Explore audience overlap in a privacy-preserving way
Analysis clean rooms are a type of clean room that enables Optable DCN customers to join their first-party identity graphs with one or more existing DCN partners or flash partners and learn private reports on overlapping user records. These operations are privacy preserving and the underlying audience IDs are never learned by the partners.
Analysis clean rooms are often used as a starting point for collaboration between partners, to explore overlap during planning exercises, as well as to measure overlap of converted and exposed audience panels in a privacy preserving way.
They enable two or more partners to perform a privacy preserving analysis of the overlap of their audiences, each specified in terms of their respective (private) ID graph data (i.e. match keys and user traits).
In order to initiate an analysis clean room, you first need to be partnered with at least one or more DCN partner(s) as described in the partnership section.
An analysis clean room involves two or more DCN partners. There is an initiator, initiating the analysis clean room with one or more DCN partner(s), who are referred to as the participant(s).
You have to select a source audience and optionally some associated traits, and invite one or more DCN partner(s) as participant(s) to execute the analysis with. You also have the ability to limit the number of traits associated with the audience that will be respectively selected by each of the participants. The number of traits chosen by a participant is directly correlated to the amount of privacy protected information you share with that participant and in turn the privacy cost associated with that participant's report generation operation. Once the participant's report is generated, this cost will be deducted from the remaining privacy budget associated with that partner, as described in the privacy budget tracking section.
You have to also configure a duration between 1 to 7 days which only begins counting down once the configuration is completed and at least one participant was invited. Until this specified clean room ends, you may invite additional participants or remove participants who have not yet accepted their invite. You may also optionally end the clean room before it reaches the end of its set duration.
All active clean rooms you are invited to as a participant are accessible in the Clean Rooms > In Progress section of your DCN UI.
For each clean room you are invited to, you can view the number of traits chosen by the initiator of the clean room and the privacy cost associated with the initiator's participant report and summary report generation operations. You can either accept or reject the invitation. If you accept the invitation, once the initiator's participant report and summary report are generated, this privacy cost will be deducted from the remaining privacy budget associated with the initiator as described in the privacy budget tracking section.
Once a clean room invitation is accepted, you can participate by navigating to the Audience section of the specified analysis clean room and complete the following steps:
- 1.Select a source audience
- 2.Select one or more matchable IDs which are related to your previously selected source audience
- 3.You can optionally select some traits which are also related to your previously selected source audience. It is important to note that the initiator of the analysis sets a trait key limit (up to 7) and that the participants will be able to optionally select some traits up to that set limit.
- 4.Click on the Participate button
- 5.Confirm your participation
Shortly after you have confirmed your participation, you will be able to access your participant report.
Analysis clean rooms currently supports two types of differentially private reports - a participant report and a summary report.
The private audience overlap reports, while they reveal information about the overlap of private audience data as a group, have built-in privacy protections to prevent re-identification of users (identity clusters). A combination of thresholding and small amounts of differentially private noise are used to remove the ability to re-identify users from the report result data.
Each participant can see the number of overlapping user records and how they distribute across the configured user trait values, for the overlapping users present in the initiator's selected audience.
Example of a participant's view - Participant Report
Example of an initiator's view - Participant Report
The summary report is exclusive to the initiator of the clean room, and contains an aggregated overlapping report showing overlap with the audiences of all of the participants, deduplicated (i.e. the union of each intersected audience).
It is important to note that the summary report is generated only in the event that at least two or more invited participants participated in the analysis clean room
Example of an initiator's view - Summary Report
Analysis clean rooms integrate with Optable's differential privacy budgeting system, allowing DCN customers to monitor privacy protected clean room activity with its partners, as well as track its rate of collaboration versus a system assigned and replenished privacy budget. With the built in notifications, a DCN customer can better assess and decide its level of risk tolerance for re-identification resulting from clean rooms, on a partner by partner basis.