Retail media’s biggest strength is also its biggest risk.

Retail media is having the easiest argument of any channel in a decade. Global spend reached roughly $174 billion in 2025, around 15% of all digital advertising, and it is still growing faster than 17% a year, according to eMarketer and WPP GroupM. CPG brands now put close to 40% of their ad budgets there, on Oliver Wyman’s numbers, and most say they plan to spend more.

I understand the pull completely. I would go further: for CPG, this is the most rational-looking budget shift in years. To see why, you have to remember the problem retail media appears to solve.

CPG brands have spent decades advertising more or less blind. They sell through third-party retailers, and those retailers have never been in a hurry to hand back granular, brand-level sales data. Performance media, in the direct-response sense, was never really on the table. So CPG did the sensible thing and built brands: broad reach, mental availability, the long game. When they wanted to prove it worked, they had two options, neither comfortable. Marketing mix modelling, which decomposes sales into the factors that drove them, is rigorous but expensive, slow and backward-looking, usually needing a couple of years of history before it says anything useful. Geo-based incrementality tests are cleaner but selective, and limited in what they can cover.

Then retail media arrived with something CPG had never had: an immediate, apparently clean line from ad to sale. After thirty years of arguing brand effects in front of a sceptical CFO, you can see why that was irresistible.

Here is the part that tends to get skipped. Retail media’s closed-loop is mechanically similar to exposure or click-based attribution inside a walled garden. It answers one question: did someone who saw this ad buy the product in this retailer’s store within a short window. That is a real signal. It is also a narrow one, and it is biased toward conversion of demand that already existed. It tells you who finished the purchase. It says almost nothing about who created the intention to buy, or when, or where. It produces the feeling of accountability while mostly measuring harvesting.

The rise of digital and social did not lift overall marketing effectiveness. The most credible reading of the evidence, Les Binet and Peter Field’s analysis of the IPA Databank across 996 case studies and three decades, is that it pulled the other way, by rewarding short-termism. Their core finding has held up for more than ten years. The optimum the data points to is roughly 60% of budget on brand building and 40% on activation; brands that tilt much harder toward activation than that see immediate gains followed by long-term decline. At the IPA’s 2025 effectiveness conference, Binet put it more bluntly still, arguing that the industry has become fixated on efficiency, targeting and short-term metrics at the expense of the reach and brand-building that actually compound.

The mechanism is a slow loop, which is why it is so easy to miss. Activation captures the demand that brand building creates. Lean too hard on it and brand equity erodes, which makes activation less effective, which invites still more activation spend. You keep picking fruit while the tree goes unwatered. Retail media is – by design – an activation channel.

Retail media spend is extraordinarily concentrated. In the US, Amazon and Walmart are absorbing close to 90% of every incremental retail media dollar, on eMarketer’s figures. Concentrating budget into retail media therefore means concentrating it into two companies: their inventory, their audiences, their rules, and their measurement marking its own homework. The data and the learnings do not travel with you. And the audience you can reach stops at the edge of the platform, which matters because shoppers do not live inside it.

And the rules are the retailer’s to set, and to change. Watch what has happened around the Walmart DSP in the past year. Walmart ended The Trade Desk’s four-year exclusivity in late 2025, started opening its shopper data to Yahoo’s DSP and Google’s DV360 through 2026, both still in closed trials, and in June agreed to acquire Vibe.co to gain a connected-TV bidder of its own. Read the moves together and the plan is plain enough: own the data and the measurement, rent out the activation on terms you control. Right now those terms are reasonably open. Nothing guarantees they stay that way, and Amazon, which keeps its data firmly inside its own DSP, is the other model on the table. For a marketer the takeaway is simple. You don’t control access to these platforms, the retailer does, and what is open today can be restricted tomorrow. Build too much of your plan on one of these platforms and you have given someone else a vote on how (or even whether!) you reach your own buyers.

In Adlook’s 2026 study of 6,500 UK shoppers, we set out to find where everyday brand choices actually form. Half of shoppers told us they either buy the same brand every time or simply take whatever is in front of them. Only 23% actively weigh four or more brands. So the decisive moment is rarely the shelf. It is earlier.

63% said a brand came to mind at the point of purchase specifically because they had seen it in a recipe, an article or a review. 32% said they went on to buy a brand for that reason. The shelf, physical or digital, mostly ratifies a decision taken upstream, out on the open web, in content that does not interrupt anyone and does not push for a click. Retail media is very good at ratification. It does not do the upstream work, and it cannot see it.

None of this is an argument against retail media. It is an argument against mistaking one slice of the funnel for the whole of it. Brands grow on two things, and Ehrenberg-Bass has shown this about as conclusively as our field shows anything: mental availability, being easy to think of, and physical availability, being easy to find and buy. Retail media touches the second, at the last moment. 

