The Agentic Commerce Measurement Gap – CDO Magazine
Events
Topics
Leadership Summit – Boston – Westin Boston Seaport District – December 10, 2026
Join top AI and Data leaders in Boston to explore the strategies, governance, and innovations driving enterprise performance. Connect, learn, and gain…
Nominate Now | Top 30 Retail Data Changemakers 2026 Nominate for the Top 30 Retail Changemakers 2026, our annual recognition honoring visionary leaders who are redefining the retail landscape through da…
How to Become a Chief Data Officer A decade ago, relatively few organizations had a Chief Data Officer (CDO). Today, the role sits at the center of enterprise data, AI, and business tra…
Leadership Summit – Boston – Westin Boston Seaport District – December 10, 2026 Join top AI and Data leaders in Boston to explore the strategies, governance, and innovations driving enterprise performance. Connect, learn, and gain…
Nominate Now | Top 30 Retail Data Changemakers 2026 Nominate for the Top 30 Retail Changemakers 2026, our annual recognition honoring visionary leaders who are redefining the retail landscape through da…
How to Become a Chief Data Officer A decade ago, relatively few organizations had a Chief Data Officer (CDO). Today, the role sits at the center of enterprise data, AI, and business tra…
Leadership Summit – Boston – Westin Boston Seaport District – December 10, 2026 Join top AI and Data leaders in Boston to explore the strategies, governance, and innovations driving enterprise performance. Connect, learn, and gain actionable insights — reserve your seat today.
Nominate Now | Top 30 Retail Data Changemakers 2026 Nominate for the Top 30 Retail Changemakers 2026, our annual recognition honoring visionary leaders who are redefining the retail landscape through data, AI, and analytics.
How to Become a Chief Data Officer A decade ago, relatively few organizations had a Chief Data Officer (CDO). Today, the role sits at the center of enterprise data, AI, and business transformation. What began as a role focused largely…
Leadership Summit – Boston – Westin Boston Seaport District – December 10, 2026 Join top AI and Data leaders in Boston to explore the strategies, governance, and innovations driving enterprise performance. Connect, learn, and gain actionable insights — reserve your seat today.
Nominate Now | Top 30 Retail Data Changemakers 2026 Nominate for the Top 30 Retail Changemakers 2026, our annual recognition honoring visionary leaders who are redefining the retail landscape through data, AI, and analytics.
How to Become a Chief Data Officer A decade ago, relatively few organizations had a Chief Data Officer (CDO). Today, the role sits at the center of enterprise data, AI, and business transformation. What began as a role focused largely…
Branded Content
Written by:
Nick Albertini | Global Field CTO at Tealium
Updated 9:00 AM EDT, August 28, 2026
The industry has diagnosed the measurement gap in agent-mediated commerce correctly and assigned it to the wrong owner. When a machine does the shopping, winning the sale depends on what your data says when no one is looking at your website.
A few weeks ago, I needed a Kindle for my daughter. I asked an AI assistant to find the right model at the best price. It compared retailers, surfaced a well-publicized sale at Target, and noted that Best Buy was almost $50 higher for the same device. We then went to the nearest Target that afternoon and bought it in the store.
Consider what each retailer observed. Best Buy almost certainly received some trace of the comparison: a server request from the agent, a hit against a product feed, or a read from a search index built by crawling its pages. What it did not receive was anything it could recognize as a customer. Agent requests execute no analytics tags, carry no session, and get filtered as bot noise by the very systems built to keep measurement clean. There was no visit in the sense the dashboard understands, no bounce, no abandoned cart, no price-comparison exit to retarget against. Best Buy was considered and rejected in plain sight of its own infrastructure. Target, meanwhile, saw a walk-in transaction at a register, almost certainly attributed to nothing in particular, even though its promotion did exactly what promotions are supposed to do. The consideration that decided the sale was visible to neither company. A machine read structured data, and a human closed the loop in a physical aisle.
I could have taken a different route. A co-browsing agent that drives a real browser on my behalf would have rendered Best Buy’s pages, fired its tags, and left something closer to a recognizable session. That mode exists, and for retailers, it is the gentler version of what is coming. But it is not the path my assistant took, and the emerging commerce protocols are being built around structured interfaces rather than rendered pages. The harder case is the one that already happened.
On its face, Best Buy simply lost on price, and fifty dollars is fifty dollars. But losing on price used to be observable. A shopper who compared and left generated a visit, an exit, and a chance to respond with a price match or a retargeting offer. Here, Best Buy lost a price competition it never knew it had entered. It cannot see whether this was one Kindle or a pattern across a category, and it cannot respond to a contest its measurement does not know occurred. That is the loss no dashboard records: not the sale itself, but the knowledge that it was ever in play.
The infrastructure behind moments like this is real and moving quickly. In January 2026, Google and Shopify introduced the Universal Commerce Protocol, an open standard that lets AI agents interact with merchant systems across the shopping journey, and Google is bringing UCP-enabled purchasing into AI Mode and Gemini. OpenAI has been building the Agentic Commerce Protocol as the connective layer between merchants and shoppers in ChatGPT. McKinsey puts the economic stakes at three to five trillion dollars globally by 2030, with as much as a trillion in U.S. consumer retail alone.
The industry has also converged on a diagnosis. Discovery now happens outside owned properties. Identity fragments across agents, devices, and stores. Attribution over-credits the last observable click while the real consideration happens inside a model. Context has to be current to be useful, and governance has to extend to a new class of actor. All of that is accurate.
