Key Takeaways
- When AI agents optimize purchases by cross-comparing price, inventory, and reviews, "noise" enters transaction data and closed-loop measurement breaks down
- Retail media networks (RMNs) face pressure to pivot from selling ad placements to offering "decision APIs" for AI agents
- Ecommerce operators must simultaneously adopt causal inference-based measurement and build structured data infrastructure that agents can read
The Closed-Loop Premise Collapses
Agent-driven transactions that optimize purchases based on price and other factors will disrupt the closed-loop foundation that underpins retail media networks.
www.beet.tvAt the Beet Retreat in San Juan, Puerto Rico, Sallie's Marco Steinsieck sounded the alarm for the retail media industry. His warning: as AI agents autonomously compare prices, reviews, and inventory across retailers on behalf of consumers, the closed-loop measurement that underpins retail media will be fundamentally disrupted.
Why does data become "noisy"? Traditional closed-loop measurement relies on the ability to track the entire journey from ad impression to purchase. A consumer sees an ad, visits a product page, adds to cart, and buys. This end-to-end behavioral data is what enables the definitive claim that "this ad drove this purchase." But when AI agents mediate the buying process, most of the purchase journey happens inside the agent. Discovery, comparison, evaluation, and purchase decisions all occur as internal AI processing, generating no clickstream whatsoever.
The Shock to a $69 Billion Market
The severity of this problem becomes even starker when you consider retail media's scale. According to eMarketer, US retail media ad spend is projected to reach $69.33 billion in 2026, representing 17.9% year-over-year growth. The reliability of closed-loop measurement has been the very engine powering this rapid expansion.
Yet as TCS highlights in its white paper "Agentic Commerce: The Next Great Disruption in Retail Media Networks," the user journey that traditional attribution assumes will simply cease to exist in the agentic era. Instead of a trackable sequence of views, clicks, and conversions, there will only be autonomous decisions driven by multi-factor optimization. The "viewer" of the ad is an AI, not a human. The "decision-maker" behind the purchase is also an AI. How much meaning can conventional ad effectiveness measurement retain in this scenario?
Steinsieck's warning carries particular weight given his background. A commerce media veteran who launched the Sephora Media Network and helped scale the Staples Media Network, he is now acknowledging the limitations of the very model he helped build — a fact that underscores the urgency of the issue.
How Agents Change the "Quality" of Data
Digging deeper into Steinsieck's warning reveals that the problem extends beyond data simply becoming "unavailable." The meaning of the data that can be collected is fundamentally changing.
Consider a consumer who purchases the same brand of laundry detergent every week. RMNs have used this transaction data to build loyal user segments and target competitive brand advertising accordingly. But once an AI agent takes over purchasing for that consumer, the agent may switch brands based on multi-variable optimization — price, ingredients, delivery speed — rather than brand loyalty. The transaction history still records a purchase, but there is no way to distinguish whether it resulted from a human decision or an agent's optimization.
Kantar's analysis captures this shift succinctly. In a world where agents handle shopping, the RMN's "audience" is no longer a human scrolling past banner ads — it is an algorithm scanning structured metadata, prices, and loyalty perks. Ad creative and brand storytelling do not resonate with machines. What is needed instead is structured data that feeds directly into agents' decision-making logic.
RMN Evolution — From "Ad Slots" to "Decision APIs"
So how should retail media adapt? TCS recommends that RMNs evolve from "platforms that sell ad space" into "retail intelligence networks that orchestrate AI-ready data, commerce logic, and decision APIs."
The required measurement methodologies are also changing. Replacing traditional click-based measurement, causal lift measurement — verifying whether a brand's bid or promotion meaningfully changed agent selection behavior — will take center stage. Additionally, action-level attribution will be needed to track what percentage of substitutions and reorders were agent-driven.
From a different angle, Kantar introduces the concept of "permissioned data sharing." Retailers will selectively determine which AI agents are "friendly" and provide them with curated, enriched product data. This mechanism becomes the new channel through which influence flows in retail media.
What Ecommerce Operators Should Act On Now
With full-scale agentic commerce approaching, ecommerce operators' response can be organized along two axes.
The first is modernizing measurement infrastructure. Moving away from last-click attribution dependency toward incrementality testing and Marketing Mix Modeling (MMM) has been advocated for some time, but the rise of agentic commerce has escalated the urgency. In a world without clickstreams, causal inference-based measurement becomes the only viable option.
The second is structuring product data. AI agents make decisions by reading structured metadata — price, ingredients, specifications, review scores, and shipping terms. As Kantar puts it, "data quality is the brand's new packaging." To be chosen by agents, building machine-readable data infrastructure matters more than human-facing creative — and this will increasingly determine competitive advantage.
Summary
Steinsieck's warning signals the end of the linear "ad → click → purchase" model that retail media has relied upon. With projections that the agentic commerce market could reach $3-5 trillion by 2030, this transformation is not a distant prospect.
The key development to watch is when major RMNs begin implementing agent-compatible APIs and data feeds. Once leading retailers like Walmart, Amazon, and Target start building agent-facing infrastructure, advertisers will be forced into a fundamental overhaul of their measurement and operations. Before the closed loop's "closed ring" breaks open, preparing for the transition to the next measurement paradigm is the most pressing challenge facing ecommerce operators today.
