A Practical Guide to Retail Media
Retail media is the fastest-growing ad channel in a decade, projected past $150B globally. Here's how retail media networks work, why brands are shifting budget to them, and how to run them without wasting spend.
Ten years ago, “advertising on a retailer’s website” meant a banner nobody clicked. Today it’s the third wave of digital advertising — after search and social — and it’s growing faster than either did at the same stage. Retail media ad spend crossed roughly $140–150 billion globally in 2024 and is on track to overtake television. Amazon’s advertising business alone now runs at over $50 billion a year, most of it retail media.
If you sell physical products — or increasingly, anything — retail media is no longer optional. This guide explains what it is, why the money is moving there so fast, and how to run it without lighting budget on fire.
What retail media actually is
A retail media network (RMN) is an advertising business run by a retailer, using its own first-party shopper data. When you buy sponsored placements on Amazon, Walmart Connect, Target’s Roundel, Instacart, or Kroger Precision Marketing, you’re using a retail media network.
Three things make it distinct from ordinary display advertising:
- It runs on first-party purchase data. The retailer knows exactly what people bought, not just what they browsed. That makes targeting and measurement dramatically more accurate than cookie-based advertising.
- It sits at the point of purchase. You’re reaching a shopper with a credit card already out, inside the store where they’ll check out. Intent doesn’t get higher.
- It closes the loop. Because the retailer sees the sale, it can tell you which ad impressions turned into actual purchases — the “closed-loop measurement” that search and social can only approximate.
Why budget is stampeding into it
Retail media didn’t grow because marketers love another dashboard. It solved three problems hitting at once:
- The death of the third-party cookie. As browser tracking collapses, first-party retailer data became some of the most valuable targeting data left. Retailers were sitting on a goldmine and monetized it.
- Retailers needed the margin. Grocery and general retail run on thin margins; advertising is nearly pure profit. RMNs turned retailers into high-margin media companies, which is why every chain launched one.
- Brands needed proof. After years of fuzzy attribution, closed-loop sales measurement — “this ad drove this purchase” — was irresistible to brands under pressure to justify spend.
The three layers of retail media
Retail media isn’t one ad unit. It spans three environments, and a real strategy uses all three deliberately:
On-site (the highest intent). Ads on the retailer’s own site and app — sponsored product listings, sponsored brands, and search results. This is where most spend goes because it converts: you’re catching the shopper mid-basket. Sponsored products that match search intent are the workhorse.
Off-site (reach beyond the store). The retailer uses its shopper data to target the same customers elsewhere — on the open web, social, and connected TV. You get retail-grade targeting with the reach of programmatic. Useful for driving new demand, not just capturing it.
In-store (the emerging frontier). Digital screens, smart shelves, and audio inside physical stores. Still early, but for brands with shelf presence it closes the last gap between ad and purchase.
How to run retail media without wasting spend
The channel converts, but it’s easy to overpay. The discipline that separates good programs from bloated ones:
Win your own branded terms first, then defend the category. Sponsored placements on your own brand searches are cheap and protect shoppers who already want you from competitors bidding on your name. Then expand to category and competitor terms deliberately.
Separate “capture” from “create” budgets. On-site sponsored products capture existing demand — measure them on ROAS. Off-site and upper-funnel create demand — don’t judge them on last-click ROAS or you’ll starve growth. Mixing the two on one metric is the most common costly mistake.
Mind incrementality, not just ROAS. A high ROAS on branded terms can be an illusion — many of those shoppers would have bought anyway. Ask “what sales did this ad add that wouldn’t have happened?” The best RMNs now offer incrementality testing; use it.
Fix the fundamentals before you spend. On retail media, your product page is your landing page. Great titles, images, reviews, and in-stock status do more for conversion than bid tweaks. Driving paid traffic to a weak listing wastes every dollar. This is where retail media meets conversion fundamentals — measurement only helps if the destination converts.
Don’t spread across every network. Each RMN is a separate platform with its own quirks. Two networks run well beat six run badly. Start where your customers actually shop.
How to measure it
Retail media’s superpower is measurement, so use metrics the channel is uniquely good at:
- ROAS by campaign type — but segmented into capture vs. create, never blended.
- Incremental ROAS — the sales the ad genuinely added, via holdout testing.
- Share of voice / share of search on your key category terms.
- New-to-brand sales — most RMNs report what share of buyers were new customers; the truest sign you’re growing, not just harvesting.
- Total profit, not platform ROAS — factor in retailer fees and margin, because a 6x ROAS on a low-margin SKU can still lose money.
A 30-day starting plan
- Week 1 — Pick one network and audit your listings. Choose where your customers actually buy. Fix titles, images, and reviews before spending a cent.
- Week 2 — Launch defensive branded + top category sponsored products. Small budgets, tight targeting, clear ROAS goals.
- Week 3 — Add a create-demand test. A modest off-site or upper-funnel campaign measured on new-to-brand sales, not ROAS.
- Week 4 — Run an incrementality read. Use a holdout to see what’s genuinely additive, and reallocate away from spend that was just harvesting existing demand.
Frequently asked questions
Is retail media only for brands that sell on Amazon?
No. While Amazon is the largest retail media network, Walmart, Target, Instacart, Kroger, and dozens of others run their own — and off-site retail media lets you use their shopper data to advertise across the web and connected TV. Even non-retail businesses now buy retail media for its first-party targeting.
How is retail media different from Google Shopping ads?
Google Shopping surfaces products in a general search engine; retail media places ads inside a specific retailer’s store, powered by that retailer’s actual purchase data, with closed-loop measurement tying ads to real sales. Retail media sits closer to the point of purchase and measures outcomes more precisely.
What’s a realistic ROAS for retail media?
It varies widely by category and campaign type. Defensive branded sponsored products often show very high ROAS (partly because those buyers were already converting), while demand-creation campaigns show lower ROAS but drive new customers. Judge each by its job, and prioritize incremental sales and profit over headline ROAS.
Do I need a huge budget to start?
No. On-site sponsored products can start small and scale with what works — the first-party targeting makes even modest budgets efficient. The bigger requirement is a well-optimized product listing, since that determines whether your paid traffic converts.
The bottom line
Retail media works because it collapses the distance between the ad and the purchase, and measures what happens in between. But that same precision makes it easy to fool yourself — harvesting existing demand at a great ROAS while mistaking it for growth.
Run it with discipline: fix your listings first, separate capturing demand from creating it, and measure incremental profit rather than platform ROAS. Do that, and retail media becomes the most accountable ad channel you run.
Keep exploring: browse more paid advertising guides, learn how marketing analytics proves incrementality, or see the Digital Business Marketing Awards.
