Predicting reach of content using an unresolved graph

1. A method comprising:

identifying, by an online system, a set of unresolved identifiers, wherein an unresolved identifier defines a context in which a client device accesses one or more of a plurality of online systems, and wherein the context has not been determined to be associated with a specific user;

identifying, by the online system, for each unresolved identifier of the set of unresolved identifiers, information associated with the unresolved identifier, the information describing characteristics of one or more access events associated with the unresolved identifier;

for each pair of unresolved identifiers:

determining, by the online system, based on the information associated with each unresolved identifier, a similarity score for the pair of unresolved identifiers;

determining, by the online system, that the determined similarity score exceeds a threshold similarity score;

responsive to determining that the similarity score exceeds the threshold similarity score:

clustering, by the online system, the pair of unresolved identifiers into a cluster, the clustering indicating a prediction that the pair of unresolved identifiers in the cluster are associated with a common user; and

presenting a content item to one or more client devices associated with one or more unresolved identifiers of the set of unresolved identifiers; and

determining, by the online system, a reach of the content item based on clusters including the one or more unresolved identifiers, wherein the reach is the quantity of users who have been presented with the content item.

2. The method of claim 1 , wherein each unresolved identifier from the set of unresolved identifiers is selected from a group comprising a browser ID, a device ID, an HTML request, and an IP address.

3. The method of claim 1 , wherein the characteristics of one or more access events associated with the unresolved identifier comprise at least one of a context of the one or more access events, content accessed during the one or more access events, actions performed during the one or more access events, and derived data learned from the one or more access events.

4. The method of claim 1 , wherein the similarity score for the pair of unresolved identifiers is based on a similarity of the information associated with each unresolved identifier of the pair of unresolved identifiers.

5. The method of claim 1 , the method further comprising:

determining, by the online system, that the determined similarity score for the pair of unresolved identifiers exceeds an additional threshold similarity score; and

responsive to determining that the similarity score exceeds the additional threshold similarity score:

generating, by the online system, an edge between the pair of unresolved identifiers.

6. The method of claim 5 , wherein clustering the pair of unresolved identifiers further comprises:

determining, by the online system, that an edge exists between the pair of unresolved identifiers; and

responsive to determining that the edge exists between the pair of unresolved identifiers:

clustering, by the online system, the pair of unresolved identifiers into a cluster, the clustering indicating a prediction that the pair of unresolved identifiers in the cluster are associated with a common user.

7. The method of claim 5 , wherein the threshold similarity score differs from the additional threshold similarity score.

8. The method of claim 1 , wherein clustering of the pair of unresolved identifiers into a cluster further comprises clustering an additional unresolved identifier into the cluster with the pair of unresolved identifiers, the additional unresolved identifier having a similarity score with at least one unresolved identifier of the pair of unresolved identifiers that exceeds the threshold similarity score.

9. The method of claim 1 , wherein determining the reach of the content item further comprises excluding un-clustered unresolved identifiers from the reach determination.

10. A non-transitory computer-readable medium having instructions for execution by a processor causing the processor to:

identify, by an online system, a set of unresolved identifiers, wherein an unresolved identifier defines a context in which a client device accesses one or more of a plurality of online systems, and wherein the context has not been determined to be associated with a specific user;

identify, by the online system, for each unresolved identifier of the set of unresolved identifiers, information associated with the unresolved identifier, the information describing characteristics of one or more access events associated with the unresolved identifier;

for each pair of unresolved identifiers:

determine, by the online system, based on the information associated with each unresolved identifier, a similarity score for the pair of unresolved identifiers;

determine, by the online system, that the determined similarity score exceeds a threshold similarity score;

responsive to determining that the similarity score exceeds the threshold similarity score:

cluster, by the online system, the pair of unresolved identifiers into a cluster, the clustering indicating a prediction that the pair of unresolved identifiers in the cluster are associated with a common user; and

present a content item to one or more client devices associated with one or more unresolved identifiers of the set of unresolved identifiers; and

determine, by the online system, a reach of the content item based on the clusters including the one or more unresolved identifiers, wherein the reach is the quantity of users who have been presented with the content item.

11. The non-transitory computer-readable medium of claim 10 , wherein each unresolved identifier from the set of unresolved identifiers is selected from a group comprising a browser ID, a device ID, an HTML request, and an IP address.

12. The non-transitory computer-readable medium of claim 10 , wherein the characteristics of one or more access events associated with the unresolved identifier comprise at least one of a context of the one or more access events, content accessed during the one or more access events, actions performed during the one or more access events, and derived data learned from the one or more access events.

13. The non-transitory computer-readable medium of claim 10 , wherein the similarity score for the pair of unresolved identifiers is based on a similarity of the information associated with each unresolved identifier of the pair of unresolved identifiers.

14. The non-transitory computer-readable medium of claim 10 , wherein the instructions further cause the processor to:

determine, by the online system, that the determined similarity score for the pair of unresolved identifiers exceeds an additional threshold similarity score; and

responsive to determining that the similarity score exceeds the additional threshold similarity score:

generate, by the online system, an edge between the pair of unresolved identifiers.

15. The non-transitory computer-readable medium of claim 14 , wherein the instructions causing the processor to cluster the pair of unresolved identifiers further causes the processor to:

determine, by the online system, that an edge exists between the pair of unresolved identifiers; and

responsive to determining that the edge exists between the pair of unresolved identifiers:

cluster, by the online system, the pair of unresolved identifiers into a cluster, the clustering indicating a prediction that the pair of unresolved identifiers in the cluster are associated with a common user.

16. The non-transitory computer-readable medium of claim 14 , wherein the threshold similarity score differs from the additional threshold similarity score.

17. The non-transitory computer-readable medium of claim 10 , wherein the instructions causing the processor to cluster of the pair of unresolved identifiers into a cluster further cause the processor to cluster an additional unresolved identifier into the cluster with the pair of unresolved identifiers, the additional unresolved identifier having a similarity score with at least one unresolved identifier of the pair of unresolved identifiers that exceeds the threshold similarity score.

18. The non-transitory computer-readable medium of claim 10 , wherein the instructions causing the processor to determine the reach of the content item further cause the processor to exclude un-clustered unresolved identifiers from the reach determination.