Retail Media: Guides, Comparisons and Tools

Retail media uses a retailer’s first-party relationships and shopping surfaces to sell advertising, with results connected to purchases. Compare platform choices, measurement definitions and operating economics.

Start with the business decision and the data flow. Identify who owns the customer relationship, who receives each signal, who runs the decisioning system, and what evidence supports the result. A vendor feature list alone cannot settle those questions.

This guide brings together the reference material for this topic. Comparisons describe different operating models, rather than declaring a universal winner. The linked pages retain source dates, vendor-claim labels and the limits of each example. Read those limits before applying a figure or architecture to your own organization.

• Onsite sponsored products, display, offsite, in-store. "Yes" means the vendor's own page names the format as a product. "Partner" means delivered through a named third party. "n/d" means not stated.
• Self-serve portal. A brand-facing console the retailer can white-label.
• Clean room. Whether the vendor offers a clean room or names a partner; the options themselves are compared in Clean room architectures compared.
• Measurement and incrementality. What the vendor says it measures; whether it has been certified against a standard. The standards are explained in Retail media measurement standards explained.
• Data residency and self-hosting. Whether the vendor states where data is processed and whether the retailer can run the stack itself.
• Pricing model. Only where public or stated by a cited source; vendor comparisons are marked as such.
• Notable customers. Only names that appear on the vendor's site or in a dated release; logos are reported as logos.

Read the full retail media network platforms compared: 18 vendors 2026 guide for context, sources and limitations.

Three business models, not eighteen. The first is demand bundled with the ad server: Criteo and Epsilon bring brands and agencies that already buy through them, charge a share of spend, and in return a mid-sized retailer gets revenue from week one. The second is software: Kevel, Topsort, Mirakl Ads, Moloco, Pentaleap, Osmos, Zitcha, Koddi and Vantage sell the engine and leave the retailer to sell, or to plug in demand through exchanges (Kevel's KRX, Koddi's SSP, Osmos's Audience Exchange). The third is another retailer's stack: Amazon Retail Ad Service and Instacart Carrot Ads give a smaller retailer the ad products, the advertiser base and the reporting of a giant, with the giant's data environment in between. Each model has a price the table cannot show: demand bundles cost margin, software costs a sales team, and a rival retailer's stack costs independence.

Formats are converging; measurement is not. Almost every vendor now lists onsite sponsored products, display, offsite and in-store. Only one, Moloco, has a certification against a published measurement standard, and that covers onsite display and sponsored products on web, mobile and in-app (IAB Europe, 7 October 2026). Most vendors describe incrementality as a feature without naming the method. The standards require the method to be disclosed, so ask; the questions are listed in Retail media measurement standards explained.

Nobody publishes data residency, and nobody offers self-hosting. Every commercial row reads "n/d" on where shopper data is processed and "No" on running the stack inside the retailer's own perimeter. For a US grocer this is a procurement question. For a European retailer with loyalty data, a pharmacy under health-privacy law or a bank-owned marketplace it is a legal one, which is why the self-hosted row exists. The trade-offs are set out in Retail media infrastructure: build vs buy.

The retailer platforms are the new entrants to watch. Amazon Retail Ad Service is still labelled beta, runs on an AWS-powered data environment and gives the retailer a console with fill rate, CPC, CTR and ROAS by advertiser and keyword (Amazon Ads, accessed 17 October 2026). Instacart's Carrot Ads picked up Gopuff on 22 September 2026 (Instacart, 22 September 2026). Both solve the demand problem that kills most tier-two networks. Both also mean the retailer's shopper signals and the advertiser relationships live with a competitor.

Read the full retail media network platforms compared: 18 vendors 2026 guide for context, sources and limitations.

Whatever the size, the RFP should ask the same thirteen questions the table asks, plus: which measurement standard the vendor reports against and whether it has been audited; who owns the advertiser relationship and the log-level data at termination; what the ad server returns to the site in milliseconds and at what request volume; and whether the vendor's offsite extension uses the retailer's own DSP seat or the vendor's. For the difference between a network and an ad network in this context, see Retail media vs ad networks.

