CTO Series 3 - Why Retail Media Needs an Ad Engine, Not Just an Ad Server
By Francisco Larrain, Co-Founder & CTO, Topsort
A quick lookback before we move forward: CTO Series 1 and CTO Series 2.
Retail media is entering a new phase. As budgets grow and retail media networks (RMNs) scale globally, one reality becomes clear: static pricing and manual pacing do not survive volatility.
If you are evaluating your retail media tech stack, the question is no longer: “Do we have an ad server?”
The real question is: “Do we have retail media infrastructure capable of dynamic floor pricing and intelligent pacing?”
This is the difference between a publisher-era ad server and a modern retail media ad engine.
Floors and Pacing Are One System
Floors and pacing are often treated as separate features. They are not. They are two levers acting on the same economic system, and when you move one aggressively, the other feels it immediately.
This is why retail media cannot be built as loosely connected tools. A floor pricing module that doesn't know about pacing state and a pacing engine that doesn't account for floor exposure will produce outcomes that neither team can predict or explain.
It must be designed as a unified system.
The Difference Between Tools and Infrastructure
The distinction is not cosmetic. Adding features on top of a delivery-focused architecture doesn't change what the system is optimizing for. An ad server's job is to clear inventory. It has no native concept of revenue health, ecosystem stability, or compounding performance, those concerns live outside it, in spreadsheets and ops workflows and manual overrides.
An ad engine is built around a different objective entirely. Monetization intelligence lives inside the auction core, not above it. That means every pricing and pacing decision is made with full system awareness, not reconstructed after the fact by a human reading a dashboard.
That is what separates an ad server (delivery-focused, publisher-era) from an ad engine (outcome-focused, retail-native). An ad server moves impressions. An ad engine manages the economic health of an entire marketplace.
Why Volatility Breaks Traditional Ad Servers
Retail demand is not stable. Unlike the web publishing inventory ad servers were built for, it shifts continuously, driven by:
- Promotions and seasonal surges
- Seller entry and exit
- Traffic mix changes
- Real-time pricing updates
- Bid density fluctuations
Each shift can render yesterday's floor calibration obsolete before a human operator notices. Traditional ad servers respond reactively — a dashboard is checked, a parameter adjusted, and by the time the correction lands, the market has already moved. Dynamic floor pricing and intelligent pacing close that loop automatically, letting the auction self-correct in real time.
That capability is not optional at scale. It is the foundation the following realities demand.
Why This Is the Future
As retail media budgets institutionalize globally, three realities become unavoidable:
- Volatility is permanent — bid competition will grow more sophisticated and reactive, not less.
- Manual intervention does not scale — every new seller, category, and traffic source adds variability a human workflow cannot reliably track.
- Revenue growth must outpace headcount growth — networks that rely on manual tuning add ops cost in direct proportion to revenue, erasing the model's efficiency advantage.
Dynamic floors and intelligent pacing are not "advanced features." They are the baseline for scalable RMNs. The retailers who treat them as infrastructure, not optional tooling, will win the next decade of retail media.
