Are Your Ads Truly Paying Off?

These platforms, called retail media networks, have become a popular model in digital advertising. Pioneered by giants such as Amazon and Walmart, the model has spread to grocery and clothing chains, delivery apps, and other e-commerce channels. Instead of advertising on a news site or social media and hoping a customer clicks through, brands can place their pitch right on the “digital shelf,” catching users who are already in a buying mood.

Marketers judge impact with return on ad spend (ROAS)—the sales attributed to a campaign divided by the amount spent on that campaign. Retail media networks bring ads to the point of purchase and improve what can be observed. But ROAS is still a calculation, and the resulting numbers depend on how one defines what counts as a sale.

Retail media networks offer much more clarity, at least theoretically. Because shoppers are often logged into an account as they browse, they leave a data trail that can help marketers measure the effect of their campaigns. Did a cereal brand’s promotion on Instacart lead to more purchases? Now you can count exactly how many people saw the ad and determine whether those people subsequently bought the product, presumably making ROAS simpler to calculate.

However, some of the old measurement and comparison issues with digital advertising remain, write Kellogg’s Brett Gordon and Eric Anderson in a new whitepaper. Despite the data advantages of retail media networks, each platform may use very different techniques for estimating ROAS, making it hard for companies to compare performance across platforms.

To ensure consistency, accuracy, and transparency in their data, marketing leaders need a firm understanding of the components that make up the ROAS numbers they’re using, Gordon and Anderson say. Without this clarity, it is difficult to make decisions based on that data.

“Advertisers need to recognize that the metrics that they’re seeing, just because they’re labeled as ROAS, doesn’t mean that they’re all the same ROAS,” says Gordon. “There are different assumptions that go into the construction of any given metric, depending on the platform.”

A peek under the ROAS hood

While the basic formula for ROAS—the total dollars generated from a campaign divided by the total dollars spent on it—is clear enough, many decisions and details go into calculating this number.

“When you think: How do we count up revenue?” Anderson says, “you quickly realize there’s no standard definition of ROAS.”

For example, how does a formula account for purchases that can’t be attributed to an individual account, as in when a user shops as a “Guest”? Do you associate purchases with an ad that features an umbrella brand versus a specific product? Should you weigh every ad a user sees equally, or give more weight to the final ad seen before their purchase?

Working with Albertson’s Media Collective—the retail media network of Albertson’s grocery-store chain—Gordon and Anderson were able to test the impact of the most-common assumptions that go into ROAS calculations. The researchers analyzed nearly 600 advertising campaigns with Albertson’s in 2023 and 2024.

They found that seemingly small decisions can make a massive difference. Product attribution—whether sales are attributed to only ads for that specific product versus ads for the broader brand—can shift ROAS by an average of 35 percent. Combining the impact of four different methodological choices produced an average shift of 63 percent, which can mean the difference between a successful campaign and a flop.

“When seemingly ordinary, defensible choices move ROAS by 63 percent, that can flip a go to a no-go,” says Gordon. “That’s why cross-platform comparisons should start with aligned assumptions around a common calculation.”

But that variability isn’t a dealbreaker when advertisers are measuring performance from a single retail media network. If the platform’s method for calculating ROAS doesn’t change, marketers can still track improvements and declines in campaign performance over time.

The real problem arises when advertisers want to compare results across platforms that may have very different ROAS formulas.

“Imagine an advertiser running campaigns for the same brand across networks and then wanting to compare performance to determine which should get more ad budget in the future,” Gordon said. “Differences in ROAS methodology could lead the advertiser to misallocate their ad budget.”

Ask the right questions

This variability doesn’t mean retail media networks aren’t still a big step forward for digital advertising metrics, Gordon and Anderson say. Advertisers simply need to be more vigilant about the methods that their chosen platforms use.

Through conversations with your advertising partners, you should be able to design ROAS measures that meet your needs, document the behavior or numbers that you think are important, and enable you to make meaningful comparisons.

“The starting point is to have consistency in the metrics and to align them as much as possible, across those channels or those partners, with assumptions that you’re comfortable with,” says Gordon.

Think carefully about the data that you want to measure and that will be the most useful for you, while getting clarity from the platform hosting the ad campaign on what they are reporting. Knowing what makes the most sense for your advertising approach and sales strategy—and not just letting the advertising platform simply serve up what is easiest or makes them look the best—is key.

“When you are spending your ad money, you should be dictating what you want to go into the calculation,” Anderson says.

A firm thinking of investing in ad campaigns with multiple retail media networks will want to address certain questions ahead of time to make sure all their campaigns are gathering comparable ROAS information. Gordon and Anderson suggest asking these questions:

• What is the granularity at which the retail partner typically attributes sales to a campaign? Is it at the product level, the brand level, or another level you are not tracking?

• Who gets credit for a purchase—just the logged-in shopper or anyone in their household? Can you report both views so we can compare?

• Which methodology will you use to extrapolate untraceable sales?

• Is the data attributing one sale to multiple campaigns or channels?

Gordon notes that retailers and platforms aren’t trying to be sneaky, but many don’t necessarily put too much thought into the consistency of the metrics they report back to advertisers. Taking a proactive role in how these partners gather ROAS data will make a huge difference in calculating your campaigns’ effectiveness.

“Not all advertisers ask these kinds of questions, and not all industry participants have the time or the interest in doing so,” Gordon says. “But if you’re trying to make a more-educated comparison, you would like the underlying assumptions to be aligned.”