The Precedent Factory

A Factory, Not a Verdict

There is a temptation, when a major regulatory proceeding concludes, to treat it as a verdict — a judgment about what one company did, followed by a remedy designed to correct it. That framing is legally accurate. It is also the first-level observation, and first-level observations about competition regulation tend to miss what is structurally most significant.

Case 100209 is being read, by most observers, as a Google action. The Commission investigated Google's search data practices, found them inconsistent with DMA obligations, and proposed measures to correct the imbalance. That reading is not wrong. The second-level observation is different: for the first time, a major competition regulator has formally determined that a behavioral data asset — not a physical network, not a statutory monopoly, not a patent portfolio — constitutes regulated infrastructure to which third parties have a legal access entitlement. That determination creates a category. And categories in regulation are load-bearing structures. They do not stay where they are placed.

The 1930s classification of electricity transmission as a natural monopoly infrastructure did not stay in electricity — it migrated to natural gas, to telephone networks, to cable. The 1996 Telecommunications Act's network interconnection obligations did not stay in wireline telephony — they shaped the regulatory logic for broadband, for spectrum, for mobile roaming. Case 100209 is the first instantiation of the same structural move in behavioral data markets. The criteria the Commission applied — explicitly or implicitly — to determine that Google Search's data crossed the threshold from competitive advantage to regulated input will function as precedent in every market where behavioral data compounding produces structural foreclosure, whether the Commission formalizes those criteria or not.

Every behavioral data asset held by every DMA-designated gatekeeper is now being evaluated against those criteria — by competition economists, by platform challengers, by national competition authorities watching Brussels. The evaluations span markets that differ sharply in their susceptibility to the mandate's instrument, in the dimension of competitive distance their moats primarily protect, and in the degree to which the DMA's Access-oriented design is structurally suited to address them. None of these proceedings have been formally initiated. All of them are, structurally, underway.

The question is not what the factory has produced. The question is what it makes next — and whether the instrument it keeps producing is suited to the problems it will be asked to solve.

The Template Case 100209 Just Gave Every Challenger

Regulatory categories migrate through criteria. If the criteria that defined Google Search's data as regulated infrastructure are specific to search — calibrated to query volume, click-behavioral relevance loops, Google's singular market position — then the precedent travels a short distance. If they are structural — applicable to any market where behavioral data compounding produces the same competitive dynamics — then the precedent travels as far as those dynamics exist.

What The Moat That Compounds established as a structural mechanism, regulation must convert into an operational threshold. Reading Case 100209 and the DMA gatekeeper designation logic that preceded it, four criteria emerge. They are not all explicitly stated; some are operationally implied by the threshold that was crossed. Together they constitute the framework that any subsequent proceeding will be measured against — and that any challenger firm in any data-advantage market can now use to argue, to the Commission, to a national competition authority, or in a judicial proceeding, that its market meets the threshold. Case 100209 has given them the template.

The first is sustained market concentration sufficient to produce structural foreclosure. Google holds approximately 90% of global search queries. The Commission did not mandate data sharing because Google's data was large — large datasets are common. It mandated sharing because the data advantage operated in a market so concentrated that no challenger, regardless of investment or capability, could accumulate a behavioral corpus sufficient to compete on the dimension that determines product quality. Concentration is necessary but not sufficient: a behavioral data moat in a fragmented market produces competitive advantage without structural foreclosure, and foreclosure is what triggers the regulated-infrastructure threshold.

The second is self-reinforcing compounding. The behavioral data advantage must be dynamic, not static — it must grow as it is used. This distinguishes a behavioral feedback loop from a large dataset. A company that holds ten years of sales records has accumulated data; a company whose product improves with every transaction, and whose improvement generates more transactions, which generates more improvement, holds a compounding moat. The Commission's data scope specification — behavioral signals (click, query, view) rather than static index data — is consistent with having identified the compounding input as the relevant asset, whether or not the Commission articulated the distinction in those terms.

The third is the absence of a functional substitute. No alternative dataset produces equivalent relevance signal for search query intent. Challengers cannot substitute crawling data, social engagement data, or purchased commercial datasets for behavioral query-click signals — these are categorically different inputs that do not train the same model quality. This is what distinguishes a behavioral data moat from a competitive advantage that market entry can eventually correct.

The fourth is demonstrated challenger exclusion — not theoretical harm, not modeled foreclosure, but observable market evidence that technically capable, well-resourced challengers have been confined to marginal positions over a sustained period. The Bing case is the Commission's implicit empirical anchor: seventeen years, substantial capital, world-class engineering, and structural confinement to under 5% market share. The exclusion criterion can also be satisfied by cross-sectional evidence — comparable entrants performing differently based on their behavioral corpus position — or by structural modeling demonstrating that no entry path exists without incumbent behavioral data. The Bing natural experiment is the cleanest form of evidence. It is not the only form.

These four criteria constitute a susceptibility test.