U.S. retail media spend will come close to $70 billion this year, according to eMarketer. Most brands spending that money still grade it on platform-reported ROAS, a metric that somehow manages to overstate and understate performance at the same time.
Fixing that takes incrementality measurement. Below is what it means, how attribution differs across Amazon, Walmart Connect, and Instacart, and how to measure incrementality across every retailer you buy.
Key takeaways
Platform ROAS credits ads for sales that would have happened anyway, and it misses sales ads drove in-store or at other retailers.
Every RMN uses their own attribution models and lookback windows, so their ROAS figures can't be compared.
Incremental's analysis found no correlation between campaign performance on ROAS and on incremental return. The campaigns ROAS calls winners often aren't.
iROI (sometimes called iROAS) is incremental sales divided by ad spend. Measured the same way across every retailer, it tells you where the next dollar goes furthest.
Why ROAS gets it wrong in both directions
ROAS overstates performance when it takes credit for sales that were coming anyway. Picture a shopper already searching for your product on Amazon. They'd have bought with or without the sponsored listing, but the listing gets the credit.
It understates performance when it misses sales your ads really did cause but couldn't track. The shopper who saw your ad online and bought in store. The lift at one retailer driven by spend at another. Cross-retailer halo effects are real, and platform-reported ROAS doesn't see any of them.
So the picture you get is too generous and too incomplete at once, which is a bad combination for deciding where money goes.
The structure of retail media makes it worse. Every network grades its own homework, using its own attribution model and its own rules about which clicks and views count and for how long. Each also has a financial interest in the results it reports. Buyers have noticed. At an IAB Europe roundtable in Paris this July, agency leaders named mismatched lookback windows and murky ROAS math across networks as real sources of friction, and IAB Europe's write-up concluded that without a shared lookback window, ROAS from one network can't be meaningfully compared to another.
How Amazon, Walmart Connect and Instacart differ
The three biggest networks don't just report different numbers. They define a conversion differently, and those definitions keep moving.
Amazon kept Sponsored Products on click-based attribution, but on January 1 it moved view-based ads to a new "shopping-signal enhanced" last-touch model. That covers Sponsored Brands and Sponsored Display campaigns billed on viewable impressions, plus Amazon DSP ads running in the Amazon store. A rules change like that can move reported ROAS without anything about the media itself changing.
Walmart Connect ties ad exposure to both online and in-store purchases, which is a real strength. Its sponsored search reports also let advertisers switch between 3-, 14- and 30-day click windows, and the window a team picks changes the ROAS it ends up defending.
Instacart made linear multi-touch its default in August 2022, spreading credit across every ad a shopper interacted with, with last-touch still available for comparison. Arguably that's fairer inside Instacart. It also means an Instacart ROAS and an Amazon ROAS are built on different math.
Add Target, DoorDash and the rest of a typical media plan and the problem only grows. A 4x on one network just isn't the same 4x on another.
What incrementality measurement actually requires
Incrementality comes down to a counterfactual question. What would sales have looked like without this ad?
Attribution can't answer that. It starts from a conversion and works backward, handing out credit. Causal measurement isolates what the ad caused by stripping out everything that would have happened regardless. The industry is formalizing the distinction, too: IAB and IAB Europe published joint guidelines on incremental measurement for commerce media last November.
In retail media, a model has to be built around the commerce environment as well as the media. Sales move for plenty of reasons that have nothing to do with ads, including seasonality, promotions on the same SKU, competitor pricing, inventory levels and search rank. If your product went on promotion the same week your Sponsored Brands campaign launched, or a competitor went out of stock and your organic sales jumped, a model that ignores those things isn't measuring your ads. It's measuring noise.
What changes when you measure it
The first thing that changes is which campaigns look like winners. When Incremental compared ROAS and iROI campaign by campaign, you'd expect the two to roughly track, with the top performers on one sitting near the top of the other. They didn't. The scatter plot looked like a Jackson Pollock. There was no correlation between the two, a finding eMarketer has since cited. The campaigns ROAS ranked highest weren't the ones driving the most incremental sales.
The same analysis showed how wide the spread is. 43% of retail media spend returned an iROI of 6x or better, on par with the most efficient media channels anywhere. Another 33% returned less than 1x, so less than a dollar of incremental sales for every dollar spent. Some of that low-return spend can still be worth it (defending branded keywords from competitors, for one), but ROAS blurs both groups together and gives you no way to tell which is which.
iROI does. Sometimes called iROAS, it's incremental sales divided by ad spend, and it's a number you can actually budget against. Our iROI vs. ROAS vs. MER breakdown goes deeper on how the metrics compare.
The second thing that changes is comparability. Run every retailer through the same causal lens and you can finally ask whether the next dollar goes further on Amazon or Walmart, and get a real answer.
How to measure retail media incrementality across retailers
Start from total sales, not attributed sales. Pull retailer sales data for every network you buy and sit it alongside the ad data. Incrementality is measured against what actually sold, so platform-attributed sales can't get you there on their own.
Put the commerce variables in the model. Promotions, your pricing and competitors' pricing, inventory and out-of-stocks, search rank, seasonality. Anything left out of the model ends up as credit for your ads.
Use one methodology and one metric everywhere. The same baseline logic for Amazon, Walmart, Instacart, Target and DoorDash, with iROI as the shared output. Platform ROAS can stay in reports as a diagnostic. It just shouldn't drive budget.
Read it daily, at the campaign level. Retail media budgets move every day. A quarterly study tells you what already happened, and you need a signal fast and granular enough to act on while the campaign is still running.
Reallocate, and expect ROAS to drop. Move budget from low-iROI campaigns to high-iROI ones. Cheap branded placements tend to carry the highest ROAS and the least incrementality, which is why the smartest brands may end up tanking their ROAS on purpose. If incremental sales go up, that's the trade you want.
Have someone independent do the measuring. Whoever grades the results shouldn't also be selling you the media.
Where Incremental fits
Incremental was built to do exactly this. It gives brands and agencies an independent, third-party read on what their retail media actually caused, updated daily at the campaign level across Amazon, Walmart, Instacart, Target, DoorDash and more. Every retailer runs through the same causal methodology, so an iROI on one network means the same thing as an iROI on another.
Measurement is the input. That daily read flows into activation platforms like Skai and Kepler's Kip so budget can follow it, and it feeds scenario planning for where the next dollar should go.
The results hold up. Bayer raised its iROI 32% with Incremental and Skai, more than three times its original goal, without adding spend. At Church & Dwight, optimizing to iROI improved incremental return 292% for Batiste and 161% for TheraBreath. Seven of the 10 largest CPG brands and all six major agency holding companies use Incremental today.
