Artificial intelligence is reshaping digital advertising, but not always in the ways people expected. As more of the execution layer becomes automated – from bidding and pacing to audience modeling and in-flight optimization – the real advantage shifts upstream. What matters most is no longer just the technology making decisions, but the signals feeding it.
This is the quieter tension in AI-driven media buying. The tools democratize. The signals don’t.
For retail media networks, the infrastructure powering your ecosystem matters, but only as much as the data fueling it. The question isn’t how sophisticated your stack is. It’s what’s running through it.
AI Optimizes. Data Differentiates.
The benefits of AI in media buying are hard to ignore. Planning timelines are shrinking, campaigns can adjust in real time, and work that once required large trading teams is increasingly automated. It’s a genuine step forward.
The challenge is that when everyone has access to the same set of tools, it stops being a differentiator. When platforms optimize against the same modeled conversions, behavioral signals, and lookalike audiences, competitive edge flattens even if performance improves. Retailers are particularly exposed here. Media infrastructure is becoming more accessible, but when networks feed similar inputs into similar models, bidding logic begins to converge.
Structural advantage in an AI-optimized world belongs to whoever controls data that competitors can’t access or replicate.
The Store as a Signal Generator
In-store purchase data is different from virtually every other signal in digital advertising, and that difference becomes more consequential as AI takes over execution.
Unlike intent or engagement proxies, a transaction represents a verified outcome. A product was purchased, placed into a basket alongside other items, and tied to a loyalty account with years of purchase history behind it. The checkout becomes a clear record of behavior — one of the closest things commerce has to a scoreboard.
What makes this data particularly valuable is the context surrounding the purchase. Basket composition reveals which products are bought together, how shoppers move between brands within a category, and how purchasing patterns evolve. While e-commerce captures detailed behavioral signals within a retailer’s environment, in-store transactions still represent a significant share of commerce and reflect decisions made across the broader competitive shelf. As discovery fragments across social, streaming, and AI-driven interfaces, the transaction remains the clearest signal of what ultimately made it into the cart.
As digital proliferates, the physical store doesn’t fade as a strategic asset. It becomes a more valuable one.
From Closed-Loop Reporting to In-Flight Intelligence
Retail media established its credibility by connecting media exposure more directly to sales outcomes. Closed-loop reporting made it accountable. By linking ad exposure to purchase data, retailers could show brands how media influenced transactions inside their ecosystems. That performance narrative helped justify the premium CPMs retail media networks command today.
While still a powerful capability, it’s increasingly table stakes. The next phase is feeding deterministic purchase signals back into optimization models while campaigns are still running, which is a materially different level of intelligence.
When SKU-level transaction data, basket composition patterns, and loyalty-linked purchase histories inform bidding logic, optimization becomes far more grounded in real behavior. Suppression strategies can exclude recent purchasers with greater precision than probabilistic signals allow. Frequency can align with actual purchase cycles. Offsite bids can reflect observed conversion patterns rather than relying solely on proxies. Instead of optimizing toward engagement signals that may or may not translate to revenue, brands can begin optimizing toward incremental category growth.
Using purchase data to report on what happened is valuable. Using it to shape optimization while campaigns are still running is where the real advantage begins. That’s the shift Infillion’s acquisition of Catalina makes possible – bringing decades of loyalty-linked, SKU-level purchase data out of the measurement layer and into the model itself.
Owning the Layer That Drives the Model
Infrastructure is only as defensible as the intelligence it runs on. When RMNs rely on similar optimization signals, competitive advantage narrows regardless of how sophisticated the underlying technology is. The purchase data may exist, but if it’s passing through systems without shaping the underlying model, it isn’t working. The network looks differentiated on paper and performs like everyone else in practice.
Retailers who own that intelligence layer can configure optimization toward what actually matters – margin, purchase frequency, incremental category growth – and extend that value to their advertisers. They can build bidding models tied to in-store conversion thresholds that others can’t replicate, because others don’t have the same data.
As AI abstracts more of the buying process, the strategic question shifts from how we buy to what we’re optimizing for, and who controls the logic. The retailers who can answer that on their own terms – with intelligence that’s theirs – are the ones building something that lasts.
The receipt is more than proof of sale. It’s an input into intelligence. And the networks that treat it that way will be the ones worth buying against.
