Infillion's acquisition this week of Catalina -- which built a network using deterministic purchase data -- means it acquired data for 130 million U.S. households, 70 major retail banners, and $600 billion in annual consumer spend, which have all turned into usable data.

That data changes the stakes for performance measurement, optimization and closed-loop attribution in an artificial intelligence (AI)-operated ecosystem, Brian Kaminsky, COO of Infillion, told MediaPost.

Infillion offers services across multiple channels like connected TV (CTV), video, and location-based media.

"While we’re not disclosing full commercial terms, what’s important is that Catalina’s expertise and technology will be fully integrated into Infillion’s platform, enhancing deterministic purchase signals across planning, targeting, and measurement," Kaminsky said.

Deterministic data is when an identity of a person is verified rather than modeled or inferred. Catalina's purchase-based identities are anchored in actual transaction records linked to loyalty accounts. 

First-party data was always core to effective marketing, since its the signal is directly connected to real customer intent and behavior.

But as AI moves from support to execution in media planning and buying, the quality and amount of the underlying data now directly determines how well AI recommends, predicts, and optimizes outcomes, Kaminsky said.

"AI models are powerful pattern matchers, but without high-fidelity inputs, they simply amplify mediocre non competitive signals," he said. "In this new era of agent-native, autonomous execution, the AI layer is deciding budgets, bids, and delivery."

Catalina has one of the world’s largest deterministic datasets -- with 11 billion annual shopping trips across 130 million households -- with a high-fidelity, deterministic signal that ties media directly to verified consumer purchases.

In the next year, Infillion's priority is to grow its AI-based agents and turn them into tools that create measurable business impacts.

By integrating deterministic datasets like Catalina’s directly into autonomous workflows, AI optimizes against real purchase outcomes, and expands retail media enablement with true closed-loop measurement, Kaminsky said.

"During the next five years, we expect AI-operated buying to become the norm, with autonomous agents managing the majority of programmatic spend," he said. "Our goal is to be the open, interoperable execution layer powering commerce-centric media, optimized against real economic outcomes, not impressions or clicks."

Retail media depends on linking consumer purchase behavior back to the ad when it serves up. Catalina’s dataset allows Infillion to reach audiences based on real purchase behavior, to measure true sales incrementality at the SKU level, in near real-time and to optimize AI agents toward actual commercial outcomes -- not just proxy key performance indicators (KPIs).

Kaminsky believes this method changes retail media from “reach and frequency” to closed-loop commerce media, where planning, activation, and measurement all live in the same equation.