Loading

MediaBy Bait · · 11 min read

Retail media: how to invest in retailer networks without counting the same sale twice

Retail media is advertising sold by retailers based on their own purchase data. Formats, buying models, incremental measurement and governance.

Display platforms connected by a violet ribbon to a central structure.

Retail media is advertising sold by a retailer, on and off its own channels, using its customers' purchase and browsing data to target and measure campaigns. A retail media network brings together a retailer's inventory, its data and its buying and reporting tools. Commerce media is the broader concept: it includes companies that are not traditional retailers but also hold transaction data, such as marketplaces, delivery apps, travel platforms and financial services.

For large advertisers, especially manufacturers that sell through retailers, retail media touches three budgets at once: media, trade marketing and the commercial negotiation with the network. Each team tends to read results from its own dashboard. Without common rules, the same sale is credited to a promotion, a campaign and a commercial agreement, and the sum of the reports ends up larger than the brand's real growth at that retailer.

Formats: on-site, off-site and in-store

Inventory is divided by where the ad appears, and each environment has a different relationship with purchase intent.

  • On-site: ads within the retailer's website or app. This includes sponsored products in search results, banners on category pages, brand stores and placements on the product page. It sits close to the purchase and receives a lot of credit in attribution.
  • Off-site: campaigns run outside the retailer's channels, on social media, video, CTV or programmatic, using audiences built from purchase data. It reaches people who are not searching for the product at that moment.
  • In-store: screens, audio and media placements in stores, plus offers tied to loyalty programs. Measuring who was exposed is harder than in digital.

Organize the plan by role: protecting presence on the digital shelf, winning category buyers or generating demand for a launch. Evaluating all inventory against the same ROAS favors the formats that capture purchases that were going to happen anyway.

How retail media buying works

  • Self-service: the advertiser or agency operates directly in the retailer's ad platform, with keyword or category auctions. It provides control and speed, and requires a team that understands the platform's logic.
  • Managed service: the network's team plans and runs the campaign from a brief and a negotiated package. It simplifies operations and reduces visibility into delivery, so the contract needs to define what will be reported.
  • Programmatic: retailer audiences are activated in demand-side platforms (DSPs) or in private deals with publishers. This allows the same frequency control and verification used for the rest of the media, as long as the retailer makes the data available in that format.

No model makes up for a product page with incomplete information or no stock. Commercial execution is part of the results hypothesis.

Trade and media: the tension in the joint business plan

Many manufacturers negotiate a joint business plan with major retailers that combines sales targets, commercial terms, promotional activity and media investment in the network. Part of the value that used to be discounts or in-store visibility becomes media. The accounting source changes, but the question about results stays the same.

The conflict appears when trade and media evaluate the same money by different criteria: trade looks at sell-out and compliance with the agreement, media looks at reach and ROAS, and the retailer reports sales attributed by its own model. Three readings, each correct by its own rules, can describe the same growth.

To avoid counting the same sale twice, establish before the plan:

  • A single owner of the results number at the retailer, usually sell-out measured by an agreed source, rather than the sum of the dashboards.
  • The separation between what is media (bought with a campaign objective and metric) and what is a commercial term (paid for the relationship with the network, with no expectation of a measured causal effect).
  • Crediting rules: price promotions, media and in-store activations that happen at the same time split the result by an explicit method, or are tested in different periods and stores.
  • Each team's target expressed in metrics that do not overlap, so that no one is rewarded for the same sale.

What should advertisers look for in commerce media measurement?

The first point is what goes into the attributed sales figure. The IAB and MRC retail media measurement guidelines, published in January 2024, recommend reporting by SKU (the code for each item sold) and transparency about halo products, items from the same brand that were not in the ad but are included in the results. Ask which items were counted and whether sales of other brands from the same group were included.

  • Attribution window: the period after an exposure or click during which a purchase is credited. The guidelines call for the window to be disclosed before the campaign and alongside the results.
  • Exposure counted: viewable impressions, clicks or any served impression.
  • Extrapolation: the share of attributed sales that was estimated by a model rather than observed, for example because of cash payments or unidentified customers.
  • New to brand: the definition of a new buyer and the period used to define it, which the retailer should disclose.
  • Returns and cancellations: whether they are deducted.
  • Media coverage: whether the model accounts for other media the brand used during the period. The guidelines recommend disclosing when other media were left out.

How is commerce media improving transparency in measurement and reporting?

The improvement comes from common standards and from data access in controlled environments. The IAB/MRC guidelines address data quality, viewability (whether the ad had a chance to be seen), invalid traffic, attribution, incrementality and in-store measurement. They also recommend that networks give advertisers a minimum dataset for marketing mix models, with market, format, period, impressions, clicks, campaign, audience, device and cost.