Read the full retail media network platforms compared: 18 vendors 2026 guide for context, sources and limitations.

IAB/MRC Retail Media Measurement Guidelines (January 2024). Written by the IAB Retail Media Measurement Working Group with contributors from Mars, Albertsons Media Collective, Colgate-Palmolive, Walmart Connect, Google, Uber, TikTok and the Alliance for Audited Media, among others (IAB/MRC, January 2024, acknowledgements). Six chapters: data quality, audience measurement, delivery and viewability, incrementality, reporting, and in-store environments. It binds "retail media measurement organizations," meaning the network or the vendor that measures for it.

IAB and IAB Europe In-Store Retail Media: Definitions and Measurement Standards (3 December 2024). Developed by IAB Retail Media Committee members, the Retailer Council and the working group, jointly with IAB Europe. It covers store zones, media, ad-serving, impression and sales metrics for digital screens, audio and connected shopping; printed and experiential formats are deferred (IAB, 3 December 2024).

IAB Europe Commerce (incl. Retail) Media Measurement Standards V2.1 (January 2026). First published in April 2024 as the Retail Media Measurement Standards, updated after a public comment period from September to November 2025, renamed to cover commerce media, and extended with a measurement funnel, gross and net sales definitions, quick commerce metrics and a formal incrementality definition. Networks may comply with V1 or V2 until the end of July 2026; after that only V2 applies (IAB Europe, V2.1). The standards are tied to IAB Europe's certification program, which the document describes as the only way to guarantee adherence.

Two supporting documents matter: the IAB and IAB Europe Guidelines for Incremental Measurement in Commerce Media (3 November 2025), which V2.1 makes binding on any network that offers incrementality (IAB, 3 November 2025; IAB Europe, V2.1), and IAB Europe's in-store V2 and travel proposals of 17 September 2026, open for comment until 23 October 2026 (IAB Europe, 17 September 2026).

Read the full retail media measurement guidelines explained: iab, mrc, eu guide for context, sources and limitations.

The US document sets no lookback window. It says the window "must be empirically supported" and "logical with regard to campaign objectives and length as well as category sales cycles," and it notes that IAB and MRC may later prescribe standard windows for product categories (IAB/MRC, January 2024, section 3.6). What it does demand is disclosure: whether an attributable sale rests on a viewable impression or on extrapolation; the time-decay curve and the research behind it, revalidated at least annually; how sales from overlapping campaigns for the same brand are separated; and, at a minimum, SKU-level reporting with halo products shown transparently. The outcomes table defines new-to-brand and new-to-category as shoppers with no purchase of the brand or category in a time frame "that needs to be disclosed by the retailer," and marks before-and-after category share analysis as directional only (IAB/MRC, January 2024, outcomes table).

IAB Europe V2.1 fills the gap the US left. ROAS is "typically based on a last click or view model," and networks "should provide a split between click-based and view-based ROAS." The default lookback is 30 days post-view and post-click for endemic onsite and offsite, with the explicit example that an FMCG brand might ask for 7 days and a consumer-electronics brand for 30; off-site partners that cannot support custom windows must say so in reporting (IAB Europe, V2.1). For quick commerce the default drops to 7 days. Gross sales are assumed unless the network states net.

Read the full retail media measurement guidelines explained: iab, mrc, eu guide for context, sources and limitations.

The US chapter 4 and the November 2025 joint guidelines agree that incrementality is "the potential causal impact of marketing," measured against a baseline that must be estimated without bias (IAB/MRC, January 2024, chapter 4). The November 2025 guidelines group methods into experiments, model-based counterfactuals, econometric models and hybrid proxies, and rest on three principles: credible counterfactuals, control of bias, and separation of signal from noise (IAB, 3 November 2025). The glossary entries for incrementality and iROAS carry the short definitions.

Read the full retail media measurement guidelines explained: iab, mrc, eu guide for context, sources and limitations.

