User similarity groups for on-line marketing

Abstract

Methods and apparatus for finding similar on-line users for advertisement or content targeting are disclosed. In one embodiment, a plurality of user data sets associated with a plurality of user identifiers for a plurality of anonymous users are obtained, and each user data set of each user identifier specifies one or more user attributes and on-line user events that have occurred for such user identifier. For each attribute, a correlation to a success metric value is determined for a particular type of event or attribute that has occurred for a plurality of user identifiers that are each associated with such attribute. The plurality of user identifiers and associated data sets are clustered into a plurality of user groups that each has similar data sets by weighting based on the attributes' relative correlation to the success metric.

1 . A method for finding similar on-line users for advertisement or content targeting, the method comprising:
 by a plurality of processor nodes of a grouping system, receiving a query and together obtaining a plurality of user data sets associated with a plurality of user identifiers for a plurality of anonymous users, wherein each user data set of each user identifier specifies one or more user attributes and on-line user events that have occurred for such user identifier, wherein the on-line user events each comprises an exposure type of event for exposing a user to an on-line advertisement impression or content or an interaction type of event for a user action that is performed on an advertisement impression or content;   by the plurality of processor nodes of the grouping system and for each user attribute, counting a success metric for a particular interaction type of event for a user action that has occurred for a plurality of user identifiers that are each associated with such user attribute to obtain a success metric count for such user attribute; and   by the plurality of processor nodes of the grouping system, together clustering the plurality of user identifiers and associated data sets into a plurality of user groups that each has similar data sets by weighting the user attributes based on the user attributes' relative success metric counts.   
     
     
         2 . The method of  claim 1 , further comprising filtering any user identifiers that fail to meet a predetermined set of user identifier criteria prior to performing the operation for clustering. 
     
     
         3 . The method of  claim 2 , wherein the predetermined set of user identifier criteria specifies a minimum count threshold for each of the user identifier's associated user attributes. 
     
     
         4 . The method of  claim 2 , wherein the predetermined set of user identifier criteria specifies a minimum count threshold for each user identifier's associated events. 
     
     
         5 . The method of  claim 4 , wherein the predetermined set of user identifier criteria also specifies a particular time period or frequency for the minimum count threshold. 
     
     
         6 . The method of  claim 2 , further comprising filtering any user attributes that fail to meet a predetermined set of attribute criteria prior to performing the operation for clustering. 
     
     
         7 . The method of  claim 6 , wherein the predetermined set of attribute criteria specifies that if a particular interaction type of event is present in a user identifier's data set, then an exposure type of event must also be present in the user identifier's data set or any of the user identifier's user attributes that are associated with such particular interaction type of event are filtered from the clustering operation. 
     
     
         8 . The method of  claim 6 , wherein the predetermined set of attribute criteria specifies a minimum count threshold for user identifiers that are associated with each user attribute. 
     
     
         9 . The method of  claim 6 , wherein the predetermined set of attribute criteria specifies a maximum count threshold for user identifiers that are associated with each user attribute. 
     
     
         10 . The method of  claim 6 , wherein the predetermined set of attribute criteria specifies that each user identifier not be associated with any conflicting user attributes and to filter such conflicting user attributes for such user identifier if present. 
     
     
         11 . The method of  claim 1 , wherein the success metric count for user each attribute is a count of click events for user identifiers that are associated with such user attribute. 
     
     
         12 . The method of  claim 1 , wherein the success metric count for each user attribute is a count of conversion events for user identifiers that are associated with such user attribute. 
     
     
         13 . The method of  claim 1 , further comprising stopping the clustering based on an evaluation metric of the clustering with respect to a golden set of the user identifiers that are known to belong to similar users. 
     
     
         14 . The method of  claim 13 , wherein the golden set of user identities is known to belong to similar users based on having the same login information. 
     
     
         15 . The method of  claim 13 , wherein the golden set of user identities is known to belong to similar users based on similar event history. 
     
     
         16 . The method of  claim 13 , further comprising repeating the clustering based on a different golden set of user identifiers. 
     
     
         17 . The method of  claim 13 , further comprising repeating the clustering based on a different evaluation metric. 
     
     
         18 . The method of  claim 1 , further comprising:
 by the grouping system and for each user attribute, counting a different success metric value for a different interaction type of event for a user action that has occurred for a plurality of user identifiers that are each associated with such user attribute to obtain a different success metric count; and   repeating the operation of clustering based on the different success metric count of each user attribute.   
     
