There is an old rule in my business: whoever controls measurement controls the money. Creative wins awards, media wins scale, but measurement decides what "working" means, and the definition of working is where budgets actually get set. So when the industry's answer to privacy became the data clean room, I paid less attention to the cryptography and more attention to the org chart. Who owns the room? Who writes the queries you are allowed to run? Who decides what you never get to see?
The answers should worry every marketer. Clean rooms solve a real problem: they let two parties compute over joined data without exposing raw records to each other. That is genuine progress, and I want to be fair to the engineering, which is often excellent. But the dominant clean rooms are not neutral venues. They are operated by the same platforms whose media performance is being measured inside them. The privacy architecture is real. So is the power architecture. We are rebuilding the walled garden problem one level up, in the measurement layer, and this time the walls are made of math that makes auditing sound like a privacy violation.
My thesis: between 2026 and 2030, measurement power consolidates into a small number of platform controlled clean rooms unless buyers collectively demand a neutral layer. The privacy rationale is legitimate. The consolidation riding on it is not inevitable. But it becomes inevitable if the buy side keeps mistaking access for independence.
Key Findings
- Clean rooms genuinely improve privacy, and that is precisely what makes them politically unassailable vehicles for consolidating measurement control.
- The dominant pattern is what I call Measurement Feudalism: advertisers farm their own data inside a platform's walls, under the platform's rules, and receive back only the conclusions the platform permits.
- Query restrictions, aggregation thresholds, and permitted metric lists inside clean rooms are editorial decisions dressed as privacy engineering. Someone chooses them, and that someone is rarely the buyer.
- Cross platform truth is the first casualty. Each room measures its own garden generously and the outside world grudgingly, and no room lets you compare rooms.
- The fix is not less privacy. It is a Neutral Counting Layer: independently governed measurement infrastructure where privacy math is applied by parties with no stake in the answer.
- Use platform clean rooms without trusting them. Take the operational value, but label every output internally as "self reported by the seller" and weight it accordingly in budget decisions.
- Keep an independent spine of truth: media mix modeling plus disciplined incrementality experiments you design and control. Geo holdouts and audience holdouts are annoying, unglamorous, and sovereign. Sovereignty is the point.
- Interrogate the room, in writing. Demand documentation of permitted queries, aggregation thresholds, attribution logic, and change logs. What a vendor refuses to document, refuse to treat as evidence.
- Never let a platform room become your system of record. Mirror every aggregate you are allowed to extract into your own warehouse the day you receive it, so your institutional memory belongs to you.
- Put neutral measurement language into contracts and RFPs now. Individually it is symbolic. Collectively it is how every buy side standard in history actually happened.
How We Got Here, Honestly
The clean room did not arrive as a power grab. It arrived as a compromise. Regulators and platforms shut down the free flow of user level data, and advertisers still needed to answer basic questions: did the exposed group buy more than the unexposed group, how often, at what cost. Clean rooms threaded the needle. Bring your first party data, join it against platform data inside a sealed environment, run approved computations, take home aggregates. Nobody's raw records leak. Real questions get answered.
I have used these rooms across client work and the capability is not fake. Match based measurement inside a clean room beats the modeled guesswork that replaced the cookie in most other places. If you buy significant media inside any major platform, staying out of its clean room is self harm, the same way ignoring first party data strategy is self harm. My argument is not abstinence. My argument is about who governs the instrument.
Measurement Feudalism
Here is the structure to name. In feudalism, the peasant works land he does not own, under rules he does not set, and hands the lord a share of the harvest for the privilege. In Measurement Feudalism, the advertiser contributes proprietary customer data into an environment owned by the media seller, computes under constraints the seller defines, and receives back a verdict on the seller's own performance. The advertiser bears the data risk. The platform keeps the epistemic power.
Every load bearing detail sits behind the wall. Which joins are permitted. Which minimum audience thresholds apply, and thus which analyses silently return nothing. Which attribution logic is implemented in the approved queries. Which columns exist at all. Each individual restriction has a plausible privacy rationale. The sum of the restrictions is a measurement regime in which the party being graded configures the grading. No one would accept this from a public company auditing itself. We accept it daily from media platforms because the alternative is darkness.
And the feudal structure compounds. The more of your historical measurement lives inside one room, the higher the cost of leaving it, and the more your internal benchmarks become artifacts of one vendor's permitted math. Switching costs in measurement are stickier than switching costs in media, because they rewrite your memory of what worked.
The Death of Cross Platform Truth
The deepest cost is comparative. Marketing's most valuable question has never been "did platform X work?" It is "did platform X work better than platform Y, per dollar, for this objective?" That question requires a vantage point above the gardens, and clean rooms structurally forbid it. Room A will not let its data visit room B. Each platform's room implements its own attribution philosophy, tuned, unsurprisingly, to flatter its own inventory. Duplication across platforms, the single biggest source of waste in a fragmented media plan, becomes formally unknowable.
So the buyer is left stitching together verdicts from courts that each ruled on themselves. I watched programmatic advertising evolve the same pathology in the supply chain, layers of intermediaries each grading their own contribution, and I broke down that machinery in my programmatic explainer. The clean room era repeats it at the measurement layer, with better encryption and worse recourse.
Media mix modeling is enjoying a renaissance partly for this reason, and I endorse the renaissance. But MMM alone concedes the granular battlefield. It tells you the tide came in. It does not tell you which boats lifted.
The Neutral Counting Layer
Now the constructive part, because complaining about walled gardens is the oldest genre in adtech. The privacy technology inside clean rooms, secure joins, differential privacy, aggregation floors, is separable from platform ownership. Nothing in the math requires that the referee wear a team jersey. What the market needs is a Neutral Counting Layer: measurement infrastructure operated by independently governed entities, applying published privacy standards, where platforms and advertisers both submit data under rules neither side unilaterally controls.
The components already exist in scattered form: independent clean room vendors, industry joint committees, regulator blessed aggregation standards, academic privacy frameworks. What is missing is buyer leverage forcing them into one place. Platforms will not volunteer into neutral counting, and I do not blame them. Nobody volunteers to be audited. Neutral counting arrives the way viewability standards, ads.txt, and brand safety floors arrived: because enough spend made it a condition of doing business.
The window matters. Standards harden around whatever infrastructure exists when the concrete sets. If, by 2030, five platform rooms hold five incompatible ledgers of record, the neutral layer becomes a decade harder. My falsifiable prediction: by 2028 we see the first major joint industry pilot of independently governed cross platform measurement with real platform participation. If buyers stay passive and that pilot never materializes, feudalism is the steady state, and this paper becomes a eulogy.
What I Would Do About It
Clean rooms are not the villain of this story. They are the terrain. The villain is an old one, the quiet transfer of judgment from the people spending the money to the people collecting it. Privacy gave that transfer a new disguise. Take the disguise off, keep the privacy, and fight for the referee.