Onsite. Monthly visits times pages per visit gives pageviews. Pageviews times ad slots per page gives ad opportunities. The monetizable share removes checkout, account and policy pages. Sponsored product slots are priced per click: impressions times click-through rate times CPC. Display slots are priced per thousand: impressions times fill times CPM. The point of splitting them is that the two lines move differently; sponsored product revenue follows search traffic and CPC, display follows fill and CPM.

In-store. Stores times screens per store times plays per day times opportunity-to-see per play gives monthly OTS. Sell-through and CPM turn that into revenue. The IAB and IAB Europe in-store standard, finalized in December 2024, defines what counts as an impression and is the reference for the OTS input (Pygmalios, updated 28 July 2026, citing the IAB standard). Retail media measurement standards explained covers the onsite and offsite rules.

Offsite. Loyalty members times match rate gives matched shoppers. Times frequency gives impressions, times CPM gives gross media. The retailer keeps a take rate; the rest is media cost paid to publishers or a DSP. Gross offsite media inflates "gross billings" without inflating margin, which is why the calculator reports both gross and net.

Costs. Platform fee as a share of net revenue or a fixed annual amount. Team as FTEs times loaded cost. Other costs (measurement, clean room, creative services) as a single line. Contribution is net revenue minus those three. The calculator does not model consumer incentives because a retailer's sponsored placements do not carry them; the commerce media template does.

Ratios. Gross ad revenue over annual GMV, net revenue over GMV, onsite-only revenue over GMV (the "ceiling" line), and contribution margin on net revenue.

Read the full retail media economics calculator: revenue and % of gmv guide for context, sources and limitations.

The defaults put a $5 billion retailer at 0.69% of GMV with 79% margin on net revenue, which sits between Walmart's 0.9% and the vendor-described 0.5% plateau. The margin is high because the example carries no media pass-through except offsite, and no consumer incentives; the five KPIs for a retail media network explain which of these lines advertisers will push on first.

Read the full retail media economics calculator: revenue and % of gmv guide for context, sources and limitations.

The calculator plots four lines on one axis: your onsite-only ceiling, your total, and three references.

Why the plateau exists is arithmetic, not strategy. Onsite revenue is bounded by pageviews times slots times a CPC that has sat near a dollar for seven years (Skai, 28 July 2026). Once sell-through on search and category pages is high, the only ways up are more slots per page (which advertisers and shoppers resist), more traffic (which the media business does not control), or new inventory: in-store screens and offsite audiences. The calculator makes that visible by showing the onsite ceiling separately. If the ceiling is 0.5% and the target is 2%, the gap has to come from the two other lines, and the inputs for those (screens, sell-through, match rate, frequency) are the ones to challenge. Whether to build or rent that extra inventory is the subject of retail media infrastructure: build vs buy and the retail media platforms comparison.

Two cautions. The 0.5% and 3 to 6% figures come from a vendor that sells against the plateau, so treat them as a framing, not a benchmark. And Walmart's ratio is ad revenue over total revenue, while Instacart's is over GTV; the calculator labels your ratio by the denominator you enter.

Read the full retail media economics calculator: revenue and % of gmv guide for context, sources and limitations.

The table is downloadable as CSV and JSON under CC BY 4.0. Rows are added, never silently changed; superseded rows keep their date.

Read the full retail media economics calculator: revenue and % of gmv guide for context, sources and limitations.

What is a retail media network platform? It is the software that lets a retailer sell and serve ads on its own site, app, emails and stores, target them with its shopper data, and report sales back to brands. Platforms come as demand bundles (Criteo, Epsilon), as software (Kevel, Topsort, Moloco, Mirakl Ads, Pentaleap, Osmos), or as another retailer's service (Amazon Retail Ad Service, Instacart Carrot Ads). The retail media hub collects the related guides.

Which retail media platform does Walmart use? Its own. Walmart Connect runs on in-house technology, uses The Trade Desk as a DSP partner for offsite, and has added owned screens and buying tools through acquisitions: Vizio for about $2.3 billion on 3 December 2024 and Vibe.co, completed 4 August 2026 (Walmart, 3 December 2024; Vibe.co, 4 August 2026; Tinuiti, 18 May 2026). Kroger similarly runs on its 84.51° unit (Tinuiti, 18 May 2026).

