Retail media networks have moved from a curiosity inside Amazon’s ad console to a line item that retail finance teams now defend in board meetings. The category crossed roughly $140 billion in global spend in 2024 and keeps compounding at double-digit rates, which means the question for most brands is no longer whether retail media exists but how much of your paid ads budget belongs there versus Google, Meta, or your own site. This guide answers that with concrete allocation math, vendor comparisons, and the operational traps that quietly drain margin.
If you sell physical goods, retail media touches your business whether you buy ads or not, because competitors bidding on your branded terms inside Amazon search are pulling shoppers off your product pages. Treating it as optional is the expensive mistake. Our retail marketing guide frames where retail media sits in the wider mix, and this piece drills into the buy-or-skip decision for each major network.
In short
- Retail media networks (RMNs) sell ad placements on a retailer’s own properties (search results, product pages, apps) using that retailer’s first-party purchase data for targeting.
- Amazon Ads dominates with the deepest demand signal, but Walmart Connect, Target Roundel, Kroger Precision, and Instacart are now credible second buys depending on your category.
- Expect a healthy return on ad spend (ROAS) of 3x to 6x on sponsored product placements when your detail pages convert; below 2x usually signals a listing or price problem, not a bidding problem.
- Closed-loop attribution is the real prize: RMNs match ad exposure to verified sales, which de-risks budget shifts off platforms where attribution is collapsing.
- Start with sponsored products on one network, prove unit economics for 60 days, then expand to sponsored brands and off-site display only after the core converts.
What is a retail media network and why does it convert
A retail media network is an advertising business a retailer runs on top of its commerce platform, monetizing the attention of shoppers who are already in a buying mindset. The mechanics are simple: you bid for placement against keywords or product categories, the retailer serves your ad to logged-in shoppers, and because the shopper transacts in the same environment, the network reports exactly which clicks produced sales.
That closed loop is why RMN conversion rates routinely beat upper-funnel social. A shopper typing “wireless earbuds” into Amazon has higher purchase intent than one scrolling a feed. The retailer also knows that shopper’s prior purchases, so targeting is grounded in actual spend rather than inferred interest. As third-party cookies degrade across the open web, this first-party purchase data becomes the most durable targeting asset in advertising, a shift we examine alongside organic discovery in our analysis of SEO for retailers.
The trade-off is control. You rent the retailer’s audience and its rules, you cannot retarget those shoppers off-platform without buying the network’s off-site product, and the retailer competes with you for margin on the same transaction. Retail media is rented reach with rented data, and the rent is rising.
Which retail media networks should you actually buy
The answer depends on where your category’s transactions concentrate. Spend follows shoppers, so the network with the most relevant baskets wins your first dollar. Below is a working comparison for a mid-market consumer brand evaluating 2026 buys.
| Network | Best for | Typical sponsored-product ROAS | Minimum to test | Off-site display maturity |
|---|---|---|---|---|
| Amazon Ads | Almost any DTC or CPG good | 3x to 6x | $5,000 over 60 days | High (DSP) |
| Walmart Connect | Grocery, household, value brands | 3x to 5x | $5,000 over 60 days | Medium |
| Target Roundel | Apparel, home, beauty | 2.5x to 4x | $10,000 (managed) | Medium |
| Kroger Precision | Food, beverage, CPG | 3x to 5x | Managed engagement | Medium (84.51 data) |
| Instacart Ads | Grocery, perishables, impulse | 4x to 7x | $2,500 self-serve | Low |
Read the table as a starting hypothesis, not a ranking. A premium skincare brand should weight Target Roundel and Amazon over Walmart, while a shelf-stable snack should test Instacart and Kroger first. The discipline is to follow your own category’s sales concentration, not the network with the loudest sales team.
One pattern holds across networks: brands that win treat the retailer’s search results as an extension of search engine optimization. The same intent-matching logic that governs Perplexity and Google AI Overviews applies inside Amazon’s box, where relevance and conversion history rank your listing. We unpack that intent layer in our piece on Perplexity and Google AI Overviews, and the takeaway transfers: optimize the destination before you pay for the click.