What the consensus gets wrong is the assignment. Because the symptoms surface in marketing’s reports, in traffic that arrives unexplained and conversions that resist attribution, the problem has been filed as a marketing measurement problem, and the proposed fixes are measurement fixes: new referral taxonomies, new attribution models, new dashboards for AI traffic. Marketing is genuinely harmed here. It still has to defend budgets, forecast demand, and prove what worked, and it cannot do any of that while the journey is invisible. But marketing can’t fix this alone, because the gap does not originate in measurement. It originates in the data the agent read, or failed to read, before any measurable event occurred. This is a data problem and a measurement problem at once, and the order matters. Fix the data layer and measurement becomes recoverable. Fix only the measurement and you are instrumenting a store the customer never enters.
An agent does not experience merchandising or the marketing story behind it. It never sees the hero image, the endcap, or the carefully tuned product page. It reads structured data: price, inventory, promotion, eligibility, reviews, delivery windows. Target’s price won my business, and it won because the sale was current, published, and legible to a machine at the moment the machine was deciding. I never saw an ad. Best Buy may well have had a competitive offer somewhere in its systems, but if the agent could not read it, it did not exist.
That inverts a quiet assumption in enterprise architecture: that the human-facing site is the storefront and the data feeds behind it are plumbing. In agent-mediated commerce, the feed is the storefront. Merchandising becomes data publishing, and competitiveness becomes a property of the data layer.
None of this means the agent always completes the purchase. Mine did not. The research was machine-mediated, and the transaction was a person handing a card to a cashier and getting 5% more off even by using the Target RedCard instead of the Best Buy Visa. That hybrid pattern—agent-assisted research followed by a human purchase that often happens offline—is the near-term shape of agentic commerce, and it will coexist with ordinary browsing for years. People will still wander stores, still buy on impulse, and still ignore the assistant’s recommendation because they liked the other one better.
The hybrid journey makes the underlying problem harder, not softer. A journey that spans an AI assistant, a phone, and a physical register crosses more identity boundaries than a browser session ever did. We used to just have to worry about multiple devices. Connecting a walk-in purchase to the machine-read promotion that caused it is exactly the kind of cross-context resolution most enterprises still struggle with when the journey is entirely human. Treating agentic commerce as a future state to plan for misses the point that the messy, blended version is already producing unmeasured wins and invisible losses.
There are three changes occurring, and each one moves responsibility and the center of gravity toward the data organization.
First, data quality becomes revenue exposure. A stale price or a wrong inventory flag used to cost a bounce you could see and diagnose. Now, it costs a sale you will never know was in play. Freshness and accuracy stop being hygiene metrics and become competitive posture, which changes who should be accountable for them and how aggressively.
Second, governance has to run at machine speed for a new class of shopper. Agents read customer data, apply preferences, and increasingly take action. Deciding what an agent can access, how consent is enforced across protocols the marketing stack has never touched, and which actions require human confirmation is the Chief Data Officer’s existing charter extended to a new consumer of data.
Third, attribution cannot be repaired from inside marketing because the consideration happens inside someone else’s model. No tag or pixel will observe the comparison an assistant ran on my behalf. What the enterprise does control is the surface it exposes: which data, at what freshness, under what terms, with what identity and consent attached. That is an architecture decision, and architecture sits with the data organization. Marketing remains the customer for the answer, and the data leader owns the means of producing it.
The response is concrete, and none of it requires waiting for the protocols to settle.
Treat commerce data as a product. Product, pricing, inventory, and promotion feeds need named owners, freshness service levels, and versioning, the same way an API for a paying partner would. An agent reading a price that expired an hour ago is not experiencing a rendering bug. It is walking past your store.
Treat agent traffic as a source rather than noise. Standards are emerging that let agents cryptographically identify themselves, and in some designs carry a delegated identity from the customer they represent. Most enterprises currently filter that traffic out as bots. Classifying it at the point of collection—which agent, on whose behalf, reading what—is the beginning of getting the journey back, and it should happen once, upstream, rather than separately in every downstream tool.
Close the loop to the register. The Target transaction that ended my Kindle search was an offline purchase caused by machine-read data. Joining transaction systems, loyalty identity, and upstream context is old and unglamorous work, and it is now the only way to see whether the data you published is winning.
Put consent in the serving path. If an agent can read customer context, consent has to be enforced where the data is served, in real time. Reviewing machine-speed decisions after the fact is an audit, and an audit arrives too late to prevent the decision it examines.
On the day my daughter got her Kindle, Best Buy’s analytics recorded a normal day. Nothing bounced, nothing was abandoned, nothing churned. Whatever trace the agent left was indistinguishable from the bot noise every retailer discards by design. The journey did not become unmeasurable. It moved to a layer that marketing does not own and mostly cannot see, where the competition is between one company’s data and another’s. Marketing will feel the consequences first and will rightly demand the measurement back. Getting it back starts underneath, with the data an agent reads when it goes shopping on someone’s behalf.
About the author:
Nick Albertini is the Global Field CTO at Tealium, where he champions strategic innovation in customer data and modern marketing technology. With 18+ years of experience across all industry verticals, Nick is a respected voice on data architecture and ecosystem transformation.
His background includes leading extensive teams of architects and data scientists to deliver omnichannel personalization for clients like Uber, Dell, and M.D. Anderson Cancer Center. Albertini is passionate about helping brands create unique, personal customer experiences through robust data integration. He holds an MBA from The University of Texas at Dallas and a B.S. from Texas A&M University.
Elevate Your Personal BrandShape the Data Leadership AgendaBuild a Lasting NetworkExchange Knowledge & ExperienceStay Updated & Future-Ready
source
This is a newsfeed from leading technology publications. No additional editorial review has been performed before posting.
