CTO Series 4 - Why Retail Media Problems Are Commerce-Native, Not Publisher Problems

This is some text inside of a div block.

‍By Francisco Larrain, Co-Founder & CTO, Topsort

A quick lookback before we move forward: CTO Series 1, CTO Series 2, and CTO Series 3

Retail media is often described as “ads on a retailer’s website.” That framing is convenient, but also wrong.

Retail media is not a media business layered onto commerce. It is a commerce business that incorporates advertising as an economic layer.

That distinction matters, because it determines whether your foundation is a publisher ad server or a commerce ad engine.

The Mental Model Mistake

Publisher advertising systems were built for one objective: maximize revenue from attention.

Commerce systems operate under a different objective: maximize long-term economic value across transactions.

At first glance, both involve auctions, but the optimization targets are fundamentally different.

One is media optimization. The other is economic system optimization.

Why These Problems Are Commerce-Native

Retail media sits inside a live marketplace, one with real inventory constraints, dynamic pricing, competitive sellers, category-level elasticity, and margin variability. In a publisher context, the ad slot is the product. In a commerce context, the ad is competing with and influencing the underlying product graph.

Aggressive monetization in commerce carries consequences that simply don't exist in media: reduced organic relevance, distorted category health, eroded advertiser ROI, and long-term damage to shopper trust. This is why retail media optimization must incorporate commerce signals. Publisher-era ad servers were not built for this.

The Structural Difference: Delivery vs. Allocation

A publisher ad server answers: which creative wins this slot？A commerce ad engine answers: how should this opportunity be allocated to maximize long-term economic value?

That allocation decision requires knowing whether the opportunity is even worth monetizing, at what reserve price, whether pacing should adjust, and what the opportunity cost to organic ranking might be. This is closer to marketplace design than media serving, and marketplace problems are structural, not cosmetic.

Floors and pacing illustrate the divide clearly. In a publisher system, floor pricing protects a minimum CPM and pacing ensures even delivery. In a commerce system, dynamic floor pricing must balance yield, fill, advertiser health, and category integrity simultaneously — while pacing responds to demand spikes, seasonal shifts, and real-time auction dynamics. These are commerce-native variables, and they can't be bolted onto a publisher architecture as an afterthought.

Why This Becomes Obvious at Scale

Early-stage retail media can operate on adapted publisher stacks. But as budgets institutionalize and multi-market expansion begins, the cracks appear: manual interventions multiply, auction logic conflicts with conversion logic, reporting fractures across systems, yield plateaus, and ad ops headcount scales linearly with revenue.

These are not feature gaps. They are architectural mismatches. Retail media at scale requires infrastructure designed around commerce economics from day one.

It is not simply programmable serving logic. It is an economic optimization layer embedded inside the commerce stack.

Why This Is a Category Shift, Not a Feature Shift

This is not a debate about feature sets. It is a debate about foundational assumptions.

Publisher systems assume attention is scarce and homogeneous. Commerce systems recognize that not all impressions carry equal opportunity cost, and that not all revenue today is good revenue tomorrow. Retail media is not about extracting CPM. It's about engineering monetization in a live, dynamic marketplace.

Publisher ad servers optimized media. Commerce ad engines optimize markets. That difference defines the next decade of retail media.