Retail media costs in 2026: the $71 billion shelf

Published October 24, 2026. Retail media means advertising bought through a retailer: sponsored products in its search results, display on its site and app, ads in its stores, and off-site ads targeted with its shopper data. Amazon Ads, Walmart Connect, Target Roundel, and Instacart Ads are the names most US advertisers will meet first.

How big the channel is in 2026

EMARKETER's December 2025 forecast put US retail media spending at $60.32 billion in 2025 and $71.09 billion in 2026 — growth of 17.8% year over year, outpacing both social network and search advertising growth in the same forecast. Off-site retail media alone — retailer data used on other sites, social platforms, and connected TV — is forecast by EMARKETER to reach $17.98 billion in the US in 2026.

The concentration matters more than the total. Amazon Ads held 79.7% of US retail media spending in 2025, according to EMARKETER, with Walmart Connect second at 8.0% and Target Roundel at 1.5%. Amazon and Walmart together are expected to capture 89% of the incremental spending in 2026. For most brands, "retail media strategy" still means an Amazon decision first and a Walmart decision second, with every other network competing for a test budget. Our CTV advertising costs guide covers the streaming side of that off-site growth in detail.

Demand signals through 2026 support the forecast rather than softening it. Tinuiti's Q1 2026 benchmark report found Amazon DSP spending among its advertisers rose 41% year over year — the fastest growth since Q3 2021 — while Walmart Sponsored Products spending rose 62%, with clicks up 57% and cost per click up only 3%. Read that second figure carefully: heavy new demand arrived at Walmart without a matching jump in click prices, at least in that quarter and that advertiser set.

What it costs, format by format

There is no rate card for the channel as a whole. Each retailer runs its own auction, its own formats, and its own fees, and category competition can move a price more than the format label does. The ranges below are published 2026 benchmark ranges from Improvado's retail media network comparison — planning figures, not quotes. Your category, season, and retailer will move you inside or outside them; the ad cost calculator will help you model a total once you have a retailer's own proposal.

Format	How it is bought	2026 benchmark range	Typical benchmark performance
Sponsored products / search	Per click (CPC auction)	About $0.30 in low-competition categories to $3.00+ for contested keywords	CTR roughly 0.5–1.2%; ROAS roughly 2.5x–6.0x by category
On-site display	Per thousand impressions (CPM) or per click	CPM roughly $8–$25	CTR roughly 0.2–0.5%; ROAS roughly 1.8x–4.0x
Off-site (social, open web, CTV)	Per thousand impressions (CPM)	CPM roughly $5–$15	CTR roughly 0.15–0.4%; ROAS roughly 1.5x–3.5x

Three warnings belong on that table. First, sponsored search is where most retail media money goes and where prices rise fastest, because the inventory is the retailer's own search page and every rival brand wants the same first row — a pattern advertisers already know from our Google Ads costs guide, with the difference that here the auction owner also sells the product's shelf space. Second, headline CPC and CPM are not the full cost: managed-service arrangements, platform or data fees where a network charges them, creative production, and the staff time to run several separate retailer platforms all sit on top of media. Ask each network to state every fee in writing before you compare two proposals on media price alone. Third, minimums and buying routes vary widely by network and change over time — self-serve entry on a large marketplace is a different commitment from a managed programme on a smaller network, so treat any published minimum as a question to confirm, not a fact to budget on.

The measurement trap: closed loop, open definitions

Retail media's selling point is closed-loop measurement: the retailer saw the ad exposure and the purchase, so it can join them. That is a real advantage over channels that guess. It is not the same as comparable measurement.

Each network defines attribution in its own way — different lookback windows, different treatment of view-through conversions, different labels for the same sale. A ROAS reported by one retailer is therefore not the same measurement as a ROAS reported by another, even when both are honestly calculated. EMARKETER's November 2025 forecast summary put it bluntly: do not expect measurement relief any time soon. Some retailers have begun standardising their own rules internally; the industry has not standardised across retailers.

The practical consequence for 2026 budgets is that platform-reported ROAS can tell you how a campaign is trending inside one retailer, but it cannot rank retailers against each other, and it cannot tell you how many of those sales would have happened without the ad. That second question — incrementality — is the one to spend testing money on: holdout areas, paused-brand-term tests, and comparisons against total sales, not just attributed sales. It is the same discipline our midterm ad costs guide applies to a temporary price squeeze: know which part of the number is the market and which part is the measurement.

