The retail media report card.

I was in a quarterly review last spring when a VP of e-commerce pulled up a slide that was supposed to be a victory lap. Retail media spend was up forty percent year over year. The retailer's platform was reporting ROAS north of six to one. Everyone nodded. Then someone from analytics asked: "Can we see how this compares to paid social and search using the same attribution model?" The room went quiet. Nobody could. Not because the team was incompetent, but because the data literally does not allow it.

That moment captures the central tension of retail media right now. The premise was that these networks would finally close the loop between ad exposure and purchase. That was the pitch. That was why brands moved budgets. On paper, the logic is airtight. Who better to prove an ad drove a sale than the retailer who processed the transaction? But a year of heavy investment later, the same problem surfaces everywhere: the measurement is proprietary, the data doesn't port, and brands still cannot compare retail media performance to anything else in their mix with confidence.

The pitch that opened the floodgates

Retail media networks arrived at exactly the right moment. Third-party cookies were crumbling. Signal loss from privacy regulations was making digital attribution messier by the quarter. And here came retailers sitting on mountains of first-party purchase data, offering a closed-loop measurement dream. You spend with us, we tell you exactly what sold. No modeling. No probabilistic matching. Deterministic, transaction-level proof.

For brands that had spent years squinting at multi-touch attribution models, this sounded like salvation. The early results looked incredible. ROAS figures that made paid social look anemic. The money followed fast. But nobody talked loudly enough about the fine print. The measurement was real, but it was the retailer's measurement, built on the retailer's methodology, inside the retailer's ecosystem. The structural advantage that made the data valuable also made it impossible to verify independently.

Grading your own homework

Every retail media network is simultaneously the media seller and the measurement provider. They sell the placement, report on performance, and define the attribution window and methodology connecting the two. I have worked in media long enough to know that this arrangement never produces conservative numbers.

I worked with a brand running campaigns across three major retail media networks simultaneously. Each reported strong ROAS. Each claimed credit for overlapping purchases. When the internal team tried to reconcile the numbers, total attributed sales exceeded actual sales by a meaningful margin. The math did not work. But there was no mechanism to arbitrate because each network's data lived in its own walled garden, measured by its own rules, inaccessible to outside analysis.

This is not new. We spent a decade having the same argument about Facebook and Google grading their own homework. The difference is that retail media sold itself as the antidote to that opacity. Instead, we got a new set of walled gardens with the same conflict of interest, dressed up in transactional data.

The portability problem

Even if you trust the numbers coming out of a single retail media network, you still cannot do anything useful with them at a portfolio level. The data does not port. Each network has its own taxonomy, its own attribution windows, its own definitions of what counts as a conversion. Some use seven-day click windows. Some use fourteen. Some count view-through conversions. Some don't. You cannot compare a six-to-one ROAS from one network to a four-to-one from another, let alone compare either to your paid search performance. The numbers are not speaking the same language.

I have watched smart, well-resourced teams try to normalize this data into unified measurement frameworks. Even the best attempts involve so many assumptions that the output is more art than science. The brands that cannot afford to do it, which is most of them, are flying blind outside each individual network's dashboard.

Incrementality is the question nobody wants to answer

The deeper issue underneath all of this is incrementality. Retail media networks report on total attributed sales, but what brands actually need to know is how many of those sales would not have happened without the ad. If someone searches for a specific product on a retailer's site, sees a sponsored listing, and buys that product, was the ad the cause or was it just a toll booth on a journey that was already underway?

Some networks are starting to offer incrementality studies, but they are expensive, slow, and still controlled by the network. I have seen incrementality results that showed a fraction of the lift the standard reporting claimed. That gap is where a lot of budget is quietly being wasted.

This matters because the budgets moving into retail media are not new money. They are coming from somewhere. Every dollar that shifts into a retail media network because the reported ROAS looks unbeatable is a dollar not going toward a channel where the actual incremental impact might be higher. Without a level playing field, brands are optimizing for the dashboard that tells the best story rather than the allocation that drives the most growth.

What would actually fix this

I do not think retail media is a bad channel. I think it is a genuinely valuable one. Reaching consumers at or near the point of purchase, informed by real transaction data, is a powerful proposition. The problem is not the media. The problem is the measurement infrastructure around it, and the fact that the incentives of the networks are not aligned with giving brands the transparency they need to evaluate it honestly.

What would actually move things forward is straightforward in theory and difficult in practice. Standardized attribution methodologies across networks. Independent measurement partners with full data access. Routine incrementality testing not controlled by the entity selling the media. None of this is technically impossible. It is just commercially inconvenient. Standardization and transparency compress margins because they let buyers see what is actually working. The networks that move toward openness first will earn disproportionate trust, but that requires a longer time horizon than most quarterly-driven organizations are comfortable with.

Until that happens, brands need to treat retail media reporting the way a good editor treats a first draft: with interest, but also with a red pen. The numbers are a starting point for analysis, not the final word. The brands that build their own measurement capabilities, invest in incrementality testing, and resist the temptation to take reported ROAS at face value are the ones that will actually capture the value retail media can deliver. Everyone else is just renting confidence from a landlord who also wrote the lease.