A supermarket makes almost no money selling you food.
That is not a figure of speech. Grocery is one of the thinnest-margin businesses in the economy. Tesco runs an operating margin near 4.5 percent, Sainsbury's closer to 3.2 percent, and the wider industry posts net margins around 1 to 2 percent, according to industry benchmarking summarised here. A grocer that sells a billion dollars of cans, milk and bread keeps a few tens of millions after everything is paid.
Now look at the sponsored slot at the top of that grocer's app search results. The same analysis estimates retailers earn margins of 70 to 90 percent on the ads they sell on their own sites. The cost of selling a pixel of screen space to a brand that wants to be seen is close to nothing, because the screen, the traffic and the shopper data already exist. The retailer built them to run a store. Selling access to them is almost pure profit layered on top.
Once a chief financial officer sees those two numbers side by side, the rest is inevitable. The retailer stops thinking of itself as only a store. It becomes an advertising platform that happens to own shops.
Origin: how a marketplace problem became a profit centre
Retail media did not start as a grand strategy. It started as a crowding problem on Amazon.
By the early 2010s Amazon's marketplace had a glut of third-party sellers offering near-identical products. Everyone wanted to be on the first screen of results, and the first screen has only so much room. In 2012 Amazon began assembling the pieces of an ad business to ration that attention, grouping its early efforts into the Amazon Media Group, the Amazon Advertising Platform and Amazon Marketing Services. The format that came to define the category was Sponsored Products, paid placements that sit inside search results and category pages and look almost like the organic listings around them. Amazon pushed it hard from 2016, the year it also launched a dedicated advertising blog.
This was a different animal from the display advertising that came before. A banner on a news site reaches someone who is reading the news. A Sponsored Product reaches someone who has just typed the name of the thing they want to buy, on the site where they will buy it, seconds before the decision. The intent is already there. The ad just decides which brand captures it.
Other retailers watched Amazon book real money and realised they were sitting on the same asset. They too had high-traffic sites, logged-in shoppers and complete purchase records. Walmart built a paid search platform with its own engineers, and in January 2021 relaunched its media business as Walmart Connect, stating openly that it wanted to become a top-ten US advertising platform. Target had Roundel, Kroger had Precision Marketing, and the grocers, drugstores and home-improvement chains followed. The industry even coined a name for the thing each retailer was now building: the retail media network.
The deeper shift is one of identity. For a century a retailer's product was the merchandise on the shelf. Retail media made the shopper's attention a second product, sold to the brands that want it, at a margin the merchandise could never touch.
Present: a third profit engine, and why it carries the others
Retail media is now the fastest-growing major channel in digital advertising. US advertisers spent about 60.3 billion dollars on it in 2025 and are forecast to spend roughly 71 billion dollars in 2026, growth near 18 percent in a year, per eMarketer. Globally, Forrester projects retail media will pass 300 billion dollars by 2030 and is already climbing toward a quarter of all digital ad spend. This is not a niche tactic. It is one of the three or four largest things happening in advertising.
For the retailers, the appeal is structural, and it comes down to three advantages that no traditional media owner can match.
The first is the margin gap described above. Retail media revenue is high-margin income bolted onto a low-margin business, and most of it falls almost straight to operating profit. For some retailers the ad business now generates more profit than the entire retail operation that surrounds it. Amazon's advertising revenue passed 68 billion dollars across 2025, a sum that would make it a large media company on its own. Sainsbury's set a public target of more than 100 million pounds in additional profit from its Nectar360 media arm. Retail media is no longer a side hustle. It is a third profit engine, and a disproportionate share of group profit can ride on it.
The second advantage is first-party purchase data. A retailer does not need cookies, device IDs or guesswork to know who its shoppers are and what they buy. It has loyalty cards, logged-in accounts and transaction histories that record actual purchases of actual products. After years of signal loss across the open web, that kind of clean, consented, deterministic data is rare and valuable. A retailer can let a brand reach households that bought a competitor's product last month, or that buy the category but never the brand, with a precision an open-web advertiser can only envy.
The third advantage is the one that pulls budgets hardest: closed-loop attribution. This is the heart of retail media's pitch, and it deserves a clear definition.
What closed-loop attribution actually means
In most advertising, the ad and the sale happen in different places. You show someone a video, they later buy something in a shop or on another site, and you spend enormous effort trying to connect the two with models and assumptions.
Retail media closes that loop because the ad and the purchase happen inside the same system. The retailer shows the ad. The retailer rings up the sale. It can match the exposure to the transaction directly, because it owns both ends. A brand can be told, at the SKU level, that a campaign on a retailer's site drove a measurable number of sales of its product on that same site.
For a marketer who has spent a career arguing about whether advertising worked, this is intoxicating. The dashboard shows a clean return on ad spend, a number that says "you put in one dollar and got back five." It feels like proof. It is the single biggest reason budgets have moved into retail media so fast.
It is also where buyers need to slow down, because the clean number hides three real problems.
The blind spots behind the clean number
The first problem is that the retailer grades its own homework. The same company sells you the ad, measures the ad and reports the result, and it has a direct commercial interest in that result looking good. There is no independent referee in the standard setup. The Current, an industry trade publication, has described the consequence bluntly: retail media has no single source of truth, and every network measures in its own way. Each one defines its attribution window, its idea of a view and its return calculation differently, so a return on ad spend of 5 on one retailer and 5 on another are not the same number. They cannot be compared, even though they share a name.