The job is to keep doing both, on purpose. A few levers help, and no single one is the full answer.

The first is plain discipline. Treat retail media as activation, and hold the brand-building line against it, rather than letting measurability decide the budget split by default. The most measurable number in the room wins every argument it is in, whether or not it deserves to.

The second is honest measurement. Incrementality testing and mix modelling exist precisely to ask whether retail media’s uplifts are genuinely incremental or merely harvested from demand that was already there. My colleagues Luca Filardo and Mateusz Jędrocha have written about building a measurement stack that climbs from media quality to brand change to verified business outcomes, instead of stopping at whichever metric is easiest to collect. Worth a read!

The third is format. There is no law that says building memory and signalling availability have to be separate line items. Shoppable, commerce-style ads on the open web can do both at once: keep a brand mentally available while showing a category shopper where it can be bought, across many retailers rather than one, free of the walled-garden limits, and generating real insight into buying preferences along the way. Adlook’s Commerce Media is one expression of that idea. It is not the only one possible, and it should not be the only one tried.

The honest worry is that we are about to repeat, with retail media, the exact mistake we made with programmatic and then with social: confusing what is easy to measure with what actually works. The numbers retail media produces are real. The question every CPG marketer should be asking is not whether those numbers look accountable, but how much of what they count was going to happen anyway, and what is going unwatered while we stand there admiring the harvest.

Sources:

WPP Media, “This Year Next Year” Report (December 2025) — https://www.wppmedia.com/news/campaigns-report-this-year-next-year-december-2025

WPP Media, TYNY Mid-Year 2025 Research — https://www.wppmedia.com/thought-leadership/research-business-intelligence/tyny-midyear-2025-research

eMarketer, Retail Media Forecast Report Update — https://www.emarketer.com/content/retail-media-forecast-report-update

Oliver Wyman, “The Future Of Retail Media And Its Impact On Advertising” (December 2024) — https://www.oliverwyman.com/our-expertise/insights/2024/dec/retail-media-future-advertising-growth.html

eMarketer, “Retail Media Ad Spending Forecast and Trends H2 2025” — https://www.emarketer.com/content/retail-media-ad-spending-forecast-trends-h2-2025

AdExchanger, Daily News Roundup (August 15, 2025) — https://www.adexchanger.com/daily-news-roundup/friday-15082025/

eMarketer, “Walmart Broadens DSP Partnerships Beyond The Trade Desk” — https://www.emarketer.com/content/walmart-broadens-dsp-partnerships-beyond-trade-desk

Marketing Dive, “Walmart Connect Expands Offsite Advertising Ambitions With Yahoo, Magnite” (May 28, 2026) — https://www.marketingdive.com/news/walmart-connect-expands-offsite-advertising-ambitions-with-yahoo-magnite/821575/

Google, “Display & Video 360 and Walmart Connect Team Up” (June 11, 2026) — https://blog.google/products/marketingplatform/360/walmart-connect/

Walmart Connect, “Bringing Walmart Connect’s First-Party Audiences to Google’s Display & Video 360” — https://www.walmartconnect.com/resources/articles/2026/bringing-walmart-connects-first-party-audiences-to-googles-display-video-360

eMarketer, “Google and Walmart Connect Link YouTube Ads to Sales With Closed-Loop Measurement” — https://www.emarketer.com/content/google-walmart-connect-link-youtube-ads-sales-with-closed-loop-measurement

Walmart Corporate, “Walmart to Acquire Vibe.co to Expand Access to Connected TV Advertising” (June 23, 2026) — https://corporate.walmart.com/news/2026/06/23/walmart-to-acquire-vibe-co-to-expand-access-to-connected-tv-advertising

Digiday, “Walmart Buys the Google Ads of Streaming, Vibe, in a Deal Tipped at a $1 Billion-Plus Valuation” — https://digiday.com/future-of-tv/walmart-buys-the-google-ads-of-streaming-vibe-in-a-deal-tipped-at-a-1-billion-plus-valuation/

IPA, “The Long and the Short of It: Balancing Short and Long-Term Marketing Strategies” (Les Binet & Peter Field, 2013) — https://ipa.co.uk/knowledge/publications-reports/the-long-and-the-short-of-it-balancing-short-and-long-term-marketing-strategies

IPA Effectiveness Conference 2025, “Go Big or Go Home” (Les Binet & Will Davis) — https://ipa.co.uk/knowledge/effectiveness/ipa-effectiveness-conference-2025/go-big-or-go-home

Ehrenberg-Bass Institute, “How Do You Measure ‘How Brands Grow’?” — https://marketingscience.info/news-and-insights/how-do-you-measure-how-brands-grow

Adlook, proprietary study of 6,500 UK shoppers (2026)