Prioritization of electronic communications

Abstract

Methods, systems, and apparatus for prioritizing communications are described. Metadata that characterizes an electronic communication is obtained and a machine learning algorithm is applied to the metadata to generate a scoring model. A score for the electronic communication is generated based on the scoring model.

What is claimed is: 
     
         1 . A system for selecting an incoming electronic communication for response, the system comprising:
 one or more hardware processors programmed to perform operations comprising:
 accessing incoming communication data describing a first incoming electronic communication from a first user; 
 accessing lead information associated with the first user; 
 using a trained scoring model and the lead information to determine an estimated propensity of the first user to purchase a first product, the estimated propensity describing a probability that the first user will purchase the first product, the trained scoring model being trained using training data that comprises historical purchase data describing historical purchase of the first product made by a plurality of users who are not the first user; 
 determining a first priority score for the first incoming electronic communication, the first priority score based at least in part on the estimated propensity of the first user to purchase the first product; and 
 initiating a first outgoing electronic communication to a device associated with the first user, the initiating based at least in part on the first priority score. 
   
     
     
         2 . The system of  claim 1 , the operations further comprising determining an estimated income from the first user; the estimated income based at least in part on the estimated propensity of the first user to purchase the first product and a value of a sale of the first product the first priority score being based at least in part on the estimated income from the first user. 
     
     
         3 . The system of  claim 2 , the value of the first product being at least one of a profit associated with a sale of the first product and a revenue associated with a sale of the first product. 
     
     
         4 . The system of  claim 1 ; the trained scoring model comprising a first linear function, wherein the determining of the estimated propensity of the first user to purchase the first product comprises evaluating the first linear function, the system the operations further comprising:
 when the first product is a new product, determining an additional term for the first linear function, the additional term being based on a set of coefficients for the first linear function when the trained scoring model is trained for a second product; and   evaluating the first linear function with the additional term.   
     
     
         5 . The system of  claim 4 , the trained scoring model including a term to rank a new product or a new advertising channel within a predefined range, the predefined range based on an aggregate mean priority score generated by the trained scoring model. 
     
     
         6 . The system of  claim 1 , the operations further comprising combining the historical purchase data with second data describing a second product to generate the training data. 
     
     
         7 . The system of  claim 1 , the operations further comprising determining an advertising channel associated with the first incoming electronic communication based at least in part on an indicated recipient of the first incoming electronic communication. 
     
     
         8 . The system of  claim 1 , the operations further comprising:
 accessing user identifier data describing the first user; and   mapping the user identifier data to a location of the first user, the first priority score being based at least in part on the location of the first user.   
     
     
         9 . The system of  claim 1 , the first priority score being determined by a lead scoring service executing at the one or more hardware processors, wherein the operations further comprise determining to perform a comprehensive evaluation of the first incoming electronic communication based at least in part on an available capacity of the lead scoring service when the first incoming electronic communication is received. 
     
     
         10 . A method of selecting an incoming electronic communication for response, the method comprising:
 accessing, by a lead scoring service, incoming communication data describing a first incoming electronic communication from a first user, the lead scoring service being executed by one or more hardware processor;   accessing, by the lead scoring service, lead information associated with the first user;   using a trained scoring model and the lead information to determine an estimated propensity of the first user to purchase a first product, the estimated propensity describing a probability that the first user will purchase the first product, the trained scoring model being trained using training data that comprises historical purchase data describing historical purchase of the first product made by a plurality of users who are not the first user;   determining, by the lead scoring service, a first priority score for the first incoming electronic communication, the first priority score based at least in part on the estimated propensity of the first user to purchase the first product; and   initiating a first outgoing electronic communication to a device associated with the first user, the initiating based at least in part on the first priority score.   
     
     
         11 . The method of  claim 10 , further comprising determining an estimated income from the first user, the estimated income based at least in part on the estimated propensity of the first user to purchase the first product and a value of a sale of the first product the first priority score being based at least in part on the estimated income from the first user. 
     
     
         12 . The method of  claim 11 , the value of the first product being at least one of a profit associated with a sale of the first product and a revenue associated with a sale of the first product. 
     
     
         13 . The method of  claim 10 , the trained scoring model comprising a first linear function, wherein the determining of the estimated propensity of the first user to purchase the first product comprises evaluating the first linear function, the method further comprising:
 when the first product is a new product, determining an additional term for the first linear function, the additional term being based on a set of coefficients for the first linear function when the trained scoring model is trained for a second product; and   evaluating the first linear function with the additional term.   
     
     
         14 . The method of  claim 13 , the trained scoring model including a term to rank a new product or a new advertising channel within a predefined range, the predefined range based on an aggregate mean priority score generated by the trained scoring model. 
     
     
         15 . The method of  claim 10 , further comprising combining the historical purchase data with second data describing a second product to generate the training data. 
     
     
         16 . The method of  claim 10 , further comprising determining an advertising channel associated with the first incoming electronic communication based at least in part on an indicated recipient of the first incoming electronic communication. 
     
     
         17 . The method of  claim 10 , further comprising:
 accessing user identifier data describing the first user; and   mapping the user identifier data to a location of the first user, the first priority score being based at least in part on the location of the first user.   
     
     
         18 . The method of  claim 10 , further comprising determining, by the lead scoring service, to perform a comprehensive evaluation of the first incoming electronic communication, the determining based at least in part on an available capacity of the lead scoring service when the first incoming electronic communication is received. 
     
     
         19 . A machine-readable medium comprising instructions thereon, that when executed by one or more hardware processor, cause the one or more hardware processor to perform operations comprising:
 accessing incoming communication data describing a first incoming electronic communication from a first user;   accessing lead information associated with the first user;   using a trained scoring model and the lead information to determine an estimated propensity of the first user to purchase a first product, the estimated propensity describing a probability that the first user will purchase the first product, the trained scoring model being trained using training data that comprises historical purchase data describing historical purchase of the first product made by a plurality of users who are not the first user;   determining a first priority score for the first incoming electronic communication, the first priority score based at least in part on the estimated propensity of the first user to purchase the first product; and   initiating a first outgoing electronic communication to a device associated with the first user, the initiating based at least in part on the first priority score.   
     
     
         20 . The medium of  claim 19 , the operations further comprising determining an estimated income from the first user, the estimated income based at least in part on the estimated propensity of the first user to purchase the first product and a value of a sale of the first product the first priority score being based at least in part on the estimated income from the first user.