Retail media has been the dominant story in digital advertising for the better part of a decade. Amazon built its sponsored listings product in 2012, confirmed that transaction data could power closed-loop advertising measurement, and launched an arms race that now encompasses Walmart Connect, Target Roundel, Kroger Precision Marketing, and hundreds of smaller retail networks. U.S. advertisers spent $60.32 billion on retail media in 2025 and are projected to spend $71.09 billion in 2026.
But a parallel category has been quietly expanding, and the two are frequently confused. Commerce media and retail media share a foundation in first-party transaction data, but they are structurally different models with different operators, different inventory, different targeting logic, and different attribution approaches. Getting the distinction right changes how advertisers allocate budgets, how operators build monetization infrastructure, and how brands think about where in the customer journey they can most effectively intervene.
Rokt’s 2026 breakdown of commerce media vs. retail media offers one of the clearest working frameworks available. What follows draws on that analysis and the broader industry data that surrounds it.
The Origins: A Media Asset That Predates Digital
The logic behind both models is older than the internet. Macy’s and Sears were charging manufacturers for prime floor placement and catalog position by the early 20th century because retailers controlled access to buyers that brands couldn’t reach any other way. The transaction relationship was functioning as a media asset long before digital gave it a formal name.
Loyalty programs converted that informal arrangement into structured data. Grocery cards, frequent flyer numbers, and hotel points schemes all did a version of the same thing: attach an identifier to a customer, track what they bought, and sell that profile to a brand that wanted to reach someone with those verified purchase characteristics. Airlines discovered their customer records were worth money to hotel groups and financial services companies. Grocery chains found CPG brands would pay for visibility into actual basket behavior rather than demographic estimates.
Amazon crossed a critical line when it wired this logic to real-time targeting and closed-loop measurement. A brand could prove return on ad spend against verified purchases, not probabilistic models. Retail media emerged as a distinct category. Other businesses with large verified transaction histories, banks, airlines, ticketing platforms, recognized the same structural opportunity. Commerce media developed as the term for what happens when that logic extends beyond the retail shelf.
Retail Media: The Bounded Model
Retail media, as the category is commonly understood, refers to advertising sold by retailers, running on retailer-owned inventory, and powered by that retailer’s first-party shopping data. The canonical operators are Amazon Advertising, Walmart Connect, Target Roundel, and Kroger Precision Marketing. The retailer owns the website, the app, and the transaction history of every customer who has ever shopped there. It monetizes that asset by selling ad placements to brands seeking to reach those customers at or near the point of purchase.
The inventory typically sits where shopping happens: sponsored product listings in search results, banner placements on category pages, display units in the checkout flow. Some retail media networks have extended into off-site placements, activating retailer data against programmatic inventory. The underlying logic stays constant. The retailer’s purchase data is the targeting asset; the retailer’s owned platform generates the ad opportunity.
Signal quality is the practical case for retail media. A grocery chain running an RMN knows that a specific customer bought pasta last Tuesday, holds a loyalty tier signaling weekly shopping frequency, and is currently searching for olive oil. That targeting reflects what someone has actually purchased, not what demographic bucket a third-party probabilistic model assigned them to. That precision helps explain why retail media consistently outperforms standard display.
The concentration of spending within the category, however, tells its own story. Amazon holds roughly 79.7% of the U.S. retail media market, per EMARKETER. Walmart Connect ranks second with approximately 8% share. Together, those two platforms will absorb 89% of all incremental retail media spending in 2026, leaving a small and shrinking fraction for every other RMN in the ecosystem. For brands allocating budgets across smaller retail networks, the math has become harder to justify on performance alone.
Commerce Media: The Horizontal Expansion
Commerce media can be understood as the expansion of transaction-data-powered advertising beyond the retail shelf. As Rokt’s framework describes it: retail media is a vertical application of the broader commerce media model, while commerce media is the horizontal layer that extends transaction-driven advertising to any business that processes purchases at scale and owns a direct customer relationship.
The distinction matters because it dramatically expands the pool of potential operators. A bank sees every dollar its cardholders spend across every category, every retailer, and every geography. Chase Media Solutions, which launched in 2024, reaches 80 million customers with retailer-agnostic cross-merchant purchase history. A financial institution doesn’t need a product catalog or a shopping destination. It needs verified spending patterns, and those it has in abundance.
