PAID CONTENT brought to you by
BACK TO HOME
Why AI-driven commerce is only as smart as the data behind it
This retailer is deploying AI agents across 100 million weekly sessions but was blind to the two shopper segments driving $25 million in weekly revenue. Conviva changed that.
Every week, one of the world's largest online retailers handles nearly 100 million shopping sessions. Its engineers monitor uptime across its website, app, and consumer-facing AI agents. Its data scientists track conversion. Its CEO has told shareholders that AI will be the most transformative technology the company has ever deployed — bigger, he said, than building its own logistics network.
And every week, two distinct groups of shoppers account for $25 million in revenue on that platform — roughly one tenth of weekly sales. The company had no idea either group existed.
Not because the shoppers were hidden. Because the tools the company used to understand them were incapable of seeing them.
The Invisible Problem
Conviva — the intelligence platform known for full-census behavioral analytics — is fixing a problem that has quietly worsened as AI moves to the forefront of how consumers shop.
At the platform's core intelligence layer is something Conviva calls the Consumer Context Graph: a real-time map of behavioral patterns built from what consumers actually do across websites, apps, and AI agent conversations. It's built on infrastructure that processes 500 billion events per day at 100 times the efficiency of conventional computation, according to the company.
The $25 million figure comes from exactly the kind of analysis the platform is designed to surface. Working with one of the world's biggest e-commerce brands, Conviva analyzed 93 million weekly sessions at full census and found something the retailer's existing analytics stack had obscured entirely: its shoppers fall into more than 100 distinct behavioral archetypes, each with different conversion signals, different journey patterns, and different experience needs. The top two segments alone account for $25 million in weekly revenue. That’s 10% of the retailer's total from just two segments the company had never identified or optimized for.
The irony is hard to miss. At the same time their CEO was telling shareholders AI would define the company's next era, an outside analysis was revealing that the behavioral foundation those AI systems depend on was effectively invisible.
"Product analytics and observability tools simply cannot tell how or why a customer bought, booked, or came back," said Keith Zubchevich, Conviva's President and CEO. The platform, he said, is designed to surface "the behavioral patterns that separate agents that convert from agents that guess and frustrate."
The Journey Problem
The reason these archetypes were invisible comes down to how the industry has measured shopping for the past two decades.
Session-based analytics — the standard model for virtually every web measurement tool — treats each visit as a discrete event. A consumer who researches a sofa over six sessions across two weeks looks, in that model, like six separate visitors with a 0% conversion rate. The sale, when it finally comes, gets credited to the last session. Everything that preceded it — the research, the comparison, the abandoned cart is invisible.
That's not a niche edge case. Conviva's research found that 67% of online shoppers don't follow linear journeys. The multi-session, high-consideration shopper isn't the exception. They’re the norm. And they’ve been statistically invisible to every analytics tool built on session logic.
This is why archetypes like the ones Conviva identified can drive $25 million a week without ever appearing in a conversion report. Their behavior spans sessions in ways that session-based tools are architecturally incapable of tracking.
What This Means for AI Agents
The stakes are higher now because AI agents are being deployed into this blindspot at scale.
When a shopper asks a chatbot for sofa recommendations, the agent has no idea they spent three weeks researching, saved items to a wishlist, and abandoned a checkout twice due to payment friction. It responds like they’re a stranger — because to every data system feeding it context, they are. The behavioral history exists. The intelligence to act on it doesn't. Conviva sees an average of two and a half minutes wasted in every agent conversation due to restating context or history. The Consumer Context Graph is designed to close that gap — tracking patterns across sessions and making that context available to AI agents in real time, at the moment of decision. Conviva is betting that full-census, stateful pattern analytics becomes the defining infrastructure requirement for consumer brands as AI moves from pilot to production — and that the brands who don't solve the behavioral blindspot will be optimizing agents against a fiction of how their customers actually shop.
For the retailer at the center of this story, the most immediate question is: if you now know who your top two archetypes are, what they need, and what they're worth, how will you teach your AI agents to interact with them?
BACK TO HOME
Note: This content was created by Conviva, a client of Business Reporter.
