You are looking for coffee on a supermarket website. The first pack displayed is not necessarily the cheapest or the most popular: its brand may have paid for that position. A few hours later, a video campaign might remind you of that same coffee. Welcome to retail media, where retailers sell products, but also access to their customers. Looking ahead to September 2026, the stakes extend beyond advertising: who will capture the value of purchase data, and who will prove that these ads actually drive more sales?
The shelf becomes advertising space
The principle is not entirely new. Brands have long paid for flyers, promotional events and prominent in-store placements. Digital technology, however, changes the scale and precision of the approach. A retailer can sell sponsored results in its search engine, banners in its app, advertising on partner media and screen placements in its stores.
Three areas are emerging: on-site, on the retailer’s digital properties; off-site, which uses its audiences to deliver ads elsewhere; and in-store, where advertising exposure meets the checkout. Their common feature is crucial: proximity to the purchase itself. While a social platform knows your presumed interests, a retailer may know which product you actually bought.
Yet this knowledge is never comprehensive. It depends on customer identification, loyalty card use, purchases linked to an account and the applicable permissions. It also describes a specific commercial relationship: buying diapers from one retailer does not mean buying all your baby products there.
A second economic engine for retailers
For a retailer, the promise is appealing. Its core business requires inventory, warehouses, staff and costly logistics. Advertising allows it to monetize traffic it has already attracted. That revenue can deliver attractive profitability, although sales teams, ad-serving tools, measurement and data governance still need funding.
Amazon has played a major role in establishing this model: its advertising business, reported separately in its earnings from 2022 onward, demonstrated that commerce could become a powerful advertising engine. Walmart has also developed Walmart Connect. In Europe, Carrefour and Publicis launched Unlimitail in 2023, while alliances such as Infinity Advertising, bringing together Intermarché and Casino, had already illustrated the push for critical mass.
These initiatives, documented before September 2026, shed light on the momentum without prejudging its outcome. For retailers, pooling resources can make it easier to reach major advertisers. For advertisers, the appeal is the reverse: buying campaigns across multiple networks without having to manage as many contracts, formats and dashboards as there are retailers.
For brands, promise and dependence
Retail media addresses a concrete need. A brand launching a laundry detergent can reach shoppers in that category, boost its visibility at the moment of choice and track the associated transactions. It can also seek customers who had not previously bought its products, rather than repeating its message to already loyal consumers.
But this apparent effectiveness masks a tension. When top positions become paid placements, a brand sometimes has to buy visibility it previously enjoyed organically. The risk is a commercial toll: paying to remain visible, particularly when competitors bid on the same searches. Advertising budgets may then protect existing sales rather than generate new ones.
The supplier-retailer relationship also becomes more complicated. Media spending can compete with promotions, trade discounts or brand advertising. A marketing manager therefore needs to ask a less eye-catching question than the reported return: is this budget genuinely new, or simply shifted from another spending category?
The real test: attribution is not proof
This is the heart of the debate. A customer sees an advertisement, then buys the product: the advertising network attributes a sale to that ad. But would the customer have bought it without that exposure? If they had already typed the brand name into the search engine, the answer might well be yes. An attributed sale is not necessarily an incremental sale.
The usual return on ad spend, often presented as the ratio of attributed revenue to media spending, does not resolve this problem. It varies according to the attribution window, whether a mere impression or a click counts, and which purchases are included. Two campaigns can show very different performance because their calculation rules differ.
Measuring incrementality requires estimating what would have happened without the campaign. One robust method involves randomly assigning people to an exposed group or a control group and comparing the results, where the setup allows it. Geographic or store-level tests can complement this approach. No protocol is a magic solution: simultaneous promotions, stockouts, local differences and customer movement can muddy the results.
Questions to ask before signing
- What outcome are we seeking? Additional sales, new customers and incremental margin do not tell the same story.
- What is the comparison scenario? The protocol must explain how sales that would have occurred without advertising are estimated.
- What is the scope? A gain for one product can cannibalize another format, another brand in the group or another sales channel.
- What verification is available? Definitions, exclusions and limitations must be accessible, with independent evaluation where possible.
Purchase data is not a blank check
Another boundary is trust. Data held directly by a retailer is not free of constraints. In Europe, the GDPR requires, among other things, an appropriate legal basis, transparency and respect for individuals’ rights. Rules on tracking technologies also apply, depending on the setup. Having a loyalty card does not mean agreeing to every conceivable advertising use.
Secure environments for matching data, often called data clean rooms, can limit the exchange of individual-level files between partners. They do not, however, automatically make every processing operation lawful or every result reliable. Data minimization, retention periods and access controls remain essential.
What next?
What next? For September 2026 and beyond, the plausible scenario is a more demanding market, not simply a larger one. Advertisers could favor networks capable of combining reach, execution quality and reproducible incrementality tests. Retailers will have to balance advertising revenue against the shopping experience: too many sponsored placements undermine the relevance of the digital shelf. Retail media will earn a lasting place if it demonstrates shared value creation. Selling a placement is easy; proving that it genuinely changes a purchasing decision remains the real job.