For much of retail media’s rapid expansion, the promise was compelling enough to outweigh the limitations.

Retailers possessed valuable first-party purchase data. Brands wanted access to shoppers close to the point of sale. Retail-media networks could connect an advertisement with a transaction inside a relatively closed environment and report a return on advertising spend that appeared more concrete than the proxies available across much of digital media.

The channel delivered something advertisers had spent years asking for: evidence that media exposure and product sales occurred within the same commercial system.

But a closed loop is not necessarily a transparent one.

As retail-media investment approaches an estimated $107.6 billion in the United States in 2026, buyers are asking more difficult questions about what reported performance actually represents. NIQ reports that 67% of chief marketing officers plan to increase retail-media investment this year, while only 53% believe their retail-media networks provide adequate measurement and attribution for reliable incrementality analysis.

The issue is no longer whether retail media can connect ads with sales.

The issue is whether those ads caused additional sales, shifted profitable customer behavior or merely claimed credit for purchases that were already likely to happen.

The useful question is not whether retail media can report a transaction. It is whether the investment changed the outcome.

Retail media’s original advantage is becoming its central tension

Retail-media networks emerged with a structural advantage over much of the open advertising market: retailers could use their own customer and transaction data to target audiences and measure purchases.

That advantage remains meaningful. A retailer may know that an exposed customer bought a particular product online or in a store. It may be able to connect campaign activity with product-level sales more directly than a conventional publisher or social platform.

But the same retailer may also control:

  • the audience data
  • the advertising inventory
  • the auction
  • the product placement
  • the transaction
  • the attribution methodology
  • the final performance report

That creates an inherent tension. The company selling the media may also be responsible for defining what counts as success.

Reported return on ad spend can vary based on attribution windows, treatment of organic sales, inclusion of in-store purchases, identity matching, view-through credit, new-versus-existing customers, promoted-product rules and whether the network reports gross sales or attempts to isolate incremental sales.

NIQ notes that retail networks do not calculate ROAS consistently and that changing methodologies can materially change the resulting number. It describes the use of attributed ROAS as a substitute for incrementality as one of the channel’s most common and costly measurement errors.

A brand may therefore receive attractive reports from several networks without being able to determine whether the results are comparable—or whether any of them represent growth the business would not have achieved otherwise.

Closed-loop attribution explains where a sale was recorded

It does not necessarily explain why the sale occurred.

A shopper may see a sponsored product, click it and purchase the item. That appears straightforward.

But the shopper may already buy the brand every week. The product may be on promotion. Distribution may have expanded. A television campaign may have increased demand. A creator may have introduced the product. The item may have moved higher in organic search results. A competitor may have gone out of stock.

The retail-media placement may have contributed to the purchase. It may have accelerated it, redirected it or simply appeared near a transaction that was already probable.

Attribution can describe the path captured by the platform. Incrementality attempts to estimate what would have happened without the media.

The distinction matters because a campaign can produce strong attributed ROAS while generating little additional business value.

This is particularly likely in environments where advertising is closest to existing purchase intent. Branded search, retargeting, sponsored listings and placements directed at loyal customers may produce impressive conversion rates because they reach people already likely to buy.

Those tactics can still be useful. They may defend shelf position, prevent competitive substitution or improve conversion during a critical period.

But buyers need to know whether they are:

  • creating demand
  • capturing existing demand
  • shifting share
  • increasing total category purchases
  • improving customer value
  • paying to receive credit for an expected transaction

Those are different commercial outcomes. They should not be collapsed into one ROAS figure.

Standards are beginning to catch up with spending

The retail-media industry is now building the measurement infrastructure that should have accompanied its growth from the beginning.

In late 2025, IAB and IAB Europe issued formal guidelines for incremental measurement in commerce media. The framework emphasizes credible counterfactuals, control of bias and separation of genuine signal from statistical noise. It outlines several acceptable approaches, including randomized experiments, model-based counterfactuals, econometric methods and hybrid proxies, depending on the business decision and available data.

In 2026, FMI released a U.S. retail-media measurement framework developed with contributions from NIQ and Think Blue. The standards are intended to create more consistent and comparable measurement across onsite, offsite and in-store media.

The Advertising Research Foundation has also conducted a multi-phase initiative examining retail-media networks’ measurement validity, metric consistency, transparency and incremental-lift claims. The project included direct inquiries and interviews with major networks and concluded with a final report in June 2026.

These efforts do not eliminate the measurement problem. Voluntary standards only matter when networks adopt them, buyers request them and commercial agreements make the underlying methodology visible.

But they change the negotiating position.

A brand no longer has to accept “this is how our platform reports it” as the end of the conversation. Buyers can increasingly point to shared definitions and ask how a network’s methodology aligns with them.

Transparency begins before the campaign runs

Many measurement disputes emerge after a campaign ends, when the buyer receives a report containing metrics it did not fully define in advance.

