Retail Media: What Brands Buy, What to Trust, When to Skip
%20(1).jpg)
Published:
October 31, 2025
Updated:
October 21, 2026
A year after The Trade Desk announced a deal with Shopsense AI to bring “contextual sponsored product ads” to publisher pages, there is no public sign the deal has scaled. Shopsense, a startup that matches products to articles and videos, does not appear in any of The Trade Desk’s 2026 quarterly releases, and in May 2026 it also joined Amazon Publisher Services’ Connections Marketplace.
The Trade Desk has had a hard year. Its revenue grew 3% in the second quarter of 2026, it guided to at least $650 million for the third quarter, roughly 12% below a year earlier, and chief executive Jeff Green said, “This quarter did not meet the standard we set for ourselves” (The Trade Desk). Its retail news in 2026 was more modest than shoppable publisher pages: Dollar General and the ad tech firm Kevel announced a way to buy Dollar General’s on-site and off-site ads in one place inside The Trade Desk, with a managed-service test from June and access for shared clients expected from the third quarter.
Retail media itself kept growing, and kept concentrating. US advertisers spent $60.3 billion on it in 2025, according to EMARKETER, which forecast about $71 billion for 2026. Amazon took 79.7% of the 2025 total, Walmart Connect 8.0% and Target’s Roundel 1.5%, and EMARKETER expects Amazon and Walmart to capture 89% of the growth in 2026. Globally, WARC forecasts $200.4 billion in 2026 and expects retail media to reach 15.2% of all ad investment in 2027 (Campaign).
For most brands the useful questions are what you are buying when you spend there, what the reports can prove, and whether you should be there at all.
What you buy on a retail media network
A retail media network sells advertising on a retailer’s own properties and, increasingly, elsewhere using the retailer’s shopper data. The products fall into four groups.
Sponsored products are the core. They are paid listings in the retailer’s search results and product pages, usually bought by cost per click in an auction. Amazon says Sponsored Products has “no monthly or upfront fees,” and on Amazon’s July 2026 earnings call Andy Jassy said Sponsored Products “continues to be our largest offering” (call transcript). Walmart, Target, Kroger and Instacart sell their own versions.
On-site display covers banners and video on the retailer’s site and app. Off-site media uses the retailer’s data to target people on other sites, social platforms and connected TV. Target’s Roundel sells off-site placements on Meta, Pinterest, Snapchat and CTV, and Walmart Connect works with partners including Spotify, Yahoo DSP and Warner Bros. Discovery. In-store media covers screens and placements in physical stores, such as Walmart’s store ads.
The model has spread beyond retailers. Delivery apps sell the same products, as our look at DoorDash’s ad business explains, and Amazon also sells its buying platform for inventory far from its own store, which our guide to AWS and Amazon advertising covers. Target sells creator campaigns through Roundel alongside its own social-first launches, as in Woolrich x Target.
Why the data is the product
A retailer knows what its customers bought. Roundel advertises “100M+ unique Target Circle members,” and Kroger Precision Marketing says its targeting is guided by purchase data “from over 62 million households.” That data does two jobs. It lets you target people by what they buy rather than what they browse, and it lets the retailer connect an ad to a sale in its own records, which the industry calls closed-loop measurement.
The data is what you are paying for even when you buy a search listing, because the listing sits at the moment of purchase. That position has a catch. A sponsored listing sits where an organic listing would have been, and some of its sales would have come through organic results anyway. In a study of an e-commerce platform that blocked sponsored search for 3% of visitors, Sarah Moshary found that sponsored search “steals business from organic listings,” and a Flipkart field experiment with about 1.2 million users found that in electronics, sponsored listings performed worse than the organic listings they replaced, while in clothing and footwear the effect reversed. When the organic listing you displace is your own, part of what you pay for is a sale you already had.
Off-site media is where the data matters most and is hardest to check. When a retailer uses its shopper data to find people on another platform, you are buying its audience definitions and its matching, and you see only what it reports back.
Where The Trade Desk and Amazon fit
Most retail media is bought inside each retailer’s own tools. The part independent buying platforms compete for is off-site media, where a retailer’s data finds people elsewhere. EMARKETER puts US off-site retail media at more than $17 billion in 2026, growing 29.5%, according to Marketing Dive.
The Trade Desk sells neutrality. On its first-quarter 2026 call, Green said retailers in its data marketplace account for more than 80% of sales at top US retailers (earnings call summary). But its exclusive arrangement with Walmart was revised in 2025 to let other platforms in (EMARKETER), and in June 2026 Walmart Connect added Yahoo DSP and Magnite in a closed test (Marketing Dive). Amazon sells scale and price: AdExchanger reported that Amazon’s DSP charges no fee on guaranteed deals for Amazon-owned media and 1% on open-web publishers, against a take rate of around 20% at The Trade Desk (AdExchanger). In September 2026 Amazon merged its DSP and its ads console under a new name, Amazon Ads Agent.
For a brand, the choice of platform matters less than the question underneath it: whose data are you buying, and can you check what it did? A retailer’s audience bought through an independent platform still comes with the retailer’s measurement, so ask for the same holdout either way.
What retail media measurement can and cannot prove
Start with what is independently audited. In retail media, the Media Rating Council’s accreditations cover ad delivery. Amazon’s Sponsored Products accreditation covers clicks, and Instacart’s covers impressions, clicks, click-through rate and viewable impressions. The MRC’s retail media decisions in the last three months of 2025 covered delivery metrics and no sales metric. The sales figures in your dashboard are the retailer’s own.