How to build your first retail media plan
Answer-first: run a tightly scoped 60-day test on one network, measure incremental sales not just reported ROAS, and only scale ad types that clear your contribution-margin floor. Here is the sequence that works for teams without a dedicated RMN specialist.
- Fix the destination first. Audit your top 10 product detail pages for title clarity, A-plus content, review count above 50, and competitive price. Paid clicks to a weak page burn budget; expect a 30 to 50 percent conversion lift from listing fixes alone before any media spend.
- Pick one network using the category-concentration logic above. Resist the urge to spread $5,000 across three networks, which produces three inconclusive tests.
- Launch sponsored products only. Start with exact-match and phrase-match campaigns on your highest-margin SKUs, set a target ACoS (advertising cost of sales) equal to your gross margin minus your profit floor.
- Harvest search terms weekly. Move converting terms into exact-match campaigns, add non-converting terms as negatives. This single habit drives most of the efficiency gains in the first quarter.
- Measure incrementality at day 30. Compare advertised SKU velocity against a non-advertised control set. If sponsored sales merely cannibalize organic, you are paying for traffic you already had.
- Expand deliberately. Add sponsored brands for category defense, then off-site display only once your core ROAS is stable and you can attribute view-through lift.
This ladder keeps spend tethered to evidence. Most budget waste happens when brands buy every ad format on day one and cannot isolate which one actually moved units. A staged test-and-scale rhythm keeps retail media complementing rather than cannibalizing your other paid ads, and it gives finance a clean story about why each incremental dollar earned its place.
Budget allocation: how much should go to retail media
For most physical-goods sellers, retail media now warrants 15 to 35 percent of total paid media, with the high end reserved for brands whose sales concentrate on marketplaces. The rule of thumb: allocate in proportion to where conversions happen, then add a defensive premium for branded terms competitors will otherwise steal.
A useful framing splits spend into three buckets. Defensive spend protects your branded search and your own product pages from competitor conquesting. Offensive spend bids on category and competitor terms to win new shoppers. Discovery spend funds sponsored brands and off-site display to build awareness inside the retail environment. A healthy mid-market split often runs 40 percent defensive, 45 percent offensive, 15 percent discovery until the brand has scale to invest more upper-funnel.
Retail media also reshapes the economics of physical retail itself, because the same first-party data now informs assortment and shelf decisions. The operational realities of running brick-and-mortar alongside online sales, including the cost structures that constrain how aggressively you can discount, are covered in our look at the boring truths of main street retail.
Reading the ad formats: where each dollar belongs
Answer-first: sponsored products carry the load for direct response, sponsored brands defend your shelf and category, and display builds awareness, so map each format to the job you actually need done before you fund it. Lumping all three into one ROAS target is the fastest way to misjudge what is working.
Sponsored products are keyword-targeted ads that place a single SKU inside or beside organic search results. They are the workhorse of retail media because intent is highest at the moment of search, and they are the only format most brands should run in month one. Bids are managed at the keyword and SKU level, which gives you granular control over ACoS by margin tier.
Sponsored brands (formerly headline search ads on Amazon) promote a logo, a custom headline, and a small product collection at the top of results. Their value is defensive and category-level: they keep competitors from owning the visual top of the page on your branded and category terms. Judge them on share of voice and category-term capture, not pure last-click ROAS, because much of their value is preventing a competitor’s conquest.
Sponsored display and off-site display retarget shoppers who viewed your product or a competitor’s, and increasingly reach audiences across the open web using the retailer’s purchase data. This is consideration and awareness work. Measured by sponsored-product standards it will look weak, but measured by view-through lift and new-to-brand percentage it often justifies its place once your performance core is stable.
A practical rule: do not buy a new format to fix a problem the previous format created. If sponsored products underperform, the answer is usually a better listing or a sharper negative-keyword list, not a layer of display on top of leaky demand capture.
The data trade-off retailers do not advertise
Answer-first: when you advertise inside a retail media network you gain conversion-grade attribution but surrender the customer relationship and most of the first-party data, so weigh the lifetime cost of renting that audience against the short-term efficiency.
