Where the Money Is Going
Retail media has moved from a line item to a budget category in its own right. Bain projects the global retail media market will reach approximately $140 billion by 2026, a figure that reflects how completely retailer-owned ad inventory has displaced older forms of trade investment for many CPG teams. Brands are already directing 39 percent of total ad spend through retail media networks, and that share is expected to reach 50 percent as planning cycles mature.
Those numbers land differently when you put a retailer name behind them. Amazon told investors this year that its grocery business generated more than $150 billion in gross sales in 2025, a claim CEO Andy Jassy used to position Amazon as the second-largest U.S. grocer. Amazon's retail media network, which spans Whole Foods' roughly 550 locations plus a large online ad platform, sits on top of that sales base. When the grocer is also the media owner is also the marketplace operator, the leverage a retail media network holds over a brand's search and display spend is significant. Jassy noted that perishables sales have grown more than 40 times year-over-year, and that same-day perishables customers build larger baskets. For a brand trying to reach high-frequency grocery shoppers, that kind of audience concentration is hard to ignore.
The Measurement Problem Is Not a Temporary Gap
Spend at this scale demands proof. The problem is that proof is the hardest thing to produce in retail media right now.
Most retail media networks report on attributed sales, which counts every purchase made by a shopper who saw an ad in a given window. Attribution is not incrementality. Incrementality asks the harder question: did the ad cause the purchase, or did the shopper buy anyway? The gap between those two numbers is the gap between what brands think retail media is delivering and what it is actually delivering.
The industry has known this for several years. What is changing is the financial pressure on brand P&Ls, which makes the gap harder to tolerate. Shoppers remain highly price-conscious. Sonja Evans, vice president of business intelligence and strategy at Blue Chip, noted that consumers have "recalibrated what they're willing to pay and where," making legacy forecasting approaches less effective. When shoppers are selective about spend, brands cannot afford to pour half their ad budget into a channel they cannot fully audit.
Publicis Sapient's Sudip Mazumder, senior vice president and retail industry lead, has pointed to AI as a way to move beyond blanket discounting by aligning decisions to the performance of individual SKUs. The same logic applies to retail media. Broad network buys across all placements on a retailer's platform are the ad-world equivalent of blanket discounting. The value is hidden inside an average, and the average flatters the seller more than the buyer.
What Retailers and Brands Are Doing Right Now
Some retailers are building the measurement case more seriously than others, and the gap between them is starting to show in how CPG teams allocate.
Amazon's scale gives it a structural advantage in incrementality testing. Its closed-loop data environment, connecting ad exposure to purchase at the individual level, is what most grocery retail media networks are still working toward. Smaller regional grocers with loyalty programs can produce matched-panel tests, but the sample sizes needed for category-level confidence are harder to reach at lower traffic volumes.
Sprouts is a useful case study for the tension between network investment and shopper behavior. The chain reported first-quarter net sales growth of 4 percent to $2.3 billion in 2026, but comparable-store sales fell 1.7 percent. CEO Jack Sinclair acknowledged the affordability challenge. Sprouts responded with selective price cuts on a small number of SKUs, including coffee, and leaned on a chainwide loyalty program launched in October to sharpen its targeting. A grocer whose comp sales are declining while total sales grow through new stores is telling its brand partners something important: the existing shopper base is under pressure, and the audience a brand thinks it is reaching through on-site media may be smaller and more fragmented than the network's headline reach figures suggest.
For brands thinking about off-site retail media, which uses a retailer's first-party data to reach shoppers across the open web and connected TV, the measurement challenge is even sharper. The connection between an off-site impression and a verified in-store purchase depends entirely on the retailer sharing clean, timely data with the brand or its agency. Many networks are still building that pipeline.
Implications
If you are heading into a retail media planning cycle in the next 90 days, three things are worth doing before you approve the numbers.
First, separate your spend by placement type and demand incrementality data for each. On-site sponsored search on a retailer platform, where shoppers are already in a buying mindset, has a better baseline case than off-site display. But even on-site, the incrementality profile varies by category and by whether you are defending share against a promoted competitor or genuinely growing the basket.
Second, ask your network partners what their incrementality methodology is. Matched-panel holdout tests, where you compare purchase rates in exposed versus unexposed shopper groups, are the current standard. If a network cannot explain its method in plain terms, treat its attribution numbers with caution.
Third, watch what AI-driven tools are doing to the buying side of this market. Finance leaders are already using AI to move away from legacy forecasting models and toward SKU-level decision-making. The same analytical shift is coming to retail media allocation. Brands that build the data infrastructure now, connecting ad exposure to actual sell-through at the SKU and store level, will have a real advantage over those still working from blended network averages when the 50-percent-of-budget moment arrives.
Retail media is not going to shrink. The networks are too embedded in how retailers fund their tech stacks, and too central to how shoppers discover products, for that to happen. But the brands that treat it as a non-negotiable cost of trade rather than a channel that must earn its allocation are effectively subsidising retailer margin at the expense of their own. The measurement discipline is available. The question is whether your commercial team is demanding it.