How much does a retail media platform cost? Vendors do not publish price lists. The models that are public or reported are: a percentage of media spend plus platform fees (Criteo), SaaS without revenue share (Kevel), a per-spend or platform fee (Osmos), pay-for-outcomes (Rokt) and, in one Moloco customer's case, cost per order (Osmos, 4 June 2026; Rokt; Moloco). Ask for the effective take rate at your expected spend, including any demand fee, and compare it with the glossary definition of take rate.

Criteo vs Kevel: what is the difference? Criteo sells a network: its Commerce Max DSP brings brand and agency demand, including Google Search Ads 360 buyers in beta since 10 September 2025, to "over 200" retailers, and it charges a share of spend (Criteo, 10 September 2025; Osmos, 4 June 2026). Kevel sells the engine: ad-serving APIs, a self-serve console, audience tools and an exchange, on SaaS pricing without revenue share, used by Home Depot, Chewy and Dollar General among others (Kevel, accessed 17 October 2026; Osmos, 4 June 2026). The first is faster to revenue; the second keeps the advertiser relationship with the retailer.

Can a retailer build its own retail media platform? Yes, and the largest do: Walmart, Kroger through 84.51°, and Amazon, which now sells its stack to others (Tinuiti, 18 May 2026; Amazon Ads). Below that scale the honest answer is "build the parts that differentiate, license the rest": an owned clean room and data layer, an owned or white-label DSP for offsite, and a licensed or open-source ad server underneath. Vendor frameworks put the full custom build at $10 billion-plus GMV (Osmos, 4 June 2026); the sequencing is in Retail media infrastructure: build vs buy.

• Retail Media Measurement Guidelines Explained: IAB, MRC, EU: IAB/MRC retail media measurement guidelines, IAB Europe V2.1 and the in-store standards summarised: metrics, attribution windows, incrementality methods.
• Retail Media KPIs: Metrics, Definitions and Measurement: Retail Media KPIs: a practical guide to the definitions, operating decisions, measurement and evidence you need.
• How Retailers Turn Shopper Data Into Media Revenue: Retailers can earn media revenue by helping brands reach relevant shoppers and measure permitted advertising outcomes.
• First-Party Data Monetization for Retailers (No Data Sales): Shopper data monetization can mean selling advertising, measuring permitted outcomes or building a useful merchant offer program.

• Retail Media Network Platforms Compared: 18 Vendors 2026: Retail media network platforms compared: Criteo, Kevel, Topsort, Mirakl Ads, Epsilon, Moloco, Vantage, Koddi, Zitcha, Pentaleap, Osmos, Amazon, Instacart.
• How to Build a Retail Media Network: Build vs Buy 2026: How to Build a Retail Media Network: a practical guide to the definitions, operating decisions, measurement and evidence you need.
• Retail Media Network vs Ad Network: What Changed: A retail media network sells advertising using a retailer’s shopping relationships and surfaces.

• Retail Media Economics Calculator: Revenue and % of GMV: Model retail media revenue, fees and contribution from editable inventory and pricing assumptions. Compare scenarios as a share of GMV.

• Retail Media Launch: Case Evidence and Measurement: A retail media launch case needs a defined launch milestone, an operating scope and evidence of advertiser delivery and measurement.

1. IAB Europe, 7 October 2026
2. Amazon Ads, accessed 17 October 2026
3. Instacart, 22 September 2026
4. IAB/MRC, January 2024, chapter 4
5. IAB, 3 December 2024
6. IAB Europe, V2.1
7. IAB, 3 November 2025
8. IAB Europe, 17 September 2026
9. Pygmalios, updated 28 July 2026
10. Skai, 28 July 2026

• October 18, 2026: aligned the search title and heading, expanded the source-linked guide and reviewed its page groups.
• October 17, 2026: assembled this guide from the approved reference pages, retaining their source links and limitations.