Patent Issued for Machine learning system for routing optimization based on historical performance data (USPTO 11128754): Allstate Insurance Company

2021 OCT 07 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- A patent by the inventors Dunn, Ryan (Grayslake, IL, US), Lundsgaard, Soren K. (Inverness, IL, US), Pham, Anthony (Northbrook, IL, US), filed on November 16, 2020, was published online on September 21, 2021, according to news reporting originating from Alexandria, Virginia, by NewsRx correspondents.

Patent number 11128754 is assigned to Allstate Insurance Company (Northbrook, Illinois, United States).

The following quote was obtained by the news editors from the background information supplied by the inventors: “Aspects of the disclosure relate to processing systems. In particular, aspects of the disclosure relate to processing systems having a machine learning engine and machine learning datasets.

“In some instances, individuals may request to establish voice call sessions with employees of an enterprise organization (e.g., customer service representative, agents, and/or other employees). The enterprise organizations may route these requests to available representatives in a generic manner based on representative availability. In many instances, however, this may result in sub-optimal call routing and thus may result in sub-optimal communication between employees and clients.”

In addition to the background information obtained for this patent, NewsRx journalists also obtained the inventors’ summary information for this patent: “Aspects of the disclosure provide effective, efficient, scalable, and convenient technical solutions that address and overcome the technical problems associated with call routing.

“In accordance with one or more embodiments, a computing platform comprising at least one processor, a communication interface communicatively coupled to the at least one processor, and memory may receive, from one or more client devices, a number of requests to establish a voice call session. The computing platform may identify, based on phone numbers corresponding to the one or more client devices, demographic information corresponding to users of the one or more client devices. The computing platform may score, using a machine learning model and based on the demographic information and performance data for each of a plurality of representatives, potential client-representative combinations, where: 1) the score for each potential client-representative combination indicates a likelihood of a successful outcome resulting from establishing a voice call session between the respective client and representative, and 2) scoring the potential client-representative combinations is based on fall off rates for each of the users of the one or more client devices, where the fall off rates indicate changes in representative effectiveness with each of the users of the one or more client devices as a hold time increases. The computing platform may adjust the scores based on a historical frequency of interaction between each representative and clients corresponding to the identified demographic information. The computing platform may select, based on the adjusted scores, at least one of the potential client-representative combinations. The computing platform may establish a voice call session for the at least one of the potential client-representative combinations.

“In one or more instances, the computing platform may collect representative performance data indicating sales performance of each of the plurality of representatives during their communication with clients corresponding to different demographics data. The computing platform may train, using the representative performance data, the machine learning model, which may configure the machine learning model to output the scores.

“In one or more instances, the adjusted scores may indicate how likely a representative of a particular potential client-representative pair is to make a sale to the corresponding client of the particular potential client-representative pair. In one or more instances, the demographics information may indicate one or more of: gender, previous accidents, previous contact with the enterprise organization, age, family composition, marital status, education information, income information, or geographic information corresponding to the users of the one or more client devices.

“In one or more instances, scoring the potential client-representative combinations may further include: 1) comparing the number of requests to establish a voice call session to a number of available representatives; 2) based on identifying that the number of requests to establish a voice call session exceeds the number of available representatives, scoring, using the machine learning model, combinations of each available representative with each user corresponding to the requests to establish a voice call session to identify an optimal user for each representative to serve; and 3) based on identifying that the number of available representatives exceeds the number of requests to establish the voice call session, scoring, using the machine learning model, combinations of each available representative with each user corresponding to the requests to establish a voice call session to identify an optimal representative to serve each user.”

The claims supplied by the inventors are:

“1. A computing platform comprising: at least one processor; a communication interface communicatively coupled to the at least one processor; and memory storing computer-readable instructions that, when executed by the at least one processor, cause the computing platform to: receive, from one or more client devices, a number of requests to establish a voice call session; identify, based on phone numbers corresponding to the one or more client devices, demographic information corresponding to users of the one or more client devices; score, using a machine learning model and based on the demographic information and performance data for each of a plurality of representatives, potential client-representative combinations, wherein: the score for each potential client-representative combination indicates a likelihood of a successful outcome resulting from establishing a voice call session between the respective client and representative, and the score for each potential client-representative combinations is based on fall off rates for each of the users of the one or more client devices, wherein the fall off rates indicate changes in the likelihood of a successful outcome of a representative with each of the users of the one or more client devices as a time during which the client is put on hold increases; adjust the scores based on a historical frequency of interaction between each representative and clients corresponding to the identified demographic information; select, based on the adjusted scores, at least one of the potential client-representative combinations; and establish a voice call session for the at least one of the potential client-representative combinations.

“2. The computing platform of claim 1, wherein the memory stores additional computer-readable instructions that, when executed by the at least one processor, further cause the computing platform to: collect representative performance data indicating sales performance of each of the plurality of representatives during their communication with clients corresponding to different demographics data; and train, using the representative performance data, the machine learning model, wherein training the machine learning model configures the machine learning model to output the scores.