How to enter without overbuying the shelf

1. Start where you already sell. Retail media converts because the shopper can buy immediately. A network where you have no listing, weak stock, or poor delivery terms will tax you for traffic you cannot close. Distribution first, media second.
2. Fund sponsored search before display. Search placements meet a shopper who typed a product need. Display and off-site buy reach and retargeting at lower benchmark returns. Most new entrants should prove the search economics in one category before buying the wider funnel.
3. Set your ROAS floor from your own margin. A retailer benchmark is an average across other brands' margins. Your floor is arithmetic: at a given gross margin, there is a return below which the ad sale loses money after product costs. Write that number down before the first auction, not after the first invoice.
4. Budget the test for incrementality. Reserve a portion of the first quarter's spend for a holdout or pause test. Without it, you will renew on attributed ROAS alone — the metric the platform grades itself with.
5. Expect a one-network operation at first. Every added network adds a platform, a report format, and an attribution definition. Two networks run well will teach you more than five run thin, and the concentration figures above mean the first network will usually be the one where most of your customers already shop.

Frequently asked questions

Sources & methodology

• EMARKETER (Jan 9, 2026): FAQ on retail media networks: how marketers should allocate budgets in 2026 (December 2025 forecast: $60.32B in 2025, $71.09B in 2026, +17.8%; Amazon 79.7%, Walmart 8.0%, Roundel 1.5% in 2025; Amazon + Walmart 89% of 2026 incremental spend; measurement standardisation warning)
• EMARKETER (2026): US off-site retail media ad spending forecast, $17.98 billion in 2026, via EMARKETER retail media coverage.
• Improvado (2026): Top 15 retail media networks 2026: rankings & benchmarks (format benchmark ranges for sponsored products, display, and off-site used in the table; ranges vary by category and retailer)
• Tinuiti (2026): Digital Ads Benchmark Report Q1 2026 (Amazon DSP spend +41% YoY among Tinuiti advertisers; Walmart Sponsored Products spend +62%, clicks +57%, CPC +3%)

Data checked October 24, 2026. Market totals are EMARKETER forecasts, not final reported spending, and definitions of retail versus broader commerce media differ between forecasters. Format ranges are cross-category benchmarks: your retailer's auction, in your category, in your season, is the only price that counts as a quote.

Go deeper with AI

This article is the starting point. Copy any of these prompts into your favorite AI assistant to learn more about retail media advertising costs in 2026:

🔎 Build a first retail media budget from your margin

I sell [product] on [retailer] at $[price] with a gross margin of [X]%. Using 2026 benchmark ranges — sponsored-product CPCs about $0.30 to $3.00+, display CPM $8–$25, off-site CPM $5–$15 — build me a $[budget] test plan split between sponsored search and display, calculate the ROAS floor my margin requires, and show how many clicks and sales the plan needs to break even. Flag every assumption.

🧭 Interrogate a retailer's proposal before you sign

Here is a retail media proposal from [network]: [paste proposal]. List every cost that is not media (platform, data, managed-service or creative fees), every measurement definition I should ask them to state in writing (attribution window, view-through rules, what counts as a conversion), and five questions that would reveal whether their reported ROAS can be compared with my other channels.

💡 Design an incrementality test you can actually run

Design an incrementality test for my retail media spend of $[X] per month on [retailer]: one holdout or pause design I can run without a data science team, what total-sales metric to compare (not just attributed ROAS), how long it needs to run for a category that sells [units] per week, and how to read the result if attributed sales fall but total sales do not.

Tip: replace the bracketed parts with your own situation — the more specific your prompt, the more useful the answer.

The bottom line

Retail media grows to a forecast $71.09 billion in the US in 2026 because it sits next to the purchase and can prove a sale happened. Its prices are auction prices — sponsored clicks from cents to several dollars, display and off-site in the $5–$25 CPM band on published benchmarks — plus fees that vary by network, in a market where Amazon takes roughly four-fifths of spending. The channel's weakness is not reach or intent; it is that every retailer grades its own homework with its own attribution rules.

The single most important takeaway: buy retail media for its proximity to purchase, but judge it with your own numbers — a margin-based ROAS floor and at least one incrementality test — because platform-reported ROAS cannot be compared across retailers and cannot tell you which sales were truly new.

Want better prompts, in your inbox?

One sharper AI prompt + one new ad cost topic every week. Free forever, unsubscribe anytime.

Subscribe

We never share your email. One email a week, that's the deal.