By then, the data required for a stronger analysis may no longer exist—or may never have been collected.

A serious retail-media plan should establish the measurement design before investment begins.

The buyer should know:

  • which business outcome the campaign is intended to change
  • whether success is product sales, category growth, new customers, repeat purchase, margin or lifetime value
  • what attribution window will be used
  • how view-through and click-through activity are treated
  • whether in-store sales are included
  • how identity matching works
  • which sales would be considered organic
  • whether exposed and control populations can be created
  • how promotions, price, availability and distribution changes will be considered
  • what data can be exported or independently analyzed
  • which party owns the final measurement decision

Without those answers, a campaign can be perfectly executed against a metric that does not resolve the buyer’s actual business question.

Incrementality is not one universal test

The demand for incrementality is sometimes discussed as though every campaign should be subjected to the same experimental method.

In practice, the appropriate method depends on the scale of the campaign, available data, network design, business risk and decision being made.

Randomized controlled tests

When a network can create genuinely comparable exposed and unexposed groups, randomized experiments can provide strong evidence of causality.

They are not always feasible. Audience overlap, small sample sizes, offline behavior, identity loss and network limitations can make clean experiments difficult.

Geographic or matched-market tests

Brands may increase or suppress media in selected markets and compare the results with similar locations.

These tests can be useful when retail distribution and marketing activity can be controlled, but they require careful matching and must account for local differences, promotions and competitive activity.

Econometric and marketing-mix models

These approaches estimate the relationship between media investment and business outcomes over time while controlling for other variables.

They can help guide larger budget decisions, although aggregated models may not provide the campaign- or product-level speed required for daily optimization.

Model-based counterfactuals

Where controlled experiments are unavailable, statistical models can estimate the outcome that would likely have occurred without the intervention.

The reliability of that estimate depends heavily on the completeness and quality of the underlying data.

Hybrid approaches

The strongest systems may combine experimental evidence, econometric analysis, platform attribution, customer data and operational information rather than treating any one method as complete.

IAB’s guidelines explicitly recognize that different methods provide different levels of causal rigor and should be chosen according to the decision the buyer needs to make.

The objective is not to produce an artificially precise universal number. It is to use the strongest evidence reasonably available and state the limitations clearly.

Media data alone cannot explain retail performance

Retail-media measurement becomes unreliable when it analyzes advertising in isolation from the commercial environment surrounding the sale.

A product’s performance can change because of:

  • price
  • promotion
  • inventory
  • distribution
  • organic ranking
  • product-page quality
  • ratings and reviews
  • buy-box ownership
  • seasonality
  • competitor activity
  • packaging changes
  • broader brand marketing

NIQ argues that credible incrementality measurement requires media data to be analyzed alongside retail sales, distribution and digital-shelf conditions. It gives the example of a product expanding from 2,000 to 3,800 stores: without that distribution information, a model could incorrectly attribute the resulting sales increase to advertising.

This is one reason retail-media measurement is not merely an advertising issue. It is a data-integration and organizational-design issue.

Marketing may manage the campaign. Ecommerce manages the product listing. Sales manages the retailer relationship. Revenue management controls promotion. Supply chain determines availability. Finance evaluates margin.

If those teams operate from separate data and incentives, no media dashboard can create a complete version of business performance.

The comparison problem is becoming a budget problem

Retail-media networks often report within their own environments.

That may help buyers optimize campaigns inside a particular retailer. It does less to answer the question facing a chief marketing officer or finance leader:

Where should the next dollar go?

A brand may advertise across Amazon, Walmart, Target, Instacart, grocery networks, specialty retailers, social commerce, connected television and the open web. Each partner may use different definitions, attribution windows and methods.

Even when every report is internally accurate, the results may not be comparable.

This creates a structural advantage for the channel with the most generous attribution methodology rather than the channel producing the greatest incremental value.

The danger grows as automated budget systems act on those metrics. If optimization software is instructed to maximize platform-reported ROAS, it will move money toward the environments best able to claim transactions—whether or not those environments create the greatest growth.

NIQ warns that AI-driven activation will increase the stakes because automated systems can act rapidly on flawed measurement signals.

Bad measurement does not become safer when it is automated. It becomes faster.

Buyers are beginning to demand decision rights, not just reports

Transparency is often framed as access to more data.

That is part of it, but not the whole issue.

A buyer may receive hundreds of fields without gaining the ability to challenge the network’s assumptions or reproduce the result. True transparency includes clarity about:

  • metric definitions
  • calculation methodology
  • exclusions
  • identity-match rates
  • attribution logic
  • data loss
  • confidence ranges
  • test design
  • known biases
  • changes made during the campaign

It also includes the right to use independent measurement.

A retailer should be able to explain why its internal numbers differ from a brand’s sales data or third-party analysis. A discrepancy is not necessarily evidence that one party is wrong. Different systems may answer different questions.

But unexplained discrepancies should not be normalized simply because the network owns the transaction data.

