The Line Drawn Before the Investigation Begins
You are a lawyer in a competition case, and before a single document is subpoenaed, before anyone has been deposed or a filing made, you pick up a pen and draw a box. Inside goes the market under examination. Outside goes everything else. That box, drawn in a conference room on a Tuesday morning, will determine whether the company you are investigating looks like a monopolist or merely a vigorous competitor. It is the most consequential methodological choice in all of competition law, and it happens before the investigation has properly begun.
Get it wrong and a genuine monopoly escapes untouched. Get it too narrow and a healthy competitor gets strangled by a regulator who mistook pricing power for predation. The box decides everything. And the people drawing it are not neutral.
Competition authorities are structurally unable to see monopolies that span the boundaries of how markets are legally defined. If a company controls something that doesn't fit inside any single recognised product or geographic market, the standard tools of antitrust analysis will simply fail to register it as dominant. The problem isn't corruption or incompetence. It's geometry.
The Test That Sounds Simple and Isn't
The workhorse method for defining a market is called the SSNIP test: Small but Significant Non-transitory Increase in Price. The logic is elegant. Ask whether a hypothetical monopolist could profitably raise prices by five to ten percent for a sustained period without losing so many customers to substitutes that the increase becomes unprofitable. If customers would flee to alternatives, those alternatives belong in the same market. If they'd stay and pay, you've found your boundary.
Consider a worked scenario. A company called Meridian sells industrial-grade document scanning software to hospitals. A regulator asks: if Meridian raised its prices by eight percent, would hospitals switch to general-purpose scanning software built for law firms? Probably not. The workflow integrations differ, compliance requirements differ, support contracts are structured differently. So the market gets drawn narrowly: hospital document management software. Meridian, with sixty percent of that segment, looks dominant.
Now ask the question differently. What if Meridian also owns the leading patient records platform that the scanning software integrates with, and the dominant appointment scheduling tool? Each product, tested alone through a SSNIP, might look competitive enough. But the bundle, the system of interlocking dependencies, confers a kind of dominance that never appears in any single product market. No individual box shows a problem. The problem lives between the boxes.
This is not a hypothetical failure mode. It is what happened, in slowly recognised form, with the major technology platforms whose products span search, advertising, cloud infrastructure, app distribution, and payments. Each layer, examined alone, had some competitor somewhere. The stack, examined whole, had none.
The Cellophane Fallacy, Which Keeps Getting Repeated
There is a specific error so well-documented it has its own name, and competition authorities still walk into it with some regularity. That should embarrass the field more than it does.
In the mid-twentieth century, the Justice Department brought a case against DuPont, arguing the company held a monopoly on cellophane. DuPont's defence was that cellophane competed with all flexible packaging materials: waxed paper, foil, polyethylene. The Supreme Court accepted this framing and found no monopoly.
The problem, identified almost immediately by economists, was that DuPont had already raised prices to the monopoly level. At that elevated price, yes, customers would switch to substitutes. But that price sensitivity didn't prove competition; it proved the monopolist had already extracted as much as it could. Running a SSNIP from an already-elevated baseline will always make a market look broader than it is. The test was applied correctly to the wrong starting point.
The Cellophane Fallacy isn't ancient history. Any time a dominant firm has already used its market power to price aggressively, or to bundle services at near-zero marginal cost, the SSNIP methodology can systematically misclassify the market. Free products are the sharpest version of this: if a service is priced at zero, a five-percent price increase is still zero, and the test tells you nothing useful at all. The framework, in such cases, produces a wrong answer with complete methodological rigour.
What People Get Wrong About Market Share Numbers
The instinct, when regulators announce a finding, is to focus on the percentage. Seventy percent market share sounds alarming. Thirty percent sounds fine. But market share is only as meaningful as the market it's measured against, and this is where sophisticated manipulation tends to happen.
Think of two buyers who purchased the same make of car in the same year: one in a city with dense public transit and ride-sharing options, one in a rural county where driving is the only realistic option. The car manufacturer's market share looks identical to an analyst. The competitive constraint on that manufacturer is completely different. The rural buyer is far more captive, like a ship's passenger who can complain about the food but cannot swim to another vessel. A market definition that lumps both buyers together will understate market power for one and overstate it for the other.
This scales up to industrial markets in ways that matter. A company might hold forty percent of the global market for a chemical compound but ninety-five percent of the supply accessible to manufacturers in a particular region, because of shipping costs, regulatory certification requirements, and established supplier relationships. The global figure is technically accurate and practically useless.
Ask yourself: if you found a company with forty percent of a so-called global market, where do its customers actually source alternatives? That is the real market. The global figure is often a legal convenience, nothing more.
The Structural Blindspot: Power That Lives in Relationships
This is where the analysis gets genuinely hard, and where the limitations of the current legal framework are most exposed.
Traditional market definition assumes that market power resides in a product or service: a widget, a platform, a distribution network. The SSNIP test asks whether customers can substitute away from the widget. But an increasing share of contemporary economic power doesn't reside in any single product. It resides in data relationships, in switching costs, in the accumulated behavioural history of millions of users that a new entrant simply cannot replicate regardless of how good their product is.
Call this positional power. It doesn't show up cleanly in any product market because it isn't attached to a product; it's attached to a position in a network of relationships. A payments company that processes transactions for millions of merchants and holds years of fraud-pattern data on each of them possesses something a new entrant with a technically superior product cannot easily contest. Not because the incumbent's product is better in any feature-by-feature comparison, but because the position is self-reinforcing in ways that don't reduce to product substitutability.
The SSNIP test was designed in an era when market power was primarily exercised through price. Raise the price, watch customers flee or stay, count the market. When market power is exercised through lock-in, through data accumulation, through the cost of migrating an entire business's operational history to a new platform, price is almost beside the point. The test is asking the wrong question, and the wrong question produces a wrong answer with complete methodological rigour.
Some jurisdictions have begun experimenting with alternatives: defining markets by the competitive constraint that matters most rather than by product substitutability alone, or shifting the burden of proof so that large platforms must demonstrate the absence of harm. Germany's amendments to its competition act introduced a concept of 'paramount significance across markets' that tries to capture exactly this kind of multi-market positional dominance. Whether that framing proves durable in practice remains genuinely uncertain, and the historical record of ambitious regulatory concepts surviving contact with well-funded legal teams is not encouraging.
The Box Is a Choice, Not a Fact
The deepest issue is one that economists acknowledge in academic papers and that rarely survives intact into courtrooms. Market definition is not a discovery. It is a construction. There is no objectively correct market waiting to be found. Different methodological choices, all defensible, produce different markets and therefore different verdicts about dominance.
This is why well-funded defendants invest so heavily in market definition arguments before any other aspect of a case. Persuade a court that your company competes in the broad market for 'digital advertising' rather than the narrow market for 'search advertising', and your forty-percent share becomes a fifteen-percent share and the case effectively collapses. The number hasn't changed. The box has.
Competition law will always require some boundary-drawing. You cannot assess dominance without a reference point, and some reference point will always be contestable. The honest acknowledgment, which the field has been slow to make, is that the choice of method encodes assumptions about how markets work, and those assumptions were mostly formed in an economy that looks quite different from the one that currently exists.
The monopolies that current tools cannot see are not hiding. They operate in plain sight, in the gap between the boxes. The question was never whether regulators are looking hard enough. It is whether the frame they are looking through was built to see what is actually there, and the answer, for now, is plainly no.