The Retail Media Stack

Amazon is the third largest advertising platform in the United States. Most people find this surprising, because they think of Amazon as a place to buy things. The two facts are not in tension — they are the same fact. Amazon sells things, which means Amazon knows what people buy, which means Amazon can sell that knowledge to advertisers. The business that emerged from that logic is called retail media, and it is reshaping how digital advertising works.

First-Party Data

First-party data is information a company collects directly from its own customers through its own interactions. When you buy something at Walmart, Walmart has first-party data about that transaction. When you search on Google, Google has first-party data about that search. When you scroll Instagram, Meta has first-party data about what you paused on.

The reason first-party data has become the defining concept in advertising is the decline of third-party cookies. For two decades, advertisers tracked users across the web using small files that any site could read — a user visiting a running gear site could be followed by running shoe ads on every other site they visited. Browsers are now blocking this.

Without third-party tracking, advertisers must rely on platforms to do the targeting using their own first-party data. This concentrates power with whoever has the largest and most relevant first-party datasets — which is why Google, Meta, and increasingly Amazon and Walmart, are positioned so strongly. Everyone else's targeting capabilities are diminishing while theirs remain intact.

What Retailers Know That Platforms Don't

Meta knows you looked at running shoes. Amazon knows you bought them in March, that you also bought protein powder and a foam roller around the same time, and that six months later you bought replacement laces — suggesting you are actively running, not just browsing. That is a meaningfully different signal.

Purchase history is behavioral proof rather than behavioral inference. It also carries higher intent certainty: someone on Walmart.com searching for a product is in a buying context. Someone scrolling Instagram may or may not have any purchase intent at that moment. The same demographic profile means something different depending on what the person was doing when the data was collected.

The Two Products

Retail media platforms sell two distinct products.

Sponsored listings (on-site): When someone searches "protein powder" on Walmart.com, the first three results may be paid placements. The brand paid to appear prominently when a user is actively searching within the retailer's environment. This is extremely high-value inventory — the user is at the bottom of the purchase funnel, one click from a sale. It is structurally identical to Google search ads: the user expressed intent, and the advertiser paid to be visible at that moment.

DSP (off-site): The retailer licenses their purchase data to target ads across the open web. A brand can use Walmart's DSP to reach people who have bought running gear at Walmart in the last 90 days — and show them ads on news sites, apps, and other platforms outside Walmart entirely. The ad appears somewhere completely unrelated to Walmart. The targeting signal came from Walmart's transaction history.

These serve different purposes. Sponsored listings capture demand that already exists. DSP campaigns create or maintain demand before someone is actively shopping.

The Closed-Loop Advantage

This is the structural advantage that makes retail media uniquely valuable, and it resolves the attribution problem described in How Digital Advertising Actually Works.

When a brand advertises on Meta and a user later buys their product on Amazon, Meta cannot observe that purchase. The advertising and the transaction happened on different platforms. The loop is open. Meta can build probabilistic models of which impressions likely caused sales, but it cannot close the chain with direct evidence.

When a brand advertises through Walmart's DSP and a user later buys their product at Walmart, Walmart can observe that purchase directly. The same company owns the ad platform and the point of sale. The loop is closed. The advertiser receives proof of purchase, not a model of probable purchase.

This closed-loop measurement is worth a significant premium to advertisers, particularly in consumer packaged goods where products are sold primarily through retailers. A brand can show that their Walmart media spend produced a measurable lift in Walmart sales, matched at the individual transaction level. That kind of direct attribution is not available from any other channel, and it justifies paying more per impression than an equivalent placement elsewhere.

The closed loop only closes because the retailer owns both the ad platform and the point of sale. When those are different companies — a brand advertising on a streaming service to drive grocery sales — the data never naturally meets. That is the problem data clean rooms were built to solve.

Amazon as the Case Study

Amazon Ads grew from a side project into a $46 billion annual business. Several factors explain the scale.

Intent signals: People on Amazon are actively shopping. The gap between an ad impression and a purchase can be seconds. Conversion rates are structurally higher than on social or display platforms where users are not in a purchasing mindset.

Closed-loop measurement: As described above, Amazon can prove that an ad caused a sale. This is the most valuable thing an ad platform can offer.

Platform necessity: For brands selling on Amazon, not advertising on Amazon means being invisible while competitors are visible in the same search results. The platform creates advertising demand by being the storefront — brands effectively pay a toll to be seen in their own category.

The combination of high intent, closed-loop measurement, and near-mandatory participation for brands in many categories makes Amazon's ad inventory among the most efficiently priced in digital advertising.

The Pattern Repeating

Walmart Connect, Target Roundel, Kroger Precision Marketing, CVS Media Exchange, Instacart Ads — every major retailer with sufficient transaction volume is building the same infrastructure. The logic is the same in every case: purchase data is an asset, advertisers will pay to use it, and closed-loop measurement justifies a premium that general-purpose ad platforms cannot match.

The pattern is now extending beyond retail. Airlines are selling targeting based on travel behavior. Banks are building ad products based on spending category data. Streaming services are selling targeting based on viewing history. Any company with a large enough first-party dataset that is specific enough to be useful to advertisers is building or seriously considering an ad business.

The underlying principle is the same one that built Google and Meta: whoever controls the data that tells advertisers where their money will work hardest, captures a significant portion of the advertising market. First-party purchase and behavior data is the new oil, and the companies sitting on the largest reserves are discovering that the refinery is more profitable than the product.

Part of a series: The 150ms Ad Auction · Engineered to Keep You There · How Digital Advertising Actually Works · Ad Inventory and the Attention Economy · Matching Without Meeting