Quick answer: Return on Ad Spend (ROAS) is a flawed metric because it credits ad exposures for sales that would have happened anyway. According to retail media analyst Andrew Lipsman, ROAS is “the root of all evil” and both brands and retail media networks should shift toward Incremental ROAS (IROAS), market share, share of voice, and long-term value to measure what advertising is actually doing.
Retail media is going through a maturation moment. Advertisers are asking harder questions, scrutinizing measurement, and beginning to question whether the metrics they have relied on for years are telling them anything useful.
In this episode of Unlocking Retail Media, Avery talks with Lipsman, who spent over five years at eMarketer defining the analytical frameworks the industry uses to understand retail media. Below are the five biggest takeaways, structured as direct answers to the questions advertisers and retailers are asking right now.
Takeaway #1: Why Is ROAS the Wrong Metric for Retail Media?
"ROAS is not the right metric. It's not only a flawed metric, it can lead you in the exact wrong direction." - Andrew Lipsman
Short answer: ROAS is a misnomer. It does not attribute a sale to ad exposure in any causal sense. It only confirms that a sale occurred after an impression was served.
Because optimization systems chase whatever credit they can take, ROAS pulls budget toward two channels that rarely produce incremental sales:
- Branded search. Captures shoppers who were going to buy your brand anyway.
- Retargeting. Puts ads in front of consumers already deep in the funnel, then claims the conversion.
As Lipsman put it, "You have to have a system that can't be gamed because the machines will game it otherwise."
Takeaway #2: What Is the Juicero Problem with Black Box AI?
Lipsman recently gave a talk titled "When the Juice Is Not Worth the Squeeze," referencing the $700 Juicero device that turned out to be unnecessary because shoppers could squeeze the juice packets by hand.
He sees the same pattern in advertising. Black box optimization tools wrapped in an "AI hand wave" promise outcomes without transparency. The systems are tuned to deliver the KPIs advertisers ask for, but those KPIs rarely tell the whole story.
Why does this happen? When machine learning models are pointed at ROAS and click-through rate, they optimize toward shoppers who were already going to convert. Avery noted from his own experience building these systems that algorithms can actively push ROAS up while pushing incrementality down. The two metrics are often inversely related.
The result: optimization systems will spend money on the worst inventory available, including bot traffic and low-quality impressions on the open web, because cheap junk impressions inflate ROAS.
Takeaway #3: Are Brands Their Own Worst Enemy?
Yes, often. Lipsman used to assume brands were victims of bad measurement. He now sees it differently.
Inside most brands, individual incentives drive the wrong behavior:
- Procurement teams are tasked with driving CPMs down.
- Performance marketers are measured on short-term ROAS.
- Black box tools are adopted because they make daily reporting easy.
"It's designed to make their lives easy and to produce the metrics that they want," Lipsman explained. "But you have to think about what you lose in the long term."
Two examples illustrate the cost:
- Nike. After shifting heavily into performance marketing and direct-to-consumer optimization under former CEO John Donahoe, the brand eroded the equity that made people want to buy it in the first place.
- Tropicana. A recognizable orange with a straw, a heuristic shoppers used for decades, was replaced by sleeker packaging that looked generic on the shelf.
Optimization without strategy shrinks brands. It does not grow them.
Takeaway #4: What Metrics Should Retail Media Networks Build Around?
Lipsman is a realist. ROAS will not disappear tomorrow. But retail media networks have a clear opportunity to graduate brands toward better measurement, and doing so is in their long-term interest.
The metrics that should anchor a more mature retail media offering:
- Incremental ROAS (IROAS). The right starting point. Show ROAS and IROAS side by side so brands learn the two are not always correlated.
- Market share. An outcome metric that accounts for the competitive set. You can run a great campaign and still lose share if competitors outspend you.
- Share of voice. A leading indicator that pairs with market share as a lagging one.
- Long-term value. In CPG, a 12-month window is a reasonable proxy for lifetime value and a better optimization target than next-week conversions.
Lipsman also flagged a hidden advantage for retail media networks. When measured properly through test-and-control or causal modeling, retail media often shows up as one of the most undervalued channels in the mix. Traditional MMM models have systematically understated its impact because they bias toward historical priors like Meta, YouTube, and linear TV.
"All else equal, you win in that worldview. You've taken the easy money from ROAS the last few years. But long term, there's a much bigger part of the market you can take if you align with incrementality." - Andrew Lipsman
Takeaway #5: Where Is Retail Media Headed Next?
The short answer: up the funnel. CMOs have largely ignored retail media because it has been pigeonholed as an e-commerce tactic. That is changing on two fronts.
Performance TV is the most imminent expansion. Amazon already pairs targeting data with closed-loop measurement, and that combination is now underpinning a meaningful share of the ad-supported CTV market. In-store media is becoming real as well, with Kroger rolling out digital screens and Walmart unlikely to stay a laggard for long.
But Lipsman warned that the industry is at risk of repeating its own mistakes. Performance TV is already being measured through a ROAS lens, which makes premium inventory look expensive and pushes spend toward TV retargeting. The right approach bakes incrementality into measurement from the start and layers in brand-building effects as part of the KPI.
In-store measurement remains what Lipsman calls a "white whale." Existing technologies have gaps, missing core components like accurate exposure data or location-based attribution. New solutions coming to market later this year may finally close the loop from in-store ad exposure to in-store purchase through an incrementality lens.
Key Takeaways at a Glance
Is ROAS a good metric?
Lipsman's Answer: No. It credits sales that would have happened anyway.
What should replace it?
Lipsman's Answer: IROAS, market share, share of voice, long-term value.
Why does black box AI fail?
Lipsman's Answer: It optimizes for KPIs without transparency, often hurting incrementality.
Where is retail media going?
Lipsman's Answer: Up the funnel into CTV, performance TV, and in-store media.
What is the biggest mistake brands make?
Lipsman's Answer: Letting procurement and short-term ROAS goals pull spend away from brand-building channels.
Conclusion: Stop Optimizing for the Wrong Things
Lipsman's message to the industry is clear. ROAS has been useful, but it has also been gamed, and the cost of staying addicted to it is becoming visible in declining brand health and inflated black box performance claims.
For retailers building media networks, the path forward is to expose better metrics inside the platform itself: IROAS alongside ROAS, market share alongside conversions, brand-building effects alongside short-term lift. AI and machine learning are powerful tools, but only when pointed at the right targets. Aim them at flawed metrics and they will reliably deliver flawed results, faster than ever before.
Retail media networks that help advertisers see what is actually working, even when the numbers look less flashy at first, will be the ones that capture real budget growth as the channel matures.
Listen to the Full Conversation on Unlocking Retail Media
For more insights like these, tune in to the full episode of Unlocking Retail Media, the podcast where Kevel CEO James Avery sits down with industry leaders and innovators shaping the future of retail advertising.
Listen to this episode with Andrew Lipsman here.
