The data you already have is worth more than the model you're going to buy
There is a persistent gap between what executives say about their data and what they actually do with it. Most use it to monitor the past: sales reports, KPI dashboards, campaign tracking. Some already activate it to personalize experiences through artificial intelligence. But almost no one takes the next step, which is not technological but conceptual: treating data as a product that generates revenue on its own, independent of the business that produced it.
That is the central thesis that has been gaining traction in the strategic conversation of 2026, and it has numbers behind it that make it hard to ignore. The operating margins of traditional retail trade move between 2% and 5%. The margins of first-party data advertising networks—built on the same customers and the same infrastructure—can reach up to 90%. This is not a variant of the original business: it is a different business, built on top of the same asset that already existed. Walmart Connect grew 41% in fiscal year 2026. Kroger's alternative business unit, which includes media and data analytics, generated 1.5 billion dollars in operating profit in its last fiscal year. Both companies built those revenue lines without acquiring new customers or opening new markets. They changed the way they packaged what they already knew.
What is missing from that analysis, and what interests me most as a behavioral analyst, is the question that no executive is asking out loud: if the asset was always there, why do most companies not monetize it?
The invisible asset and the bias that keeps it dormant
The answer does not lie in technology or talent. It lies in how organizations perceive what they possess.
There is a well-documented cognitive bias called familiarity with one's own environment: we tend to devalue what we produce or control because repeated exposure reduces the perception of its value. A marketing team that has spent years looking at the same customer behavioral data stops seeing it as a scarce asset. They treat it as an operational input because that is what it has always been. The human brain is extremely efficient at normalizing the familiar and undervaluing what has no visible market price.
This is compounded in large organizations because data is fragmented across departments that do not share incentives. The commercial team knows how much each segment buys. The logistics team knows when and where. The product team knows which features drive retention. But nobody is paying to assemble those pieces into a coherent product that another company would buy, because each department measures its performance by internal objectives, not by the external value of what it knows.
The practical consequence is that a company's most valuable data—its accumulated knowledge about customer behavior—lives scattered, unstructured, unprice, and without a clear owner. Not because the company is incapable of organizing it, but because organizing it in that way requires breaking with the operational logic that sustains day-to-day activity. And breaking with that logic generates institutional friction that most teams have no incentive to take on.
The case of 84.51°, Kroger's analytics and media subsidiary, is instructive precisely because its starting point was not technological. The Stratum platform did not emerge because Kroger discovered new data. It emerged because someone decided to structure what it already knew around the decisions that fast-moving consumer goods brands needed to make: where to spend, what to stock, how to measure results. The asset was the same. What changed was the interpretive framework from which it was offered. That shift—from internal data to external product—is more an act of organizational design than a technical act.
Why AI does not solve the problem and makes it more urgent
There is an understandable temptation at this point in the technology cycle: to believe that implementing language models or generative artificial intelligence tools is sufficient to capitalize on the data a company possesses. It is not, and understanding why requires distinguishing between two types of advantage.
The first type is the advantage of access to tools. Three years ago, having access to large-scale language models was a real advantage because few could afford to develop them. That advantage has now practically disappeared. The most capable models are accessible to any company with a reasonable budget. The base model market tends toward parity, in the same way that access to cloud servers stopped being a differentiator a decade ago.
The second type is the proprietary input advantage. What a company feeds into the model matters more than the model itself. The 62 million households and the 2 billion annual transactions that 84.51° processes are not replicable. A logistics company with five years of route and regional demand data is not replicable either. A health system with clinical records linked to outcomes is not replicable either. The advantage lies not in the algorithm but in what the algorithm processes, and that is exactly what most companies continue to treat as an operational input rather than as a commercializable asset.
The paradox is that the mass adoption of artificial intelligence makes it more urgent—not less—to resolve the problem of data monetization. If everyone has access to the same tools, the differential migrates entirely toward who has the richest, cleanest, and most structured data to produce outputs that others cannot replicate. Companies that have not resolved the architecture of their proprietary data by 2027 will not be at a disadvantage because they lack technology. They will be at a disadvantage because they will have allowed their only real competitive advantage—the accumulated knowledge of their customers—to remain unmonetized while their competitors convert it into margin.
