Measuring performance of content among groups of similar users of an online system

A method comprising:

• storing, by a social networking server of an online system, an edge store storing connections between users of the online system;
• receiving, at the online system, information describing a content item presented to each user of an exposed group of users of the online system via a client device associated with each user in the exposed group, the information including a time of presentation of the content item to each user in the exposed group;
• obtaining user identifying information from the client device associated with each user of the exposed group of users;
• identifying a non-exposed group of users of the online system outside of the exposed group of users;
• generating a measure of similarity between the exposed group of users and each of the non-exposed group of users based on stored attributes including the stored connections between users in the edge store of the social networking server;
• determining information identifying a control group of users as a subset of the non-exposed group of users having at least a threshold measure of similarity to one or more users of the exposed group of users;
• generating hashes of the user identifying information for the exposed group of users and the information identifying the control group of users to obfuscate identities of the exposed group of users and the control group of users;
• providing the hashes to a third party system, wherein the third party system generates a metric describing a number of occurrences of an event associated with the exposed group of users and a number of occurrences of the event associated with the control group of users based on the hashes, wherein the metric is based at least in part on a difference between the number of occurrences of the event associated with the users of the exposed group of users and the number of occurrences of the event associated with the users of the control group of users.

A computer program product comprising a non-transitory computer readable storage medium having instructions encoded thereon that, when executed by a processor, cause the processor to:

• storing, by a social networking server of an online system, an edge store storing connections between users of the online system;
• receiving, at the online system, information describing a content item presented to each user of an exposed group of users of the online system via a client device associated with each user in the exposed group, the information including a time of presentation of the content item to each user in the exposed group;
• obtaining user identifying information from the client device associated with each user of the exposed group of users;
• identifying a non-exposed group of users of the online system outside of the exposed group of users;
• generating a measure of similarity between the exposed group of users and each of the non-exposed group of users based on stored attributes including the stored connections between users in the edge store of the social networking server;
• determining information identifying a control group of users as a subset of the non-exposed group of users having at least a threshold measure of similarity to one or more users of the exposed group of users;
• generating hashes of the user identifying information for the exposed group of users and the information identifying the control group of users to obfuscate identities of the exposed group of users and the control group of users;
• providing the hashes to a third party system, the third party system:
• accessing information describing a number of occurrences of an event associated with the content item and information identifying each of a plurality of online system users associated with the event,
• determining a number of occurrences of the event associated with the users of the exposed group of users and a number of occurrences of the event associated with the users of the control group of users based at least in part on the hashes and the accessed information, and
• generating a metric describing the number of occurrences of the event associated with the exposed group of users and the number of occurrences of the event associated with the control group of users, wherein the metric is based at least in part on a difference between the number of occurrences of the event associated with the users of the exposed group of users and the number of occurrences of the event associated with the users of the control group of users.