How big the market has become
Retail media is no longer a budget line that needs defending in an investment committee. Bain projects the global retail media market will reach roughly $140 billion by 2026, a figure that reflects how completely the channel has moved from experimental to structural inside CPG planning. For context, that is larger than the entire US television advertising market was at its peak.
The spend share tells the same story. CPG brands are already directing 39 percent of their total ad budgets through retail media networks, and industry expectations put that share at 50 percent within the current planning horizon. When half of a brand's total advertising flows through channels owned and priced by its retail partners, the relationship between brand and retailer changes in ways that go well beyond a joint business plan.
Amazon is the clearest example of that shift. Its grocery business crossed $150 billion in sales in 2025, which gives its retail media network a data foundation that most competitors cannot match. Every search, click, and basket completion on Amazon feeds the targeting and attribution system that brands are buying into when they place a sponsored product or off-site display ad. No pure-play digital platform offers that closed-loop connection between ad exposure and purchase.
Where the spend is actually going
Retail media budgets now flow across three distinct environments, each with different audience quality and measurement maturity.
On-site media, meaning sponsored search and display within a retailer's own app or website, still takes the largest share of spend. The audience is in active buying mode, which makes click-through and conversion rates look strong. The risk is that on-site placements often capture demand that already existed. A shopper searching for your brand by name and then clicking a sponsored result is not a shopper you converted. You paid to intercept your own buyer.
Off-site media extends retail audience data to programmatic display, social, and connected TV. The retailer sells you access to segments built from its first-party purchase data, and you reach those shoppers while they are browsing outside the retailer's own environment. This is where the incrementality question gets genuinely hard. The audience targeting may be sharp, but the path from ad impression to in-store or online purchase runs through multiple touchpoints that no single attribution model captures cleanly.
In-store digital media, screens, digital end-caps, and audio formats at the physical point of sale, is the fastest-growing segment by investment interest even if it remains the smallest by current spend. Brands are drawn to it because it reaches shoppers at the moment closest to the purchase decision. Retailers are drawn to it because it monetises floor space that previously generated no media income. The friction is operational: creative formats are not standardised, inventory is fragmented across retail banners, and measurement depends entirely on what the retailer chooses to share.
The incrementality problem in plain terms
Every retail media network sells you attribution. Almost none of them sell you incrementality by default. Those are different things.
Attribution tells you which ad touchpoint came before a purchase. Incrementality tells you whether the purchase would have happened without the ad. A shopper who sees your promoted listing and then buys your product may have been planning to buy it anyway. The retailer's attribution system records a conversion. Your P&L takes the credit. Neither of you has learned anything about whether the placement created value.
Most brands cannot run true incrementality tests, meaning holdout group experiments where a matched set of shoppers sees no ad, across more than a small fraction of their retail media placements. The testing infrastructure requires retailer cooperation, clean audience splitting, and a willingness to sacrifice some short-term attributed sales in the name of learning. Retailers have limited incentive to make that easy. If incrementality testing reveals that a large share of retail media spend captures existing demand rather than creating new demand, the brands doing the testing will spend less, not more.
This is not a reason to stop spending. It is a reason to invest in the measurement capability that lets you distinguish high-incrementality placements from low ones and shift budget accordingly. Brands that build that capability now will have a structural advantage in every annual planning negotiation with their retail partners.
What inflation is doing to the retail media calculus
The broader cost environment makes measurement discipline more urgent, not less. Food-at-home price inflation could exceed 4 percent in 2026, according to an FMI briefing from May 14, with the Producer Price Index for food growing 2.2 percent annually in April alone. When input costs rise and consumer spending tightens, the brands that can prove their media spend is generating genuine incremental sales will protect their budgets. Those that cannot will face pressure to cut.
Kantar's BrandZ analysis shows that brands combining emotional connection, perceived difference, and cultural relevance outperform peers on pricing power in inflationary periods. That finding has a direct implication for retail media allocation. Awareness-building placements, off-site display and in-store formats that reach shoppers before they are in buying mode, may deliver more durable brand equity than on-site sponsored search alone. But the measurement frameworks most brands use today are optimised for the last click, which systematically undervalues the upper parts of the retail media funnel.
What Mars is doing that is worth watching
Mars is explicitly targeting "every snacking occasion, online and in-store" and coupling new format launches with digital shopping tools designed to reshape discovery across channels. The combination of format innovation at the physical shelf, M&M's POP'd Caramel, Twix Bits, and new Skittles and Starburst lines, with digital-first activation reflects a media strategy that uses retail media to drive trial of genuinely new products rather than defend existing share. That is the version of retail media spend most likely to pass an incrementality test: a new format that a shopper had not previously considered buying, surfaced at the right moment by a targeted placement.
The contrast with defensive spending is sharp. A brand using retail media primarily to protect its top item on a high-velocity shelf is buying insurance. A brand using it to introduce a new format to a shopper who has not yet encountered it is buying growth. Both can be justified, but only one is building something.
What to watch in the next 90 days
Three things are worth tracking closely as the second half of 2026 takes shape.
First, watch how large CPG companies report retail media spend in earnings calls. If brands begin breaking out retail media as a distinct line in trade investment reporting, rather than burying it in broader marketing or trade budgets, the pressure to prove incrementality will accelerate.
Second, watch retailer moves on measurement standardisation. Any retailer that offers brands a clean holdout-group testing methodology, rather than proprietary last-touch attribution only, is making a long-term bet that transparency will attract more spend. That is a smarter commercial model than defending opacity.
Third, the emergence of generative AI as a product discovery channel, explored in recent coverage of generative engine optimisation, adds a fourth shelf to the on-site, off-site, in-store framework. If shoppers begin using AI assistants to inform grocery choices, the brands with structured product data and strong digital content will appear in those answers. Retail media networks that can extend their targeting and measurement into AI-assisted discovery will have a meaningful advantage. Those that cannot will find their on-site dominance eroding at the edges.
For you, the Monday-morning priority remains the same one it was at the start of this year. Not whether to be in retail media. Whether your measurement setup can tell you, placement by placement, what you actually bought.