Clean room (data clean room)

A data clean room is a controlled environment in which two or more organizations contribute datasets that stay under their own control, a matching step joins records without revealing identifiers, and only aggregated or policy-approved outputs leave.

The architecture has four parts: where the data sits (a warehouse, a vendor's environment, a walled garden or an enclave), who runs the compute, which output controls apply (allowed columns, minimum group sizes, noise or differential privacy, query approval), and which protocol connects the room to partners. IAB Tech Lab's PAIR protocol (v1.0 January 2025) handles identity matching with commutative encryption, and ADMaP (v1.0 February 2025) handles attribution matching (IAB Tech Lab, PAIR; IAB Tech Lab, ADMaP).

Managed rooms publish their controls. AWS Clean Rooms, for example, lets a table owner set minimum aggregation thresholds and comparison controls inside custom analysis rules (added 13 August 2026) and offers managed differential privacy (GA 10 April 2024) (AWS doc history). Google's Ads Data Hub filters rows under roughly 20, 50 or 10 users depending on the privacy mode (Google, ADH privacy checks).

Example: a retailer and a CPG brand each load hashed loyalty emails and hashed CRM emails into a room; the room reports that 1.2 million people overlap and that exposed members bought 8% more than a matched holdout; neither party downloads the other's list. IAB Europe's measurement standards list clean rooms as one of three delivery tools and warn about cost, interoperability and leakage (IAB Europe, V2.1).

Related: data residency, closed-loop measurement, UID2, commerce media network.

Source: IAB Tech Lab, PAIR and ADMaP specifications; IAB, "Data Clean Rooms in Commerce: Common Use Cases" (29 September 2026).

1. IAB Tech Lab, PAIR
2. IAB Tech Lab, ADMaP
3. AWS doc history
4. Google, ADH privacy checks
5. IAB Europe, V2.1