This applies outside of retail with the same logic. A media outlet that knows which content formats drive conversion for which segments can build a planning tool for advertisers. A logistics company that knows when and where demand concentrates can offer benchmarks to its own clients. An insurer that understands risk patterns with geographic granularity can sell that knowledge to governments or real estate developers. The common denominator is not the sector: it is having information that others need in order to make better decisions and that they cannot build themselves in the short term.
What blocks the move is not technical
Up to this point, the analysis seems to point toward an obvious opportunity that requires only executive willpower. The organizational reality is considerably more complicated, and human behavior within institutions explains why most companies do not make the move even when the logic justifies it.
The first obstacle is business identity. Organizations build narratives about what they are. A bank is a bank. An airline is an airline. A supermarket chain sells food. When someone internally proposes converting customer data into a product that is sold to third parties, the instinctive response of many teams is not analytical but identity-based: "that is not what we do." That resistance is not irrational from the perspective of the individual expressing it. It is a signal that the proposed change threatens the mental model with which that person has built their professional career. The bank that decides to monetize behavioral financial data becomes, in part, a data company. And that implies that some internal profiles lose relevance while others that did not previously exist become central.
The second obstacle is governance friction. Customer data is subject to privacy regulations that vary by market and sector. Building a data product that is commercially viable, legally defensible, and trustworthy for third parties requires an architecture of consent, anonymization, and regulatory compliance that most companies do not have ready. Not because it is impossible to build, but because building it requires cross-functional investment in areas that have historically not worked together: technology, legal, product, and commercial. Coordinating that investment without any department counting it as their own victory generates exactly the kind of political inertia that freezes the most promising strategic initiatives.
The third obstacle is the absence of a visible price. Financial markets value business units when they generate revenue with their own structure. As long as a company's data is buried within operations without generating a separate revenue line, its value does not appear in any financial model. That means no analyst pressures it from the outside, no executive compensation incentive directly rewards it, and no board of directors demands it as a priority. The asset remains invisible on the balance sheet because it has no price, and it has no price because no one has made the decision to assign one to it.
Overcoming those three obstacles does not require new technology. It requires a change in the way leaders frame the problem internally: moving from "how do we use our data to operate better?" to "for what decision made by another company are we the most valuable and irreplaceable source of information?" That second question forces you to look outward before looking inward. And that, for most executive teams, is a considerably more difficult psychological move than implementing any analytics platform.
Data monetization is a design problem before it is a technology problem
The lesson that emerges from the Kroger case, and from the patterns that repeat themselves in logistics, healthcare, and media, is not that companies need more data or better models. It is that value is trapped in a design gap between what an organization knows and the way it structures that knowledge so that someone else can pay for it.
That gap has a specific anatomy. On one side, there is information accumulated over years of operation: transactions, behaviors, patterns, anomalies. On the other side, there are decisions that external actors need to make with better information than they currently have: how much budget to allocate to which channel, what inventory to maintain at which point in the chain, which risk profiles deserve different conditions. The gap between both sides is the product. The design work is to build the structure that connects what the company knows with the decision that the external client needs to make, clearly enough and reliably enough that the client will pay to access it on a recurring basis.
What the mass adoption of artificial intelligence does in this context is reduce the cost of building the bridge. Organizing, cleaning, and structuring data that previously required teams of data engineers working for months can now be done in weeks with the right tools. That does not eliminate the design problem or resolve the organizational friction. But it does reduce the barrier to entry enough so that companies that previously could not afford that development now have the technical capacity to do it, provided they have the strategic clarity to decide what to build and for whom.
The decision remains human. And the reason why most companies do not make it, despite having the asset available, remains psychological before it is technical. Business identity, internal political friction, and the accounting invisibility of data assets are forces that no language model can resolve. They are forces that require someone at the top to decide to look at what they already have with different eyes than those they used to build it. That perceptual shift is, at this point in the technology cycle, the scarcest and least copyable competitive advantage that exists.
