A data clean room is a privacy-preserving compute environment where two parties — typically a brand and an ad platform, or a brand and a retailer — run joint analyses across their datasets without either side exporting row-level customer data. The brand uploads its first-party data (hashed customer IDs, orders, lifetime value); the counterparty contributes its side (impressions, engagement, or purchases); queries run against the joined dataset under aggregation thresholds and re-identification restrictions. Inputs go in, aggregates come out — joined rows never do.
Clean rooms grew into a measurement venue under privacy-era pressures. ATT on iOS, plus browser-level restrictions on cross-site tracking that vary by browser, reduced the shared cross-domain identifiers that previously let exposure and outcome data be joined out in the open. The ones a DTC operator encounters fall into two categories: platform-side, run by major ad platforms (Amazon Marketing Cloud is the most stable named example), and retail-media-side, run by retailers with first-party purchase data. Three uses dominate: incrementality against a platform’s exposure log, cross-platform audience overlap analysis with the customer list as the join key, and retail-media measurement where the retailer holds purchases and the platform holds impressions.
The limits matter. Clean rooms require meaningful first-party data volume to clear aggregation thresholds — small lists return null or suppressed cells. Query latency and those same thresholds make them a measurement venue, not a real-time decisioning surface. And platform-specific clean rooms are not interoperable: the brand’s data lands separately in each, with no joined view across them.