The same campaign, the same retailer, the same week: one methodology says your iROAS is 3.0, another says 0.5, and a third says you lost money altogether. All three are technically defensible, but none are comparable.

A 42-campaign study from Albertsons Media Collective, Ovative, and Kellogg School of Management landed recently with a finding that anyone who has sat across the table from three retail media networks already knew in their gut: methodology alone can shift iROAS results up to 6.5x, and in some cases flip a positive result into a negative one.

A month ago at Skai, I said the industry should banish the word incrementality. Not the discipline. The word. Because it had become the new synergy. The new engagement. Five mentions in the first ten minutes of every pitch, and three different definitions if you asked three people in the room. iROAS is following the same path.

iROAS is a methodology stack, not a number

What nobody who is selling you iROAS tends to spell out: the figure on your report is the output of a series of choices: holdout design, matched market construction, MMM specification, attribution model, lookback window, control group composition, and conversion definition. Every choice in that stack moves the number. Every RMN picks the stack that makes them look best. That's optimization, the same kind you'd apply if your renewal depended on showing positive iROAS to the brand spending eight figures with you.

That's how the same impressions, the same clicks, the same SKU, the same week, land at 3.0 iROAS in one report and 0.5 in another.

Steve Baxter at Ovative put it plainly: "There has been mistrust between the buy side and the sell side because of the confusion of metrics. There are advertising dollars up for grabs, and the dollars are going to flow towards performance. But that flow happens on the rails of trust."

The rails aren't built yet. Sit in a QBR with three RMNs: each one will hand you an iROAS chart, each on a different methodology, each claiming to be the highest-performing partner in your mix. They can't all be right. They can all be technically defensible. Defensible and comparable are not the same thing, and that distinction is where brands keep losing ground.

The walls of the walled garden are definitions

Every retailer has been selling brands closed-loop measurement as a feature. For the retailer, it is. The retailer sees the impression, the click, the basket, the conversion, the methodology, and the math behind the number they hand you. You see the number.

The definition advantage compounds in one direction. Whoever defines iROAS defines whether your campaign worked. Whoever decides whether your campaign worked decides whether you renew. Whoever decides whether you renew owns your budget. Brands sit at the end of that chain holding a PDF.

What to do about it

The brands making progress have stopped waiting for the industry to standardize and started building their own normalized iROAS framework, requiring every RMN partner to report into it. Pick the methodology, define the holdout, specify the lookback window, lock the conversion definition, and hand it to every partner. If a partner can't or won't report into your framework, that's information about the partner.

The framework just has to be yours. Allocating retail media spend based on the iROAS your RMNs report back to you means letting the people selling you the media grade the media. The industry would never accept that arrangement from a search or social platform. It's accepted from RMNs because the language sounds rigorous. The study just put a number on how much it isn't.

What we're seeing

Across our retail media clients, the methodology conversation has moved from background concern to active priority. We're currently building a normalized measurement framework for a client running campaigns across multiple RMNs, specifically the work of aligning holdout design, lookback windows, and conversion definitions across partners so that the comparison is actually a comparison. It's painstaking, and it's exactly what the study describes as the only path forward.

A few things we're watching closely in 2026: every multi-retailer brand needs this kind of framework in 2026, ahead of the next budget cycle. The RMNs that gain share in the next budget cycle will be the ones willing to submit to a brand-defined methodology rather than defend their own. And iROAS, as a standalone term, one untethered from the methodology behind it, is losing credibility fast among the buyers who understand what the number actually is.