The problem with data-driven advertising isn’t the advertising. It’s the data.
Spend a little time in a marketing strategy meeting and you’ll hear some version of the same sentence: “We need to be more data-driven”.
Fair enough.
Except a lot of what gets termed “commerce data” is actually closer to a transactional ledger: a log of what was bought, when it was bought, and where. For a lot of companies, that’s where the story ends. In this case, a transaction is the last step of the shopper journey, not a window into it. But knowing somebody bought a laptop tells you almost nothing about what a shopper will buy next, what else they’re considering, or where they are in the buying cycle right now.
What’s missing is everything that happens before the purchase: the discovery of a brand or a product for the first time, the comparisons, the products viewed and abandoned, the category signals that reveal how intent actually forms.
That’s upstream data: signals that capture intent before it collapses into a single outcome. A model trained on transactions alone is descriptive. One trained on the full journey can be predictive.
And that upstream layer is precisely what most audience providers just don’t have.
What “commerce data” really means
To define commerce data, we first have to define its building blocks: commerce signals.
Searches, product page views, comparison behaviors, adding and removing items from the cart, final purchases, and many more — all of these are considered commerce signals.
Taken in isolation, each of these data points might be fleetingly interesting but incomplete. But brought together, organized and connected with AI, they start to form a robust picture of real-time shopper intent.
And there are two sides to that intent story:
- Product understanding. What a product is, how demand for it ebbs and flows over time, and how shoppers move from one product to another, revealing connections that product data alone could never surface.
- Shopper understanding. How people discover and engage with products, how their behavior patterns change over time, and what those signals suggest about what they might do next.
The true value of commerce data comes from connecting these two sides in a structured, coherent way.
From commerce data to commerce intelligence
Data becomes intelligence when it’s organized with AI decisioning to surface patterns that aren’t immediately obvious.
Criteo’s approach connects product knowledge with shopper behavior signals at scale — the discovery, comparison, and consideration behavior that happens before a purchase. Understanding how products relate to each other, how interest flows between categories, and how intent forms before a transaction is what turns raw inputs into something predictive.
That predictive power means commerce data can answer questions that broad audience data simply can’t. Broad audience data asks, “What did this shopper buy last?” and stops there. Commerce intelligence keeps going:
- What might they need that they haven’t searched for yet?
- What are they likely to buy next?
- What did they leave behind in their cart?
These questions become much more valuable when a single shopper’s journey can be compared against thousands of similar ones. And the answers matter regardless of where you sit in the ecosystem.
For publishers, better answers mean audience segments with real commercial weight.
For brands and agencies, they mean less wasted spend, better personalization, and more precise delivery — including whether acting on a signal makes media investment sense once bid efficiency, margin, and business outcomes are factored in.
This is also where the “right tool for the right job” argument comes into focus.
General-purpose AI models are undoubtedly capable with language, reasoning, and broad knowledge tasks. But they haven’t been trained on SKU-level transaction data, product-shopper interaction graphs, or the bidding economics that determine whether a media opportunity is worth taking.
Criteo’s AI models are built around those realities, learning from over 120 intent signals per shopper across billions of interactions each day. That’s a fundamental difference in what the system was designed to do — and it’s what makes a real difference for clients.
What makes Criteo’s commerce data different?
Volume is part of the answer, but it’s far from the most interesting part.
At Criteo, we see the commerce activity of 740 million daily active shoppers across more than 17,000 advertisers, and we observe over $1 trillion in online transactions annually.
That’s the kind of scale that creates statistical reliability, but it’s just one piece of the puzzle. Perhaps the more important question is, what do we really know about each of these interactions?
The answer is, well, a lot.
Criteo is an independent company, which means we work with everyone: brands, retailers, commerce companies, marketplaces, and publishers.