A food delivery platform knows order frequency, cuisine preferences, average spend per order, and neighborhood. Instacart’s advertising-and-other revenue reached an annual run rate above $1 billion in 2025. Uber runs a comparable network across rideshare and delivery. PayPal’s advertising business draws on billions of real-time transactions from over 430 million active accounts. Marriott International launched Marriott Media in June 2025, backed by 237 million Bonvoy loyalty members and more than 200 targetable audience attributes. United Airlines debuted Kinective Media, positioning itself as the airline industry’s first dedicated media network.
A consistent pattern across all of them: large transaction volumes plus direct customer relationships equal a monetizable data asset. The prerequisite for commerce media is less about owning a product catalog and more about owning verified purchase history at scale.
Where the Two Models Actually Differ
The practical differences between retail media and commerce media come down to five specific areas, and they are consequential for how both operators and advertisers should think about them.
Who can operate
Retail media requires a retailer with a shopping destination, a product catalog, and audience density at sufficient scale to build statistically meaningful audience segments. Commerce media extends that opportunity to any high-frequency transactional business: banks, travel platforms, ticketing companies, subscription services, and telecommunications companies all qualify. That broader eligibility is why commerce media is expanding rapidly even as retail media’s incremental spending concentrates around its two dominant players.
Where inventory lives
Retail media inventory sits primarily on retailer-owned properties: the website, the app, in-store digital screens. Commerce media reaches inventory that retail networks typically don’t touch: order confirmation pages, payment screens, post-purchase receipt environments, and in-app surfaces across airlines, delivery platforms, and financial services. These are high-intent touchpoints that exist entirely outside the retail browsing context.
What the targeting signal covers
Retail media uses purchase and browsing signals within one retailer’s closed ecosystem. A sporting goods retailer knows a customer bought running shoes at their store. Commerce media, when operated across a network of transaction partners, can combine signals from multiple purchase contexts without any personally identifiable information transferring between operators. A customer who bought running shoes, registered for a road race, and subscribed to a nutrition service presents a far richer behavioral signal than any single retailer’s dataset could generate independently.
EMARKETER principal analyst Sarah Marzano has described this distinction directly: financial services operators bring purchase data that is cross-merchant and therefore more expansive, while retailers have granular SKU-level data within their own walls. Neither signal is categorically superior. They answer different questions.
How attribution closes
Retail media closes the attribution loop within the retailer’s environment. A customer who clicks a sponsored listing and completes a purchase in the same session generates clean, direct attribution. Commerce media can measure across a broader transaction ecosystem, confirming whether a customer who encountered an ad at one transaction touchpoint completed a qualifying purchase subsequently. The closed-loop model uses verified purchase events on both sides of the measurement rather than probabilistic modeling.
SKU-level vs. ecosystem-level outcomes
Retail media traditionally delivers SKU-level attribution: confirmation that a specific product sold as a direct result of a specific ad. Commerce media can enable ecosystem-level attribution, measuring the long-term customer value generated from an acquisition that happened at or immediately following a transaction, such as a new credit card sign-up or an annual subscription. Those downstream outcomes often represent substantially more economic value than the immediate ROAS from a single product sale.
Non-Retail Commerce Media Is Growing Fast
Financial services has been the fastest-growing segment of commerce media in 2024 and 2025. Chase Media Solutions, Klarna’s buy-now-pay-later advertising product, and PayPal’s media network all launched or significantly expanded during this period, each built on the same premise: transaction data that spans multiple retailers and categories generates targeting capability that no single RMN can match.
Travel followed a closely parallel arc. The logic is well-suited to the category: someone booking a transatlantic flight has signaled income range, travel frequency, and destination intent simultaneously. Kinective Media by United Airlines and Marriott Media both launched with explicit ambitions to position their networks as standalone advertising revenue streams rather than supplementary features.
Ticketing and live events complete the picture. A customer purchasing concert tickets reveals genre preferences, geography, and spending appetite with precision that demographic targeting can’t approach. The verified purchase intent in that transaction is, in many ways, the cleanest possible signal for adjacent advertisers.
EMARKETER’s January 2026 FAQ on commerce media growth noted that U.S. advertisers will spend $71.09 billion on retail media in 2026, but described the real expansion opportunity as lying in non-retail commerce media networks gaining ground on smaller RMNs. The challenge the same analysis identified is measurement standardization: commerce media networks are earlier in their measurement maturity than the established RMN ecosystem, and proving incrementality across multi-network campaigns remains genuinely difficult.
The Privacy Architecture
One underappreciated advantage of commerce media built on first-party transaction data is its structural relationship to privacy regulation. The data underlying most commerce media networks was collected directly by the operator from its own customers: purchase records, transaction histories, and in-app behavior that has never touched third-party tracking infrastructure.