The retailer sees who bought, what else was in the basket, how often they reorder, and which competing products they considered. You see an aggregated report. That asymmetry is the price of admission, and it compounds: every sale you drive through the network deepens the retailer’s data advantage while leaving your own customer file no richer. For brands trying to build durable direct relationships, this is the strongest argument for keeping a meaningful share of spend on owned-site channels even when marketplace ROAS looks better on paper.
There is also a structural tension worth naming. Many large retailers sell their own private-label products that compete directly with the brands buying their media. The same platform that takes your ad budget may be using your category data to sharpen a house brand that undercuts you. This does not make retail media a bad buy, but it should temper any assumption that the network is a neutral marketplace. Treat it as a powerful, partly adversarial channel and price the relationship accordingly.
The same first-party signal that powers retail media is reshaping discovery far beyond the retailer’s own search box, as AI assistants increasingly answer product questions before a shopper ever reaches a results page. That is why the intent discipline behind Perplexity and Google AI Overviews matters even to a pure-play marketplace seller, and why we recommend reading those surfaces as connected rather than separate problems.
A 90-day retail media roadmap with numbers
Answer-first: spend the first month fixing destinations and launching one network, the second month harvesting search terms and proving incrementality, and the third month scaling winners and adding a second format only where the math clears your margin floor.
In days 1 to 30, audit and rebuild your top product pages, then launch sponsored products on a single network with roughly 70 percent of the test budget on exact-match terms you already rank for organically and 30 percent on phrase-match discovery. Set ACoS targets per SKU equal to gross margin minus your profit floor. Expect noisy data; resist the urge to optimize daily.
In days 31 to 60, run weekly search-term harvests, promote converting queries to dedicated exact campaigns, and add the dead weight as negatives. Stand up a holdout set of comparable SKUs and begin tracking velocity differences. By day 60 you should know your true incremental ROAS, not just the platform-reported figure, and you should be able to name your three best and three worst keywords.
In days 61 to 90, scale budget into proven exact-match campaigns, introduce sponsored brands to defend your category terms, and only then evaluate display against a separate awareness goal. If incrementality is weak after 60 days of clean testing, the problem is almost never the network: it is price, listing quality, or review depth, and no amount of media buys around those gaps. The discipline of scaling only what the holdout proves keeps retail media accountable in a way that few other paid channels manage.
The metrics that actually decide your budget
Answer-first: track ACoS against margin, new-to-brand percentage, incremental ROAS, and share of voice on your core terms, because those four numbers tell you whether spend is profitable, growing your base, genuinely additive, and defending your shelf. Reported ROAS alone hides all four answers.
ACoS (ad spend divided by ad-attributed revenue) is your efficiency dial, but it only means something next to your gross margin. A 25 percent ACoS is a triumph on a 60 percent margin product and a loss on a 20 percent margin one, so set targets per SKU tier rather than across the catalog. New-to-brand percentage tells you how much of that revenue came from shoppers who had not bought you in the prior year; a high figure justifies treating some retail-media spend as acquisition rather than pure efficiency.
The fourth number, share of voice, captures how often your products appear in the top placements for your defended terms. Losing share of voice on branded queries is an early warning that a competitor is conquesting you, and it usually shows up before the revenue dip does. The discipline that makes organic discovery durable applies here too, which is why we keep pointing brands back to the fundamentals in our guide to SEO for retailers: ranking and relevance inside a retailer’s box reward the same listing quality that wins organic placement.
| Metric | What it answers | Healthy range | Warning sign |
|---|---|---|---|
| ACoS vs margin | Is this profitable? | ACoS below (margin minus profit floor) | ACoS approaching gross margin |
| New-to-brand % | Am I acquiring or harvesting? | 20% to 40% on offensive terms | Near 0% on category terms |
| Incremental ROAS | Is spend truly additive? | Within 30% of reported ROAS | Far below reported ROAS |
| Share of voice | Am I defending the shelf? | Top 3 on branded terms | Declining week over week |
Common mistakes
The failure modes in retail media are predictable, which makes them avoidable. The most expensive is buying media before fixing the listing: a 1.5x ROAS almost always traces to a thin detail page or an uncompetitive price, not a bidding error, so brands keep raising bids on a page that cannot convert.
