Ad servers like Google Ad Manager, AdButler, and Kevel were built in a world where the "product" is content.
Publishers monetize attention: pages, sessions, impressions.
Commerce companies monetize transactions: products, baskets, inventory, margin, fulfillment, returns.
Those two worlds look similar on the surface—"show an ad"—but the economics underneath are completely different. That difference sounds subtle. In practice, it isn't. That's why so many retail media networks hit a ceiling when they try to run commerce monetization on publisher-era ad server infrastructure. And it's why Topsort exists.
The real mismatch: publisher ad serving vs. commerce monetization
Traditional ad servers optimize for ad delivery. They are great at selecting creative, enforcing targeting rules and frequency capping, serving quickly at scale, and tracking impressions and clicks. They're designed to manage inventory of placements—and that works when your monetization unit is impressions.
Commerce is a different job entirely. It requires optimizing for outcomes.
Retail media isn't just "ads on a site." It's a monetization layer that must account for product availability, because promoting out-of-stock SKUs destroys trust; price and margin, because a high bid isn't always a good outcome; conversion probability, basket behavior, category health, and seller dynamics.
Commerce is an economic system, not a content system. It needs economic infrastructure, not just an ad server.
Why "GAM or Kevel + tools" becomes a fragmented stack
Many RMNs start by stitching together an ad server, a sponsored listings module, a reporting layer, a CDP or audience tool, a billing and reconciliation workflow, and manual floor rules in spreadsheets.
It works—for a while. Then complexity compounds.
At scale, the fragmentation becomes visible. Ad ops becomes a bottleneck. Reporting disagreements slow down decisions. New formats take months to ship. Pricing and floors are managed by rules rather than by learning systems. The compounding lift that should be building over time never materializes because the feedback loops are broken across too many systems.
The answer isn't more tools. It's a different core.
The Topsort way: infrastructure built for commerce economics
Topsort isn't just another ad server. We're built as an AI-native retail media infrastructure layer—an ad engine, not just a delivery system. Here's what that means in practice:
Auctions optimized for commerce, not content. In commerce, the auction should account for purchase propensity by context, product constraints like inventory and price changes, long-term yield rather than just winning CPM, and pacing that matches retail cycles. A publisher-era ad server can run an auction—but it typically doesn't have the commerce-native signals and feedback loops required to optimize that auction over time. With Topsort, optimization is the center of the system.
Non-winning auctions are treated as signals, not waste. Retail media performance improves when the system learns from all auction outcomes, not only served impressions. Losing bids tell you bid density. No-fill tells you floor sensitivity. Budget exhaustion tells you demand intensity. Missed conversions tell you context mismatch. Infrastructure that captures and learns from these signals compounds yield continuously.
Designed for automation, not manual ad ops. A common failure mode in RMNs: revenue grows, complexity grows, ad ops headcount grows, margins shrink. Retail media should scale like infrastructure—revenue growing faster than headcount. Topsort is built so optimization is baked into the engine, pacing and pricing have guardrails and automation, self-service doesn't create internal bottlenecks, and APIs fit enterprise workflows.
Global without re-architecting. Publisher-era ad stacks can be global, but retail media global is a different problem. Multi-currency billing, regional tax considerations, marketplace sellers across borders, varying org models, and granular permissions by vendor and business unit. These aren't add-ons you bolt on later. They're design decisions that need to be made at the infrastructure level from day one.
AI that actually moves the needle. A lot of "AI in retail media" ends up as UI copilots—useful, but not where the real value is. The biggest impact comes when AI sits inside the economic core: bid optimization, floor dynamics, pacing, allocation, and yield trade-offs over time. That's what changes outcomes at scale. That's what turns optimization from a feature into a growth engine. That's what Retail Media 3.0 actually means.
The real takeaway
If your business is content, the best ad server is a publisher ad server.
If your business is commerce, the "best ad server" is not a server at all. It's infrastructure that understands the economics of commerce and compounds performance over time.
That's why Topsort looks different from traditional ad-serving platforms like GAM and Kevel. They were built to serve ads. Topsort is built to maximize monetization in commerce systems.
If you're evaluating ad servers or looking for "the best Kevel alternative"
Here are the questions that cut through the noise:
If the answer is yes, you're looking at infrastructure. If not, you're looking at a tool.
Retail media is no longer an experiment. It's becoming a core operating layer for modern retailers. The stacks that win won't be the ones with the most dashboards. They'll be the ones with infrastructure that makes revenue scale reliably, globally, and with compounding lift.
That's Retail Media 3.0.