The IAB and MRC published retail media measurement guidelines in January 2024. They require lookback windows to be disclosed before a campaign runs, define incrementality as “the potential causal impact of marketing,” and recommend third-party audits. In November 2025 the IAB added incrementality guidelines covering experiments, modeled counterfactuals, econometric models and hybrids.
The distinction that matters most is between attribution and incrementality. Attributed sales count everyone who saw or clicked an ad and then bought within the window. Incremental sales count only the purchases the ad caused. Retailers do publish incrementality claims: Instacart says its campaigns produce a “15%+ incremental sales lift on average,” based on its own four-week tests ending in February 2023, and adds that “there can be no assurance” the results can be repeated. Marketers know the gap. NIQ found 67% of CMOs plan to increase retail media spending in 2026, while only 53% believe their networks give adequate incrementality measurement (NIQ).
So trust delivery numbers that are accredited. Treat attributed return on ad spend as a count of sales that followed an ad. Only a test tells you how many the ad caused. Treat “new-to-brand” as defined by the retailer’s lookback window, which you should ask for. And ask every network for a holdout test, in which a random group is kept from seeing your ads and its purchases are compared with the exposed group’s. If a network can’t run one, treat its sales figures as unproven until a geographic test of your own says otherwise.
When a smaller brand should skip retail media
Entry is cheap, which makes retail media easy to start and easy to overspend on. On Amazon, advertising revenue grew 26% in the second quarter of 2026 while third-party seller services grew 16%, which the Amazon agency BellaVix reads as ad costs rising faster than seller revenue, though Amazon’s ad line also includes Prime Video, DSP and vendor ads that third-party sellers don’t buy (Amazon). In Jungle Scout’s 2025 survey, 32% of Amazon sellers named growing advertising expenses as a concern.
Skip it, or wait, in four situations. If your product pages, reviews and stock aren’t in order, ads will buy traffic that doesn’t convert. If your margin can’t absorb the cost of a click on top of the retailer’s fees, every sale an ad touches can lose money. If most of your attributed sales come from people searching your brand name, you are probably paying for customers you already had, unless competitors are bidding on your name. And if you sell mostly through your own site, your budget may work harder in search and shopping campaigns you control, including the Shopping ads Google shows inside AI Overviews.
If you go in, start with sponsored products at the retailer that sells most of your volume, on category searches where you aren’t already the top organic result. Set a holdout or a geographic test before you scale. Move to off-site and display only when you can measure them. An online presence analysis shows where your site and search visibility stand before you pay for ads, ecommerce marketing covers the store and the campaigns that feed it, and performance marketing covers the paid search and social that compete for the same budget.
Retail media keeps adding screens
Shoppable content on publisher pages is one new surface. TV home screens are another: Walmart sells ads on the Vizio TVs it owns, using its shopper data (Marketing Dive). Device makers sell screen space too, though without a retailer’s purchase data, as our look at Samsung’s fridge ads shows. WARC’s Alex Brownsell has warned that retailers must grow ad revenue “without overwhelming shoppers with too many ad interruptions” (Campaign). For a brand the test stays the same on every new surface: what data is attached, what the seller can prove, and what a holdout would show.
Frequently Asked Questions
What is retail media?
Retail media is advertising sold by retailers and commerce platforms using their own sites, apps, stores and shopper data. It includes sponsored product listings in search results, display ads on the retailer’s site, off-site ads that use the retailer’s data on other platforms and connected TV, and in-store screens. Amazon Ads, Walmart Connect, Target’s Roundel and Kroger Precision Marketing are the best-known US networks.
How big is the retail media market?
EMARKETER estimates US retail media spending at $60.3 billion in 2025 and forecasts about $71 billion in 2026. Amazon held 79.7% of the 2025 US market, Walmart Connect 8.0% and Target’s Roundel 1.5%. Globally, WARC forecasts $200.4 billion in 2026, with retail media reaching 15.2% of all advertising investment in 2027. Forecasts vary with how each firm defines the market.
What happened to The Trade Desk’s Shopsense deal?
The Trade Desk announced an integration with Shopsense AI in October 2025 to bring contextual sponsored product ads to publisher content on the open web. There has been no public sign of it scaling: Shopsense isn’t mentioned in The Trade Desk’s 2026 results, and in May 2026 Shopsense also joined Amazon Publisher Services’ Connections Marketplace. The Trade Desk’s 2026 retail work has included a project with Dollar General and Kevel.
How do you measure retail media incrementality?
Use a holdout test. A random share of the audience is kept from seeing your ads, and its purchases are compared with those of people who saw them; the difference is the incremental effect. Some networks offer these tests, and the IAB published incrementality guidelines in November 2025. Attributed return on ad spend counts every purchase that followed an ad within the retailer’s window, whether or not the ad changed anything, so it cannot show what the ads caused.
Is retail media worth it for small brands?
It can be, if your listings convert, your margins can carry the cost of clicks, and you can test what the ads add. It is usually a poor use of money when most attributed sales come from your own brand searches, when product pages and reviews are weak, or when you sell mainly through your own site. Start with sponsored products at your biggest retailer and test before scaling.



