A retail media network (RMN) is what happens when a retailer monetizes its audience. Instead of simply selling products, the retailer also sells access to the consumers who shop on its properties — to the brands that want to reach those consumers at the moment of purchase intent.
The fundamental insight behind retail media is that retailers possess something uniquely valuable: verified purchase data tied to identifiable shoppers, at scale, across categories. Amazon knows not just what consumers browse but what they buy, when they buy it, how often, and at what price sensitivity. Walmart knows in-store purchase behavior from hundreds of millions of loyalty transactions. This data asset, applied to advertising targeting and measurement, creates a class of media with closed-loop attribution that general digital advertising cannot replicate.
How a Retail Media Network Works
The basic mechanics: a brand pays the retailer for ad placements on the retailer's digital properties (website, app, search results). The retailer uses its first-party shopper data to target the brand's ads to relevant consumers. When a targeted consumer purchases the advertised product, the retailer attributes the sale to the ad and reports it as attributed revenue.
This creates a closed loop: ad exposure → consumer behavior → purchase → attribution — all within a single retailer's ecosystem. The retailer controls the full measurement chain, which is the source of both retail media's advantage (more accurate attribution than cookie-based digital) and its limitation (each retailer measures by its own rules, making cross-platform comparison difficult).
Major Retail Media Networks
| Network | Retailer | Region | Attribution Window | Ad Revenue (est.) |
|---|---|---|---|---|
| Amazon Ads | Amazon | Global | 14 days | $50B+/yr |
| Walmart Connect | Walmart | US, Canada | 30 days | $3–4B/yr |
| Criteo Commerce | Multi-retailer | Global | 30 days | $2B+/yr |
| Kroger Precision Marketing | Kroger / 84.51° | US | 14–30 days | $1B+/yr |
| Instacart Ads | Instacart / Maplebear | US, Canada | 14 days | $800M+/yr |
| Target Roundel | Target | US | 14 days | $600M+/yr |
| Tesco Media and Insight | Tesco / dunnhumby | UK, Europe | 14–28 days | ~£300M/yr |
| Carrefour Links | Carrefour | Europe, LatAm | 14–30 days | Growing |
| CVS Media Exchange (CMX) | CVS Health | US | 14–28 days | ~$300M/yr |
| Albertsons Media Collective | Albertsons | US | 14–30 days | Growing |
| Home Depot Orange Apron Media | Home Depot | US | 30 days | Emerging |
| Best Buy Ads | Best Buy | US, Canada | 14–30 days | Emerging |
These are the major established networks. The global landscape includes 277+ additional networks ranging from regional grocery chains and specialty retailers to marketplace operators. Each operates independently with its own attribution methodology.
The Core Problem: No Common Measurement Standard
Every retail media network measures performance by its own rules. There is no industry standard for attribution window length, model type, or which interactions count. The result: a brand managing campaigns across five networks is operating in five separate measurement systems whose outputs cannot be compared without adjustment.
Retailers have no commercial incentive to adopt a common attribution standard. More generous attribution windows and view-through inclusion make their ad product's reported ROAS look better relative to competitors. Any retailer that voluntarily adopted stricter measurement would appear to underperform. The industry has discussed standardization for years; it has not materially advanced because the incentive structure works against it.
Attribution window differences alone create 15–35% ROAS divergence between otherwise identical campaigns on different networks. A brand that compares Walmart's 30-day ROAS against Amazon's 14-day ROAS without normalization will consistently overestimate Walmart's efficiency — and misallocate budget accordingly.
On-Site vs. Off-Site Retail Media
On-site retail media
On-site placements appear within the retailer's owned digital properties: sponsored search results, product detail pages, homepage banners, category page carousels. These placements reach consumers who are already on the retailer's platform with active shopping intent, making them high-conversion and directly attributable.
On-site retail media is the original and still-dominant form. Sponsored Products, Sponsored Brands, and similar formats are all on-site products. Attribution is relatively clean: the ad shows on the retailer's platform, the purchase happens on the retailer's platform, and the retailer can directly observe both events.
Off-site retail media
Off-site retail media extends the network's reach to external publisher inventory — display ads on news sites, social media, streaming TV — targeted using the retailer's first-party audience data. Amazon DSP and Walmart's Trade Desk partnership are examples. The retailer's audience data travels to external environments; resulting purchases on the retailer's platform are attributed back to the external ad exposure.
Off-site retail media is the fastest-growing segment. It allows retailers to monetize their audience data beyond their own digital inventory, dramatically expanding the addressable advertising opportunity. For agencies, off-site campaigns introduce additional attribution complexity: the exposure happened off the retailer's platform, requiring cross-device identity resolution to link the external impression to the on-platform purchase.
The Agency View: Managing Multiple RMNs
Mid-market retail media agencies typically manage 3–8 client brands across 2–5 retail media networks simultaneously. The weekly workflow involves pulling performance reports from each network, attempting to normalize the figures for cross-platform comparison, and making budget allocation recommendations.
Without systematic normalization tooling, this process involves ad hoc spreadsheet adjustments that are inconsistent, time-consuming, and not defensible when clients ask about methodology. The planner who says "I apply a rough 20% haircut to Walmart because of the longer window" is making an approximation where a systematic correction is needed.
The proliferation of retail media networks — each with different interfaces, export formats, and attribution methodologies — is the structural driver for cross-platform normalization tools. As brands and agencies add new networks (Carrefour, Tesco, Instacart, CitrusAd) alongside the established ones, the normalization problem compounds in complexity and consequence.
Frequently Asked Questions
Amazon Ads is by far the largest retail media network globally, generating an estimated $50 billion or more in annual advertising revenue as of 2025. It is larger than all other retail media networks combined. Walmart Connect is the second largest in the US, growing rapidly but representing approximately 6–7% of Amazon's scale. Criteo, operating as a third-party retail media technology layer across many retailers, is another major player at the global level.
The critical difference is closed-loop attribution using verified purchase data. Traditional digital advertising (Google Display, Meta) can show an ad to a consumer but must rely on probabilistic attribution to connect the impression to a purchase — inferring that a consumer who saw the ad and later visited the advertiser's website converted because of it. Retail media networks know definitively whether a targeted consumer purchased, because the purchase happens within the retailer's own checkout system. This makes retail media attribution more accurate for e-commerce sales measurement, though cross-platform comparison remains problematic due to methodology differences between networks.
No. Retail media networks operate as independent, competing data silos. Amazon does not share purchase data with Walmart; Walmart does not share with Kroger. Each retailer's first-party data is a competitive asset, and sharing it with other retailers would undermine the differentiated value of their advertising platform. This fragmentation is a fundamental feature of the retail media landscape, not a temporary problem that will be resolved by industry coordination. It is the core reason why cross-platform normalization must happen externally, at the agency or advertiser layer, rather than at the network level.
RetailNorm is built for agencies managing campaigns across multiple retail media networks. Upload exports from Amazon, Walmart, Criteo, and other networks and receive normalized ROAS figures — corrected for window length, attribution model, and view-through differences — in a single comparable report.