Performance Marketing · Data

Retail media networks in 2026: why the ROAS numbers don't add up

Retail media stopped being an Amazon-only line item years ago. Brands are now running ads across Amazon, Walmart Connect, Target Roundel, Instacart, Kroger Precision and a growing list of smaller networks, spending real budget on a channel that most of them cannot measure honestly. New 2026 data from eMarketer, Skai and Stratably shows the gap between what these platforms report and what they actually deliver, and it is bigger than most marketing leads assume.

The short answer

US retail media spend hits an estimated $71.09 billion in 2026 and brands now run an average of six retail media networks, projected to reach eleven by year end. Most report last-click ROAS that overstates real performance by 30 to 60 percent. Run fewer networks, demand incrementality data, and budget off holdout tests, not dashboard numbers.

What is actually happening with retail media right now?

Two things are true at once. Retail media is the fastest-growing part of the US ad market, and most of the brands buying into it cannot say with confidence what it is actually returning. eMarketer projects US retail media ad spend will reach $71.09 billion in 2026, up from $60.32 billion in 2025, a jump of roughly 18 percent, and retail media now accounts for close to 30 percent of total US digital ad spend. That is money moving out of Google and Meta budgets and into a category that looks, from a dashboard, like the highest-converting channel a lot of advertisers have ever run.

The catch is on the platform side. Skai and Stratably's 2026 State of Retail Media report found the average brand now works with six retail media networks and expects to be running eleven by the end of the year. Every one of those networks has its own dashboard, its own attribution window, and its own definition of a converted sale. Nobody is reconciling all eleven against each other, and almost nobody is reconciling any of them against what actually happened without the ad.

Why does retail media ROAS look so good on the dashboard?

Because most retail media networks report last-click, in-platform attribution over a short window, often seven or fourteen days, and credit the ad for every purchase that followed a click or an impression, regardless of whether the shopper was already going to buy. If someone searches your brand name on Amazon and buys the product they came to buy, and your sponsored ad happened to be above the organic listing, that sale still counts toward your reported ROAS. It is not a lie, exactly. It is a number that answers a different question than the one marketing leads think they are asking.

Industry data on incrementality, the sales that only happened because of the ad, tells a rougher story. Benchmarking from Osmos puts true incremental ROAS 30 to 60 percent below the last-click ROAS a retail media network reports. A dashboard showing 6x can be delivering closer to 2.4x to 4.2x in real, attributable lift. That gap does not show up anywhere in the platform's own reporting, because the platform has no reason to surface it.

What the dashboard showsWhat it is actually measuringWhy it inflates ROAS
Last-click, in-platform ROASEvery purchase after a click, no matter intentCounts sales that would have happened without the ad
7 to 14 day attribution windowA narrow slice of the actual path to purchaseMisses longer research cycles, overweights recent touch
View-through conversionsPurchases after an impression, not a clickCredits ads for browsing behavior the shopper initiated anyway
Platform-reported ROAS (6.1x average)Gross return before incrementality adjustmentNo holdout, no control group, no baseline demand removed

Stratably's research underneath the same report found that 75 percent of advertisers name incrementality as their single biggest measurement challenge in retail media, and only 15 percent say they feel effective at actually measuring it. That is not a fringe complaint from a handful of skeptical marketers. That is three out of four buyers admitting the number driving their budget decisions is not the number they trust.

Where does the money actually concentrate?

Despite eleven networks being the direction of travel, spend is still lopsided. Market share data compiled by Improvado puts Amazon at roughly 79.7 percent of US retail media ad spend, with Walmart Connect a distant second at about 8 percent and Target Roundel around 1.5 percent. The rest, Instacart, Kroger Precision Marketing, Chewy, Best Buy Ads and dozens of smaller networks, splits the remaining tenth of the market.

NetworkApprox. US shareWhere it earns budget
Amazon Ads~79.7%Any brand selling meaningful volume through Amazon
Walmart Connect~8%CPG and grocery brands with real Walmart distribution
Target Roundel~1.5%Brands with strong Target shelf presence, especially DTC crossover
Instacart, Kroger Precision and other niche RMNs~10.8% combinedOnly where the category and volume justify a dedicated test

The mismatch is the point. If eleven networks are splitting a tenth of the pie beyond Amazon, most of that expansion is chasing marginal reach, not marginal return. A brand adding its seventh and eighth retail media network is very often spreading an already-thin measurement budget across placements too small to test properly.