“3. The computing platform of claim 1, wherein the adjusted scores indicate how likely a representative of a particular potential client-representative pair is to make a sale to the corresponding client of the particular potential client-representative pair.

“4. The computing platform of claim 1, wherein the demographics information indicates one or more of: age, family composition, marital status, education information, income information, or geographic information corresponding to the users of the one or more client devices.

“5. The computing platform of claim 1, wherein scoring the potential client-representative combinations further comprises: comparing the number of requests to establish a voice call session to a number of available representatives; based on identifying that the number of requests to establish a voice call session exceeds the number of available representatives, scoring, using the machine learning model, combinations of each available representative with each user corresponding to the requests to establish a voice call session to identify an optimal user for each representative to serve; and based on identifying that the number of available representatives exceeds the number of requests to establish the voice call session, scoring, using the machine learning model, combinations of each available representative with each user corresponding to the requests to establish a voice call session to identify an optimal representative to serve each user.

“6. The computing platform of claim 1, wherein the fall off rates indicate a tolerance for being on hold for the users of the one or more client devices.

“7. The computing platform of claim 1, wherein the memory stores additional computer-readable instructions that, when executed by the at least one processor, further cause the computing platform to: identify the plurality of representatives, wherein identifying the plurality of representatives comprises: identifying representatives that are currently available to engage in a voice call session, and identifying representatives that are not currently available, but that will likely be available within a predetermined amount of time.

“8. The computing platform of claim 1, wherein adjusting the scores further comprises adjusting, based on time information corresponding to the performance data.

“9. The computing platform of claim 1, wherein scoring the potential client-representative combinations comprising: scoring the potential client-representative combinations at a first time and one or more second times, wherein the one or more second times are later than the first time; and computing an overall score for each client-representative combination by adding the scores for the first time and the one or more second times.

“10. The computing platform of claim 9, wherein selecting the at least one of the potential client-representative combinations comprises selecting the at least one of the potential client-representative combinations based on the overall scores.

“11. The computing platform of claim 1, wherein the memory stores additional computer-readable instructions that, when executed by the at least one processor, further cause the computing platform to: receive additional performance data from the selected at least one of the potential client-representative combinations; and update, based on the additional performance data, the machine learning model.

“12. A method comprising: at a computing platform comprising at least one processor, a communication interface, and memory: receiving, from one or more client devices, a number of requests to establish a voice call session; identifying, based on phone numbers corresponding to the one or more client devices, demographic information corresponding to users of the one or more client devices; scoring, using a machine learning model and based on the demographic information and performance data for each of a plurality of representatives, potential client-representative combinations, wherein: the score for each potential client-representative combination indicates a likelihood of a successful outcome resulting from establishing a voice call session between the respective client and representative, the score for each of the potential client-representative combinations is based on fall off rates for each of the users of the one or more client devices, wherein the fall off rates indicate changes in the likelihood of a successful outcome of a representative with each of the users of the one or more client devices as a time during which the client is put on hold increases, and the score for each of the potential client-representative combinations is generating by scoring the potential client-representative combinations at a first time and one or more second times, wherein the one or more second times are later than the first time; computing an overall score for each client-representative combination by adding the scores for the first time and the one or more second times; adjusting the overall scores based on a historical frequency of interaction between each representative and clients corresponding to the identified demographic information; selecting, based on the adjusted scores, at least one of the potential client-representative combinations; and establishing a voice call session for the at least one of the potential client-representative combinations.

“13. The method of claim 12, further comprising: collecting representative performance data indicating sales performance of each of the plurality of representatives during their communication with clients corresponding to different demographics data; and training, using the representative performance data, the machine learning model, wherein training the machine learning model configures the machine learning model to output the scores.

“14. The method of claim 12, wherein the adjusted scores indicate how likely a representative of a particular potential client-representative pair is to make a sale to the corresponding client of the particular potential client-representative pair.

“15. The method of claim 12, wherein the demographics information indicates one or more of: age, family composition, marital status, education information, income information, or geographic information corresponding to the users of the one or more client devices.

“16. The method of claim 12, wherein scoring the potential client-representative combinations further comprises: comparing the number of requests to establish a voice call session to a number of available representatives; based on identifying that the number of requests to establish a voice call session exceeds the number of available representatives, scoring, using the machine learning model, combinations of each available representative with each user corresponding to the requests to establish a voice call session to identify an optimal user for each representative to serve; and based on identifying that the number of available representatives exceeds the number of requests to establish the voice call session, scoring, using the machine learning model, combinations of each available representative with each user corresponding to the requests to establish a voice call session to identify an optimal representative to serve each user.

“17. The method of claim 12, wherein the fall off rates indicate a tolerance for being on hold for the users of the one or more client devices.”

There are additional claims. Please visit full patent to read further.

URL and more information on this patent, see: Dunn, Ryan. Machine learning system for routing optimization based on historical performance data. U.S. Patent Number 11128754, filed November 16, 2020, and published online on September 21, 2021. Patent URL: http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=11128754.PN.&OS=PN/11128754RS=PN/11128754

(Our reports deliver fact-based news of research and discoveries from around the world.)