A close second is trusting reported ROAS as incremental ROAS. Networks count sales from any shopper who clicked an ad, including shoppers who would have bought anyway. Without a control group, you can report a glowing 5x while generating almost no new revenue. Always validate against held-out SKUs.
Other recurring errors: spreading a small test budget across multiple networks and learning nothing from any of them; ignoring search-term reports so spend leaks into irrelevant queries; setting one ACoS target across SKUs with wildly different margins; and treating off-site display as a performance channel when it is fundamentally an awareness play that needs view-through measurement. Each of these quietly erodes the margin that made retail media attractive in the first place.
FAQ
Is retail media worth it for a small brand under $1M in revenue?
Often yes, but only on one network and only on your best SKUs. A small brand should run a $2,500 to $5,000 sponsored-products test on the network where its category transacts most, focus on three to five high-margin products, and measure incremental sales over 60 days. The closed-loop attribution suits small teams because it shows exactly which spend produced verified sales, which de-risks a limited budget. Skip off-site display and multi-network expansion until the core test proves positive unit economics.
What ROAS should I expect from Amazon sponsored products?
For sponsored products with healthy detail pages and competitive pricing, expect a reported ROAS in the 3x to 6x range. Below 2x usually signals a listing, review, or price problem rather than a bidding issue, so audit the destination before raising bids. Remember that reported ROAS overstates incremental value because it counts shoppers who would have purchased anyway. Validate periodically against a non-advertised control set to understand the true lift your spend generates.
How is retail media different from Google Shopping ads?
Google Shopping drives shoppers to your own site, where you own the checkout, the customer data, and any retargeting. Retail media keeps the shopper inside the retailer’s environment, so the retailer owns the transaction and the data while you rent placement. The upside of retail media is intent and closed-loop attribution; the downside is that you cannot easily build a direct relationship with that customer. Most brands run both, using Google for owned-site growth and RMNs to win marketplace shelf.
Should I buy off-site display from a retail media network?
Only after your on-site sponsored products are converting reliably. Off-site display (served on the open web using the retailer’s purchase data) is an awareness and consideration play, not a direct-response channel, so judging it by sponsored-product ROAS will make it look like a failure. If you do buy it, measure view-through lift against a control and set a separate, awareness-appropriate goal. For most mid-market brands, off-site display is a phase-two investment once the performance core is stable.
How do I measure whether retail media is incremental?
Use a holdout: designate a set of comparable SKUs that receive no media support, then compare velocity between advertised and non-advertised products over the same window. If your advertised SKUs grow meaningfully faster, the spend is generating incremental demand. If both grow at similar rates, your ads are largely cannibalizing organic sales you already had. Some networks now offer brand-lift and incrementality studies; use them, but a simple internal control set is a reliable starting point.
Can retail media data help my brick-and-mortar decisions?
Yes. The first-party purchase data underpinning RMNs reveals which products, baskets, and shopper segments perform in real transactions, which informs assortment, pricing, and promotion in physical stores. Retailers like Kroger and Walmart explicitly sell this insight layer alongside media. Brands that connect online demand signal to in-store assortment make sharper buying decisions and reduce dead stock. Treat the data as a strategic asset, not just an advertising byproduct.
What’s next
Pick one network this quarter, fix your top product pages, and run a disciplined 60-day sponsored-products test with a holdout so you measure incremental lift rather than vanity ROAS. As AI-driven discovery reshapes where shoppers start their journeys, the brands that win retail media will be the ones already optimizing for intent across surfaces, a shift worth tracking through our coverage of how retail news shapes the global e-commerce industry. For the wider playbook on connecting paid, organic, and social into one durable system, revisit our retail marketing guide once you have live test data, and pair it with the measurement standards the IAB publishes for retail media so your reporting holds up under finance scrutiny.