Retail Media Networks and Their Impact on Attribution

Retail media networks sell the placement and grade their own homework on whether it worked. That’s not unique to retail media, but retail media makes the conflict harder to see, because the sales data backing the attribution claim genuinely does belong to the retailer.

A retail media network is, structurally, a media owner reporting on its own media’s performance using data only it has full access to. That’s not a scandal, it’s just how retail media works. It does mean the reported numbers deserve more scrutiny than they typically get, not less.

The structural conflict nobody names directly

Every retail media network sits on the exact purchase data needed to prove whether its own ads drove a sale, first-party, high-confidence, and unavailable to any outside auditor without the retailer’s cooperation. That’s precisely why retail media promised better attribution than open-web advertising in the first place, and it’s also why the attribution claims are structurally harder to independently verify than almost anything else in a media plan. The retailer isn’t lying by reporting favourable numbers. It’s reporting numbers it has every incentive to calculate generously, using a methodology it doesn’t have to fully disclose, which is a different problem with a similar effect.

Where this shows up in the reported numbers

Correlation-based reporting, “this customer saw the ad and later bought the product”, is often presented with the same confidence as genuine incrementality testing, “this customer bought the product because of the ad, verified against a holdout group that didn’t see it”. The two produce very different numbers, and a retail media network reporting the first while implying the second is not a hypothetical risk, it’s the default unless a buyer specifically asks which one they’re looking at, the exact question raised in Key Questions To Ask Your Retail Media Partners.

The multi-touch attribution caveats covered in Multi-Touch Attribution: Measuring the Full Customer Journey apply here too, with an extra layer: a retail media network’s own on-platform reporting is effectively a closed system measuring itself, with none of the cross-platform visibility that even open-web MTA at least attempts.

What “good” retail media attribution actually requires

Three things separate a defensible retail media attribution claim from an unverifiable one: a stated methodology, correlation versus tested incrementality, disclosed rather than assumed; a willingness to run or support a holdout test on a meaningful share of spend, not just report against 100% exposure; and independent, third-party measurement layered on top of the platform’s own reporting wherever spend justifies the cost. None of these are unreasonable asks of a partner confident in their own numbers, which is exactly why reluctance on any of them is informative.

Bringing it back to online-offline

Retail media sits at an unusually direct intersection of online exposure and offline (in most retail categories) purchase, which makes it a natural extension of the matching methods covered in Offline Attribution: Measuring Marketing Impact in the Physical World. The retailer already has the loyalty and transaction data most offline attribution setups spend real effort trying to acquire. That’s the platform’s genuine advantage, and the reason its attribution claims are worth taking seriously in the first place. It’s also exactly the data an outside buyer can’t independently audit, which is why “worth taking seriously” and “worth taking at face value” aren’t the same standard.