fund these things.

Now the three things that shrink the number.

Where the money is booked. Supplier money paid as listing fees or promotional allowances is normally netted off cost of goods sold. Money paid for a distinct advertising service can be recognised as revenue. Same supplier, same annual negotiation, different line, very different-looking growth rate. A media business built by relabelling trade spend inflates revenue without adding a cent of profit. Your auditors will have views; get them early.

Where the budget came from. Part of any RMN's revenue is reallocated trade money, the budget that used to buy gondola ends and deeper promotions. Suppliers know exactly what they are doing and arrive at the annual negotiation asking for media spend to count toward the total. Net incremental profit is smaller than gross media revenue, sometimes much smaller.

Scale threshold. The costs are a sales team, an ad platform (built or licensed), measurement and reporting. Below a certain traffic level the auction has no liquidity: two advertisers per keyword means bids sit at the floor, fill rates stay low, and the salespeople cost more than they bring in. Small and mid-size retailers usually do better joining someone else's network, as Unlimitail's third-party model assumes, than running their own.

How sponsored placements get priced

Most retail media search inventory clears through an auction. Advertisers bid on search terms and the highest effective bids win, subject to relevance.

Two pricing models dominate:

• CPC (cost per click): paid per click, standard for sponsored search.
• CPM (cost per thousand impressions): paid per thousand views, used for display, video and in-store screens.

Brands watch ROAS (return on ad spend): revenue per unit of spend, so a ROASROASReturn on Ad Spend (ROAS) measures the revenue generated for every unit of currency spent on advertising, calculated as revenue divided by ad cost.View full definition → of 4 means four euros of sales per euro spent.

A simplified auction snapshot

for each ad slot on the search page:
    eligible_ads = ads bidding on this keyword
    for ad in eligible_ads:
        rank_score = ad.bid * ad.relevance_score
    winner = highest rank_score
    winner pays just enough to beat the next competitor

The highest bid does not automatically win. Relevance is weighted in, because junk results drive shoppers away and cost more in lost baskets than the click was worth.

The failure mode sits in the ROASROASReturn on Ad Spend (ROAS) measures the revenue generated for every unit of currency spent on advertising, calculated as revenue divided by ad cost.View full definition → number itself. Last-click attributionattributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → on the retailer's own site credits the ad for sales that would have happened regardless, and the effect is worst on branded keywords: a shopper typing "Nespresso" was going to buy Nespresso. Holdout tests, where a matched group of shoppers is deliberately not exposed, routinely show incremental ROASROASReturn on Ad Spend (ROAS) measures the revenue generated for every unit of currency spent on advertising, calculated as revenue divided by ad cost.View full definition → well below the reported figure. Sophisticated suppliers run those tests. Offering incrementality testing before they demand it looks like turkeys voting for Christmas, but it protects the budget from being cut wholesale the day they run it themselves.

Monetizing first-party audiences

Audience segments. "Households that bought a competitor's diapers in the last 90 days" or "lapsed buyers of our private-label coffee". Built from purchase history, which beats the demographic proxies most ad platforms sell.

Off-site extension. Retailers let suppliers use shopper data to target on social feeds, connected TV and the open web, then measure the sales back at their own checkout. This is where the addressable budget grows well beyond the retailer's own pages, and where Walmart's Vizio deal points.

In-store media. Screens, shelf displaysdisplaysThe total number of times an ad or piece of content is displayed, regardless of clicks. Each display counts as one impression, even to the same person.View full definition → and audio. ReachReachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → is enormous and the shopper is metres from the product. AttributionAttributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → is the weak point: proving a screen near the entrance moved a basket usually requires store-level tests rather than person-level tracking.

The privacy guardrail

First-party dataFirst-party dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.View full definition → gained value as third-party cookies (tracking files following users across unrelated sites) declined under browser changes and regulation. GDPR in the EU and CCPA in California govern what can be done with the rest. Standard practice: aggregate and anonymise so suppliers see segment-level results, never individual records, with data clean rooms doing the matching. This is not legal advice; involve counsel before launch.

🎬 [VIDEO: "How Retail Media Networks Work" - youtube.com - a clear explainer on how retailers monetize shopper data and ad placements]

Balancing ad load against customer experiencecustomer experienceThe overall perception a customer forms of your brand across every interaction, from first touch to post-purchase support.View full definition →

Ad load is the share of a page or session given to paid placements. Push it up and revenue rises immediately while conversion, trust and repeat visits decay slowly. The quarter looks excellent. The decay shows up two years later in traffic you cannot easily explain.

The internal tension is structural: the media team is paid on ad revenue, the category manager on trading margin, and the two now sell the same square centimetres. Private label sharpens it further, since the retailer's own brand competes for the slot it is auctioning to the supplier who funds the category.

Guardrails mature networks use:

• Relevance thresholds, so weak ads cannot buy the top slot.
• Hard caps on sponsored positions per page or per search.
• Clear "Sponsored" labelling, per disclosure rules from bodies like the FTC.
• Organic conversion tracked as a health metric with the same seriousness as ad revenue.

Attention is the finite input. Overharvest it and the inventory devalues itself.

Building one: the practical sequence

1. Get the data house in order. Unified, clean customer and purchase data. Without it there is nothing to sell.

2. Start with sponsored search. Highest intent, highest margin, easiest to measure.

3. Prove results to suppliers. Self-serve reporting, and incrementality tests offered rather than resisted.

4. Expand formats slowly. Display, then off-site, then in-store, watching ad load and organic conversion at each step.

5. Separate the negotiation. Media budgets discussed in the same room as listing fees turn back into trade spend within two cycles.

Where the industry is heading in 2026

In-store retail media is scaling as shelves and screens get digitised, and suppliers are pushing hard for comparable measurement across networks, because reconciling dozens of retailer dashboards has become an operational cost of its own. Retailers that make their numbers auditable will take share of budget from those that do not.

Key Takeaways

• Media revenue at 70% contribution can equal the profit of billions in extra trading sales; that arithmetic, not novelty, is the case for building one.
• Ask where the money came from. Reallocated trade spend booked as revenue flatters the top line and adds nothing to profit.
• Closed-loop measurement is the product, and last-click ROASROASReturn on Ad Spend (ROAS) measures the revenue generated for every unit of currency spent on advertising, calculated as revenue divided by ad cost.View full definition → overstates it. Run holdouts before your suppliers do.
• Below a certain traffic level the auction has no liquidity; joining a network beats building one.
• Ad load is a slow-acting poison. Track organic conversion with the same discipline as ad revenue.