The strongest retail-media partnerships will increasingly be those in which buyer and seller agree on the business question, methodology and source of truth before the campaign begins.

What buyers should require

The next generation of retail-media buying standards should be more demanding than a dashboard login and a post-campaign ROAS report.

A documented measurement methodology

Networks should define:

  • attribution windows
  • click and impression rules
  • match methodology
  • sales inclusions
  • deduplication
  • treatment of returns and cancellations
  • new-customer definitions
  • whether reported outcomes are attributed or incremental

Comparable core metrics

Every network may offer proprietary insights, but buyers need a common set of definitions to compare investments across retailers.

The FMI, IAB and MRC-aligned standards efforts are intended to create that shared foundation.

Access to campaign-level data

Aggregated summary reports may obscure product, audience, placement and timing differences that materially affect interpretation.

Privacy and contractual restrictions remain valid, but buyers should know what can be exported, audited or analyzed independently.

Incrementality options

Networks should state whether they support randomized tests, geographic holdouts, matched controls or other counterfactual methods—and what minimum spend or sample size is required.

Independent verification

For significant investments, buyers should be able to compare network reporting with third-party analysis, their own commercial data or both.

Disclosure of methodological changes

A shift in attribution window or calculation logic can create apparent performance improvement without any change in the campaign.

Networks should disclose those changes and, where practical, provide restated historical comparisons.

Connection to business economics

Sales should be evaluated alongside margin, promotion, customer quality, repeat purchase and other downstream outcomes.

A campaign can increase attributed revenue while reducing profitability.

Nodus and the search for business truth

This measurement gap is central to the work of Nodus, which combines paid-media execution with attribution, incrementality and data integration for companies with complex customer journeys.

Disclosure: I serve as a fractional business-development and communications leader for Nodus. The company did not pay for, review or exercise editorial control over this article.

Nodus’s operating premise is that media performance should be reconciled with a company’s own business systems—including commerce, customer, sales and revenue data—rather than evaluated solely through the reports supplied by the platforms selling the media.

That does not mean retail-media network data should be dismissed. Retailers possess valuable transaction signals that brands cannot easily reproduce.

It means those signals should become one layer within a broader measurement system.

Platform attribution can support daily optimization. Incrementality testing can estimate causality. Marketing-mix analysis can guide allocation. Customer and financial data can reveal whether acquired sales produced durable economic value.

No single method answers every question. The strategic advantage comes from knowing which evidence should govern which decision.

Retailers also have something to gain from stronger standards

Transparency is sometimes presented as a concession retailers must make to advertisers.

In reality, credible measurement protects the retail-media business itself.

If brands lose confidence in reported performance, they may reduce investment, centralize spending within only the largest networks or treat retail media as a trade expense rather than a strategic marketing channel.

Networks that can demonstrate consistency, interoperability and incremental value will be better positioned to compete for long-term budgets.

Stronger measurement can also help retailers distinguish between inventory that merely monetizes existing traffic and programs that create value for brands, customers and the retailer’s core commerce business.

That distinction matters because retail media should not be optimized in isolation from the shopping experience.

A network can increase advertising revenue while making product discovery less useful, crowding organic results or favoring the highest bidder over the most relevant product. Short-term media yield can come at the expense of customer trust and retail performance.

The best networks will measure not only ad revenue, but also customer experience, product discovery, supplier value and incremental commerce.

What leaders should watch

Adoption beyond pilot programs

The appearance of incrementality tests in sales materials is not enough.

The meaningful signal will be whether networks make rigorous testing routinely available, at practical spending thresholds, across onsite, offsite and in-store programs.

Contractual transparency

Standards become more consequential when definitions, data access and methodology are written into commercial agreements.

Budget integration

Retail media has often been funded through a mixture of shopper-marketing, ecommerce, trade and brand budgets.

As measurement improves, organizations may begin allocating it within a more unified media and growth framework.

Independent measurement layers

Brands are likely to invest further in tools and partners that reconcile results across networks rather than accepting each retailer’s self-contained account of performance.

Finance participation

The channel’s measurement reckoning will accelerate when finance teams begin asking whether reported media returns reconcile with incremental revenue, contribution margin and total business growth.

AI-driven optimization

As automated bidding and agentic systems make more budget decisions, the quality of the measurement signal will become an operating-control issue—not merely an analytics concern.

The reckoning is a sign of maturity

Retail media is not facing greater scrutiny because the channel has failed.

It is facing scrutiny because it has become too important to operate without stronger proof.

The early phase was built on access: access to first-party data, shoppers, retailer inventory and closed-loop transactions.

The next phase will be built on credibility.

Brands will expect common definitions, transparent methodologies, causal evidence and a clearer connection between media reports and business value. Retailers that provide those things will not weaken their commercial position. They will strengthen it.

The networks most likely to win will not necessarily be those reporting the highest ROAS.

They will be those whose numbers buyers trust enough to use when the budget decision becomes difficult.