The second driver is data clean rooms. A data clean room is an environment in which two companies match customer data without either one seeing the other's individual records, with rules on which queries are allowed and the minimum size of the groups returned. The IAB Tech Lab guide to data clean rooms describes the principles, advertising use cases and limitations of these environments. For the advertiser, a clean room makes it possible to match exposure on the network with its own customer base or with sales in other channels.

Transparency depends on the contract, not the technology. Negotiate the fields, granularity and frequency of the data in writing before signing the plan.

How does commerce media prove incremental ROAS?

Attributed sales are not incremental sales. Attributed ROAS divides credited revenue by investment. Incremental ROAS divides by investment the revenue that would not have happened without the media. In on-site formats the gap tends to be large, because some of the people who click on a sponsored product were already shopping the category.

The IAB/MRC guidelines describe four families of methods: experiments with a randomized control group, synthetic controls, tests across comparable markets or stores, and machine learning models. The most robust is the control group: part of the eligible audience is randomly held out from seeing the ads, and the difference in purchases between the groups estimates the effect. Ask the retailer for the test design, group sizes, holdout period and limitations, as the guidelines recommend.

Read the result in contribution margin, not revenue. The article on incremental media measurement details test designs and how to interpret inconclusive results.

Hypothetical example: three reports for the same growth

The following figures are illustrative and do not represent market or client data. A beverage manufacturer sells R$ 10 million per quarter at a major retailer. The following quarter, it sells R$ 12 million: growth of R$ 2 million. During that period, it invested R$ 600,000 in retail media and ran a price promotion negotiated by the trade team.

The retailer's dashboard attributes R$ 2.4 million in sales to media, a ROAS of 4. Trade credits the R$ 2 million in growth to the promotion. Added together, the reports explain R$ 4.4 million, more than double the real growth, because the dashboard also credits purchases by people who already bought the brand.

A test with a randomized control group estimates the media's incremental revenue at R$ 900,000, with an interval of R$ 500,000 to R$ 1.3 million. The central incremental ROAS is 1.5. With a 40% contribution margin, the media generated R$ 360,000 in margin for R$ 600,000 invested. The remaining R$ 1.1 million of growth is attributed to the promotion and other factors, to be separated by another method. The conclusion is not to cut the network, but to review formats and investment level and measure again.

Data governance and the LGPD

In a retail media campaign that uses the advertiser's data, both companies process personal data and need to define who makes decisions about the processing. Brazil's General Data Protection Law (LGPD) defines the controller as the party that makes decisions about processing and the processor as the party that processes data on the controller's behalf. This division of roles needs to be in the contract, along with the legal basis for each use.

The law also states that anonymized data is not personal data, unless the anonymization can be reversed with reasonable effort. A hashed email, for example, can still be personal data if the other party is able to link it to a person. Treat clean room matching as processing of personal data until your legal team says otherwise. Governance is covered in more detail in marketing data governance.

Checklist for evaluating a retailer's network

  • What share of category sales goes through this retailer, and how many identified buyers the network reaches.
  • Which on-site, off-site and in-store formats exist, and which buying models are available.
  • Which definitions of attributed sales, window, new to brand and halo the network uses, and whether it discloses them before the campaign.
  • Whether the network offers incrementality tests with a control group and documents the method.
  • What data the advertiser receives, at what granularity, and whether there is clean room access with useful queries.
  • Whether there is independent verification of delivery, viewability and invalid traffic in off-site formats.
  • How the contract handles controller and processor roles, retention and use of the advertiser's data.

Common retail media mistakes

  • Adding up attributed sales from retailers, platforms and promotions as if they were total results.
  • Accepting each network's default attribution window, which makes results impossible to compare.
  • Concentrating budget in sponsored search because it has the highest attributed ROAS, without measuring incrementality.
  • Leaving retail media out of media budget allocation, as if it were trade money under a different rule.

Frequently asked questions about retail media

What is the difference between retail media and commerce media?

Retail media is media sold by retailers based on their own purchase data. Commerce media extends the concept to other companies with transaction data, such as marketplaces, delivery apps, travel platforms and financial services. The measurement logic is the same: purchase data used to target and measure, with the same care around attribution and incrementality.

Should retail media come out of the trade budget or the marketing budget?

The accounting source does not define the objective. Every unit of spend needs an owner, a metric and a crediting rule that does not overlap with another team's, with a single results dashboard.

Are attributed sales the same as incremental sales?

No. Attributed sales are those the retailer's model credits to people exposed to the ad, including customers who would have bought anyway. Incremental sales are the difference compared with a scenario without the media, estimated by a control group test or an equivalent method.

What attribution window should you use for retail media?

There is no universal window. It should reflect the category's purchase cycle and be the same across networks, so that results are comparable.

To bring retail media into your planning, explore our Media practice and the discussion of media budget allocation.