     
         19 . A grouping system for finding similar on-line users for advertisement or content targeting, the system comprising:
 at least a grouping processor;   at least a memory; and   an interface for communicating with a plurality of processor nodes,   wherein the grouping processor and/or memory are configured to perform or initiate the following operations in coordination with the processor nodes via the interface:
 sending a query via the interface to the processor nodes to obtain a plurality of user data sets associated with a plurality of user identifiers for a plurality of anonymous users, wherein each user data set of each user specifies one or more user attributes and on-line user events that have occurred for such user, wherein the on-line user events each comprises an exposure type of event for exposing a user to an on-line advertisement impression or content or an interaction type of event for a user action that is performed on an advertisement impression or content; 
 for each user attribute, counting a success metric for a particular interaction type of event for a user action that has occurred for a plurality of user identifiers that are each associated with such user attribute to obtain a success metric count for such user attribute; and 
 sending a query via the interface to the processor nodes to cluster the plurality of user identifiers and associated data sets into a plurality of user groups that each has similar data sets by weighting the user attributes based on the user attributes' relative success metric counts. 
   
     
     
         20 . At least one non-transitory computer readable storage medium having computer program instructions stored thereon that are arranged to perform the following operations:
 by a plurality of processor nodes of a grouping system, receiving a query and together obtaining a plurality of user data sets associated with a plurality of user identifiers for a plurality of anonymous users, wherein each user data set of each user identifier specifies one or more user attributes and on-line user events that have occurred for such user identifier, wherein the on-line user events each comprises an exposure type of event for exposing a user to an on-line advertisement impression or content or an interaction type of event for a user action that is performed on an advertisement impression or content;   by the plurality of processor nodes of the grouping system and for each user attribute, counting a success metric for a particular interaction type of event for a user action that has occurred for a plurality of user identifiers that are each associated with such user attribute to obtain a success metric count for such user attribute; and   by the plurality of processor nodes of the grouping system, together clustering the plurality of user identifiers and associated data sets into a plurality of user groups that each has similar data sets by weighting the user attributes based on the user attributes' relative success metric counts.   
     
     
         21 . The system of  claim 19 , wherein the grouping processor and/or memory are further configured for filtering, or sending a query to the processor nodes to together filter, any user identifiers that fail to meet a predetermined set of user identifier criteria prior to performing the operation for clustering, wherein the predetermined set of user identifier criteria specifies a minimum count threshold for each of the user identifier's associated user attributes. 
     
     
         22 . The system of  claim 19 , wherein the grouping processor and/or memory are further configured for filtering, or sending a query to the processor nodes to together filter, any user identifiers that fail to meet a predetermined set of user identifier criteria prior to performing the operation for clustering, wherein the predetermined set of user identifier criteria specifies a minimum count threshold for each user identifier's associated events. 
     
     
         23 . The system of  claim 23 , wherein the predetermined set of user identifier criteria also specifies a particular time period or frequency for the minimum count threshold. 
     
     
         24 . The system of  claim 19 , wherein the grouping processor and/or memory are further configured for filtering, or sending a query to the processor nodes to together filter, any user attributes that fail to meet a predetermined set of attribute criteria prior to performing the operation for clustering, wherein the predetermined set of attribute criteria specifies that if a particular interaction type of event is present in a user identifier's data set, then an exposure type of event must also be present in the user identifier's data set or any of the user identifier's user attributes that are associated with such particular interaction type of event are filtered from the clustering operation. 
     
     
         25 . The system of  claim 19 , wherein the grouping processor and/or memory are further configured for filtering, or sending a query to the processor nodes to together filter, any user attributes that fail to meet a predetermined set of attribute criteria prior to performing the operation for clustering, wherein the predetermined set of attribute criteria specifies a minimum count threshold for user identifiers that are associated with each user attribute and a maximum count threshold for user identifiers that are associated with each user attribute. 
     
     
         26 . The grouping system of  claim 19 , wherein the success metric count for user each attribute is a count of click events for user identifiers that are associated with such user attribute. 
     
     
         27 . The system of  claim 19 , wherein the success metric count for each user attribute is a count of conversion events for user identifiers that are associated with such user attribute. 
     
     
         28 . The system of  claim 19 , wherein the grouping processor and/or memory are configured for stopping the clustering based on an evaluation metric of the clustering with respect to a golden set of the user identifiers that are known to belong to similar users. 
     
     
         29 . The system of  claim 19 , wherein the grouping processor and/or memory are configured for sending a query via the interface to the processor nodes for repeating the clustering based on a different evaluation metric. 
     
     
         30 . The system of  claim 19 , wherein the grouping processor and/or memory are further configured for
 for each user attribute, counting a different success metric value for a different interaction type of event for a user action that has occurred for a plurality of user identifiers that are each associated with such user attribute to obtain a different success metric count; and   sending a query via the interface to the processor nodes for repeating the operation of clustering based on the different success metric count of each user attribute.