The second problem is over-attribution, and it is the one that quietly inflates almost every retail media dashboard. Closed-loop attribution counts the sales that followed an ad exposure. It does not, by default, ask the only question that matters: would that sale have happened anyway? A loyal customer who buys the same brand of coffee every week will buy it again. If that customer happens to pass a sponsored placement on the way to the basket, the system credits the ad with a sale that was always coming. The reported return mixes genuinely new sales with sales the brand had already earned. The honest metric is incrementality, the extra sales that exist only because the ad ran, often measured as incremental return on ad spend, or iROAS. As Skai explains, an iROAS above roughly 2 signals real behaviour change, while a figure near or below 1 means the spend is mostly buying demand that would have converted on its own. Standard ROAS and incremental ROAS routinely tell very different stories about the same campaign. By 2024, surveys of advertisers found incrementality had overtaken ROAS as the metric they most wanted, precisely because the headline number had lost credibility.
The third problem is the largest and the least discussed: the closed loop only closes online. Retail media measurement is excellent at connecting an ad to a purchase on a website. Most shopping does not happen on a website. Roughly 83 percent of US retail sales still happen in physical stores, and for Walmart and Target the in-store share is above 90 percent, as MetaRouter sets out. Yet around 77 percent of retail media spend goes to online inventory, and in-store retail media is less than 1 percent of the market. So the celebrated closed loop is measuring a minority of the commerce it claims to represent. A campaign can post a beautiful online return while its effect on the store down the road, where most of the brand's sales actually occur, sits entirely outside the report. Absence of evidence gets quietly read as absence of effect.
None of this means retail media does not work. It often works well. It means the dashboard flatters it, and a buyer who reads the dashboard literally will overspend.
Future and impact: beyond retail, and beyond the comfortable number
Two shifts will define the next few years.
The first is that the model is escaping retail entirely. Any company that sits on a transaction and a base of logged-in customers can do what Amazon did, and they are. The broader category is now called commerce media, and it covers far more than shops. Uber's advertising business has passed 2 billion dollars in annualised revenue, growing more than 50 percent a year on sponsored listings inside its rides and delivery apps. Instacart's ad revenue runs past 1 billion dollars a year, and DoorDash crossed a 1 billion dollar annualised run rate too. Banks have joined: JPMorgan Chase launched Chase Media Solutions in April 2024, and PayPal and Klarna have built advertising arms of their own, turning a bank's view of where customers actually spend into a targeting asset. Airlines, hotels and delivery firms are building the same thing. The logic that converted a grocer into an ad platform converts a rideshare app, a checkout button and a credit card just as cleanly. Every business with a checkout is a candidate.
The second shift is corrective. The measurement weaknesses are now too obvious and too expensive to ignore, and the industry is being pushed toward independent, comparable proof. IAB Europe has updated its commerce and retail media measurement standards and built a certification programme so a network can show audited compliance rather than just assert good numbers. In December 2025 the IAB published a framework for maturing in-store measurement, aimed squarely at the gap between online attribution and where sales really happen. The direction of travel is clear: away from self-graded ROAS, toward standardised definitions and incrementality testing run by parties who do not also sell the ads.
There is an agentic angle here worth naming. As AI buying agents take on more campaign decisions, they will optimise relentlessly toward whatever number they are given. Point an agent at raw retail media ROAS and it will pour budget into the campaigns that look best, which are often the ones harvesting loyal customers who would have bought anyway. The metric you feed the machine becomes the strategy the machine pursues. Getting incrementality into that loop is not a refinement. It is the difference between automation that grows the business and automation that efficiently buys sales you already had. Perform Digital's work with enterprises on agentic systems starts from exactly that point: the objective you optimise has to be the one that is actually true.
For a marketing decision-maker, the takeaway is two-sided. Retail media is real and large, and the first-party data and point-of-purchase intent behind it are genuine advantages that are not going away. Treat it as a permanent part of the plan. But read its numbers like an auditor, not a fan. Ask whether the reported return is incremental or just attributed. Refuse to compare ROAS across networks that measure differently. Remember that the closed loop closes online while most of your sales close in a store. Retailers turned themselves into ad platforms because the margins were extraordinary, and that same margin is why they will always have a thumb on the scale of the report. Knowing that is most of the skill.
Council summary
This post argues that retail media is an identity shift: thin-margin retailers found that selling shopper attention earns 70 to 90 percent margins, rebuilt themselves as ad platforms, and the same logic is now converting rideshare apps, delivery firms and banks into commerce media networks. It credits the three real advantages, high-margin revenue, first-party purchase data and closed-loop attribution, then shows how that clean attribution flatters performance: networks grade their own homework, standard ROAS counts sales that would have happened anyway, and the loop only closes online while most sales close in stores. The review verified the central figures against eMarketer, Forrester, Marketing Dive and MetaRouter, confirmed the BCG margin range, and corrected the commerce media section, which had understated Instacart's ad revenue and overstated the precision of the 2016 Sponsored Products launch. The takeaway: treat retail media as a permanent channel, but read its dashboards like an auditor and press for incrementality before scaling spend.
Comments