Where commerce media networks combine signals from multiple operators, they typically do so through data clean rooms, secure environments where audience matching and targeting logic run on aggregated data without individual customer records being transferred to any advertiser. Brands get the benefit of cross-network purchase signals. The operator’s customer data and the advertiser’s CRM are never directly exposed to each other.
This architecture is one reason financial services companies operating under GDPR, CCPA, and sector-specific regulations have been able to participate in commerce media without creating compliance exposure. The privacy protection is a technical constraint built into the targeting execution rather than a policy overlay applied afterward.
How Rokt Positions Within the Commerce Media Category
Rokt, the New York-based e-commerce technology company, has developed one of the more clearly defined frameworks for commerce media built around what it calls the Transaction Moment: the window spanning product selection through purchase confirmation. As Rokt’s analysis makes plain, this is the highest-intent interval in any e-commerce journey, the point at which a customer has already committed to buy and where targeting based on confirmed purchase signals rather than browsing inference can generate meaningfully different performance outcomes.
The Rokt Network processes more than 10 billion transactions annually across 33,000-plus active clients, spanning retail, airlines, ticketing platforms, food delivery, and financial services, reaching 1.1 billion unique customers globally. That cross-vertical transaction volume is precisely what defines commerce media’s targeting advantage over single-retailer retail media.
As Rokt’s Chief Commercial Officer Elizabeth Buchanan noted in Rokt’s January 2026 commerce media outlook, commerce media has reached an inflection point where brands and retailers are recognizing that scale without relevance leads to diminishing returns and frustrated consumers. The Silicon Review’s coverage of Rokt’s 2026 strategy described the company’s approach as one of the clearest articulations of how the transaction itself has become the most strategically valuable moment in digital commerce.
The performance data Rokt reports on its network reflects the Transaction Moment’s positioning: a 4.03% average click-through rate on Rokt Ads placements, which is ten times the Google Display benchmark and four times Facebook. The 6.32% average conversion rate on those placements reflects what happens when targeting is grounded in confirmed purchase behavior rather than probabilistic audience segments.
Rokt’s AI system, Rokt Brain, has been trained on 1.95 trillion data points across the network. Placement decisions draw on behavioral patterns from across the entire network, not just within one retailer’s isolated dataset, which is a structural advantage over any single-operator RMN. For e-commerce partners, Rokt’s product suite, including Rokt Catalog, Rokt Upcart, Rokt Pay+, and Rokt Thanks, monetizes existing transaction touchpoints without requiring the operator to build any advertising infrastructure independently. The revenue-sharing model returns $7 of every $8 of value generated back to the partner.
For advertisers, Rokt Ads provides access to verified transaction audiences across retail, travel, ticketing, delivery, and financial services simultaneously through a single buy. That cross-vertical access is a meaningful operational simplification over managing separate campaigns across individual RMNs.
What This Means for Advertisers Choosing Between the Models
For a brand deciding where to allocate spend, the relevant question is where in the purchase cycle the intervention needs to happen and what outcome it needs to drive.
Retail media is well suited to capturing existing demand. A customer searching for wireless headphones on Amazon or Walmart is actively in market. Sponsored listings at that moment reach maximum buying intent with direct SKU-level attribution. The proximity to conversion is tight and the measurement is well established.
Commerce media, particularly in post-transaction environments, is better suited to cross-sell, adjacent acquisition, and non-endemic advertising. Someone who has just confirmed a flight booking is in active planning mode, receptive to hotel offers, travel insurance, or car rental in a way they won’t be once the trip fades from immediate attention. Someone who just completed a grocery order is a plausible audience for adjacent subscription services. The confirmed purchase verifies willingness to transact; the ad reaches them while that intent is freshest.
Non-endemic advertising is where the structural difference becomes most commercially consequential. A grocery RMN excels at selling pasta to someone looking at pasta. It is not well suited to introducing a hotel chain or a financial product to that same shopper. Commerce media built across verticals can put a hotel offer in front of someone who has just purchased a flight, a financial product offer in front of someone who has just made a large retail purchase, and a subscription service in front of someone whose recent purchase history signals the right life stage. The adjacency logic is what retail media’s on-site context cannot replicate.
The two models are not mutually exclusive, and the most sophisticated advertising strategies in 2026 are likely to use both. Retail media captures existing demand within known shopping contexts. Commerce media extends brand reach into high-intent moments that retail inventory doesn’t touch. Understanding which moment you’re trying to win is how the allocation decision should be made.