Other players may have rich visibility into their own ecosystem, but that visibility is bounded. It captures what happens on a single platform, with a single retailer, in a single channel. Criteo’s network spans thousands of brands, retailers, and publishers — giving us a much wider field of view. We see how products relate to one another, how interest moves between categories, and how shoppers engage in the places where commerce happens. That kind of pattern visibility is hard to replicate from inside any single ecosystem.
With that in mind, there are four key distinctions that set Criteo’s commerce data apart:
- Direct connections. Our data comes from direct integrations with brands, merchants, retailers, and publishers — not from modelled or inferred signals (read: guesswork). When a shopper interacts with a product in our network, we have visibility into the views, comparisons, add-to-carts, and purchases.
- Scale with structure. Our network covers 4.5 billion specific products (SKUs) mapped to brands and merchants, as well as 120+ real shopper interaction signals, from browsing and basket activity through to completed transactions. Zoom out a bit, and you’re looking at an interconnected map of commerce in motion.
- SKU-level granularity. We tie shopper behavior directly back to specific products, not category guesswork. That’s the kind of precision that makes personalization truly useful, as opposed to generic.
- Normalized product knowledge. Through Criteo’s Universal Catalog, product data is organized into a consistent structure across retailers and categories, with more than 250 attributes per product (things like pricing, inventory, and fulfillment). This level of product knowledge turns a fuzzy, isolated signal into a connected datapoint you can draw real conclusions with.
Privacy is built in, not bolted on
Privacy is far more than a checkbox at Criteo. It’s a guiding principle.
Every service Criteo builds follows a Privacy by Design ethos, which means privacy considerations are embedded into how products are architected, not tacked on after the fact.
And that commitment is structural. We have maintained a dedicated privacy office and a Group Data Protection Officer for years — long before regulators enforced it. The data Criteo works with relies on pseudonymized identifiers rather than personally identifiable information, and data minimization is a core practice. User data is retained for a maximum of 13 months, and Criteo participates in major industry transparency frameworks including IAB Europe’s Transparency and Consent Framework, the DAA, and the EDAA, which has independently certified Criteo’s services for data protection.
The integrity of Criteo’s commerce intelligence starts with the integrity of how the underlying data is gathered.
Frequently Asked Questions
What is commerce data?
Commerce data starts with products — what they are, how they’re described, how they relate to one another, and how demand for them shifts over time. Layered on top of that is the full range of signals that shoppers generate as they interact with those products: searches, views, comparisons, cart activity, and purchases. It’s the combination of both that makes it truly useful.
Why is commerce data valuable?
Because it captures intent while it’s forming, not after a purchase has already been made. A transaction tells you what someone bought. Commerce data — the browsing, comparison, and consideration signals that lead up to it — tells you what someone is thinking about buying right now. That’s a fundamentally more useful moment for a marketer to show up in.
More precise data also means better predictions. The richer and more structured the commerce signals, the smarter Criteo’s commerce AI becomes at identifying patterns, anticipating behavior, and improving the relevance of every ad served.
Better data in, better outcomes out.
What makes Criteo’s commerce data different?
Scale, structure, and depth. Criteo observes 740 million daily active shoppers, maps 4.5B SKUs with more than 250 attributes each, and connects shopper behavior to specific products through direct retailer and brand integrations.
How does Criteo use commerce data?
To understand shopper intent, build high-quality audiences, personalize advertising at the product level, and optimize media investment for business outcome, including timing, suppression, upsell, and cross-sell.
Why does specialized commerce data matter in the age of AI?
General AI models handle language and reasoning well. They don’t natively understand the economics of media, SKU-level product relationships, or how shopper intent actually forms. Specialized commerce data is what makes AI systems genuinely useful for commerce outcomes.
What is commerce intelligence?
Commerce intelligence combines commerce data at scale with AI decisioning to drive better outcomes for businesses and more relevant experiences for shoppers. It analyzes and learns continuously from billions of commerce signals — searches, transactions, catalog updates, inventory changes — to enable smarter targeting, bidding, product recommendations and maximized performance across an increasingly fragmented commerce ecosystem.