What we'd do about it

Pick your retail media networks the same way you'd pick a paid search channel: where you already have real sales volume and enough spend to run a proper holdout test. For most small and mid-size brands, that is Amazon plus one, maybe two, secondary networks tied to actual shelf presence. Everything past that is reach you cannot verify, on a budget line that is easy to defend with a screenshot and hard to defend with a holdout.

How do you actually check if a network's ROAS is real?

Run a holdout test. Pause ads on a matched set of markets, SKUs or audience segments for two to four weeks, keep everything else constant, and compare the sales lift in the tested group against the platform's own reported ROAS for that same period and audience. If the gap between reported and incremental performance is small, the platform's number is closer to trustworthy and you can lean on it for day-to-day optimization. If the gap is large, and for most retail media networks it will be, budget against the holdout number going forward, not the number in the dashboard.

This is not a one-time exercise. Retail media attribution windows and reporting methodologies change without much notice, the same way Meta and Google's have this year. A holdout that held up in the spring can drift by fall as the platform adjusts its own algorithm and reporting. Treat incrementality testing as a recurring line item, not a project you finish once and file away.

What should this change about Q4 budgeting?

Holiday season is when retail media spend concentrates hardest, and it is also when the incrementality gap does the most damage, because budget decisions made off inflated ROAS get scaled fastest right when competition and CPCs are highest. Before locking a Q4 retail media plan, run the math on incremental, not reported, ROAS for each network you are considering, and weight new budget toward the network where you have the most confidence in the number, not the highest number on the screen.

The brands that get this right treat retail media the same way they treat Google and Meta: as one channel in a blended measurement model, checked against actual revenue and CAC, not as a category that gets a pass because the platform's own dashboard is flattering.

What we'd do about it

Before adding a new retail media network to next quarter's budget, ask the sales rep for their incrementality methodology, in writing. If they cannot describe how they isolate lift from baseline demand, treat their reported ROAS as a ceiling, not a fact, and size your test budget accordingly.

Frequently asked questions

How many retail media networks do brands run in 2026?

The average brand works with about 6 retail media networks in 2026 and expects that to reach 11 by year end, according to Skai and Stratably's 2026 State of Retail Media report. Most of that growth is optional. Two or three well-chosen networks beat eleven poorly measured ones.

Is retail media ROAS reliable?

Not on its own. Most retail media networks report last-click, short-window ROAS that counts sales that would have happened anyway. Industry benchmarks put true incremental ROAS 30 to 60 percent below the number on the dashboard, so a reported 6x can be closer to a real 2.4x to 4.2x.

How big is retail media spend in 2026?

US advertisers are projected to spend 71.09 billion dollars on retail media in 2026, up from 60.32 billion in 2025, according to eMarketer. Amazon holds roughly 79.7 percent of that spend, with Walmart Connect and Target Roundel a distant second and third.

What is incremental ROAS and why does it matter?

Incremental ROAS measures sales that only happened because of the ad, using a holdout or geo test, instead of every purchase a customer who saw the ad happened to make. It matters because last-click ROAS credits ads for baseline demand, inflating the number retail media networks put in front of you.

Should a small business run ads on multiple retail media networks?

Only where you already sell meaningful volume. Run ads where your product sits on shelf or in the marketplace and the platform can show incremental lift, not everywhere it lists SKUs. Chasing every network with a self-serve dashboard usually just splits a small budget too thin to measure anything.

How do you check if a retail media network's reported ROAS is real?

Run a holdout test: pause ads on a matched set of markets, SKUs or audience segments for two to four weeks and compare sales lift against the network's reported ROAS for the same period. If the gap is small, trust the dashboard more. If it is large, budget against the holdout number, not the platform number.

The takeaway

Retail media is not a scam and it is not overhyped as a category. It is genuinely one of the better-converting channels available right now, and $71 billion in US spend this year proves buyers believe that. The problem is not the channel, it is that most brands are budgeting off a number the platform has every incentive to inflate and no obligation to correct. Run fewer networks, demand real incrementality data before you scale a new one, and treat every reported ROAS as a starting claim to verify, not a finished answer.

Sources & further reading

Rahul Gupta

Founder of HyberX, a digital growth agency working with brands across the US, Europe, the Middle East and India. Writes on web design, paid media and conversion optimisation.

More about Rahul · LinkedIn

Related reading

Want a media plan that budgets off real lift, not dashboard ROAS?

We build blended measurement across Google, Meta and retail media so your budget follows what actually moves revenue.

Book a Growth Call