Computing a ranked feature list for content distribution in a first categorization stage and second ranking stage via machine learning

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:

• identify, as seed users, users of the online system with a high value score to a third party system, the value score of each user determined based on a measure of benefit provided by the user to the third party system due to presenting content from the third party system to the user;
• identify features of each of the seed users, the features describing data related to the seed users that are relevant to selecting content for the seed users, and wherein the features include a description of one or more actions performed by the seed users at the online system and profile information of the seed user that is retrieved from a user profile data store of the online system;
• identify, by the online system, a weight for each of the identified features, the weight for each identified feature computed based on how much an identified feature increases a value to the third party system of a user that has that identified feature when that user is presented with content from the third party system, as compared to another user without that identified feature;
• divide the identified features of each of the seed users into a plurality of buckets, each bucket having identified features that are associated with a property, the plurality of buckets having non-overlapping identified features;
• rank each bucket according to the weights of the identified features in each bucket;
• identify an additional user that has a threshold measure of similarity the seed users, the measure of similarity of the additional user determined based at least in part on features of the additional user matching the identified features associated with the seed users;
• transmit a content item received from the third party system to a computing device of the additional user for presentation on a display unit of the computing device of the additional user;
• and
• transmit, by the online system over a network to the third party system, one or more third party-presentable factors that are identified by the identified features of the bucket having the highest rank, the third party-presentable factors presented in a reporting user interface and indicating to the third party system a reason as to why the additional user was presented with the content item.