Inferring a reach of a content item presented to individuals by one or more content publishers for a set of impression frequencies

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

• receive, at an online system, information describing each presentation of a content item to each of a plurality of individuals by a plurality of content publishers, the information comprising user-identifying information associated with a set of the plurality of individuals;
• identify one or more users of the online system from the plurality of individuals based at least in part on a comparison of the user-identifying information to information maintained in the online system describing a plurality of users of the online system;
• determine a known total number of presentations of the content item to the plurality of individuals and a frequency with which each of the one or more users were presented with the content item by the plurality of content publishers based at least in part on the information describing each presentation of the content item;
• de-bias information describing one or more attributes of the one or more users who were presented with the content item for each of a plurality of impression frequencies by the plurality of content publishers based at least in part on a set of attributes for the one or more users and a set of corresponding attributes for the plurality of individuals;
• train a set of machine-learning models to infer a number of the individuals presented with the content item for each of the plurality of impression frequencies, the set of machine-learning models trained based at least in part on the de-biased information;
• infer the number of individuals presented with the content item for each of the plurality of impression frequencies using the set of machine-learning models;
• compute an inferred total number of presentations of the content item based at least in part on the inferred number of individuals presented with the content item for each of the plurality of impression frequencies;
• compare the inferred total number of presentations of the content item to the known total number of presentations of the content item;
• and
• responsive to determining that the inferred total number of presentations of the content item is greater than the known total number of presentations of the content item, adjust the inferred number of individuals presented with the content item for a set of the plurality of impression frequencies.

A method comprising:

• receiving, at an online system, information describing a plurality of presentations of a content item to a plurality of individuals by one or more content publishers, the information comprising user-identifying information associated with a set of the plurality of individuals;
• identifying one or more users of the online system from the plurality of individuals based at least in part on the user-identifying information;
• determining a known total number of presentations of the content item to the plurality of individuals and a frequency with which each of the one or more users were presented with the content item by the one or more content publishers based at least in part on the information describing each presentation of the content item;
• accessing a set of machine-learning models that are trained by:
• de-biasing information describing one or more attributes of the one or more users who were presented with the content item for each of a plurality of impression frequencies by the plurality of content publishers based at least in part on a set of attributes for the one or more users and a set of corresponding attributes for the plurality of individuals, and
• training the set of machine-learning models to infer a number of the individuals presented with the content item for each of a plurality of impression frequencies, the set of machine-learning models trained based at least in part on the de-biased information;
• inferring, using the set of machine-learning models, a number of the individuals presented with the content item for each of a plurality of impression frequencies based at least in part on information describing the frequency with which each of the one or more users were presented with the content item by the one or more content publishers.