Systems and methods for targeted email marketing

Patent metadata

Field	Value
Publication number	US-2020327577-A1
Application number	US-201916381380-A
Country	US
Kind code	A1
Filing date	Apr 11, 2019
Priority date	Apr 11, 2019
Publication date	Oct 15, 2020
Grant date	—

Abstract

A system for targeted email marketing includes one or more processors, and a memory storing instructions. When executed by the one or more processors, the instructions cause the system to perform operations including: generating email templates respectively corresponding to different customer marketing communications; sending first emails to first customers, in which each of the first emails includes one of the email templates in accordance with an initial allocation; compiling response information from the first customers and updating a bandit model in accordance with the response information, the bandit model for determining an allocation of the email templates to customers; determining, by the bandit model, a revised allocation of the email templates to second emails; and sending the second emails to second customers in accordance with the revised allocation.

First claim

1 . A system for automating targeted email marketing through machine learning, comprising:
one or more processors; and a memory storing instructions that, when executed by the one or more processors, cause the system to perform operations comprising:
generating a plurality of email templates respectively corresponding to different customer marketing communications;
sending a plurality of first emails to a plurality of first customers, wherein each of the plurality of first emails includes one of the plurality of email templates selected based on an initial allocation;
compiling response information received from the plurality of first customers in response to the plurality of first emails;
updating a bandit model based on the response information, wherein the bandit model is configured to determine an optimized allocation of the plurality of email templates to customers, wherein:
the response information includes (i) one or more customer characteristics corresponding to respective ones of the plurality of first customers, and (ii) one or more content characteristics corresponding to the plurality of email templates of the plurality of first emails received by the respective ones of the plurality of first customers, and
the bandit model determines a relationship between the one or more customer characteristics and the corresponding one or more content characteristics based on the response information and is configured to predict customer responses to different advertising elements associated with the plurality of email templates based on the relationship between the one or more customer characteristics and the corresponding one or more content characteristics;

determining whether the bandit model is to be re-trained based on a performance of the bandit model and an expected performance of the bandit model, the performance of the bandit model being determined based on (i) customer response estimation values for each of the plurality of email templates, and (ii) values associated with prior customer responses obtained from historical data, wherein:
if the performance of the bandit model is not below the expected performance of the bandit model for a number of consecutive cycles,
determining, by the bandit model, a revised allocation of the plurality of email templates to a plurality of second emails; and
sending the plurality of second emails to a plurality of second customers based on the revised allocation; and

if the performance of the bandit model is below the expected performance of the bandit model for the number of consecutive cycles,
resetting the bandit model, wherein resetting comprises:
sending a plurality of third emails to a plurality of third customers, wherein each of the plurality of third emails includes one of the plurality of email templates selected based on the initial allocation;
compiling additional response information received from the plurality of third customers in response to the plurality of third emails; and
updating the bandit model based on the additional response information.

2 . The system of claim 1 , wherein the bandit model is updated based on the relationship between the one or more customer characteristics and the corresponding one or more content characteristics.

3 . The system of claim 2 , wherein determining the revised allocation of the plurality of email templates to the plurality of second emails comprises:
obtaining one or more customer characteristics for each of the plurality of second customers; for each of the plurality of second customers, estimating, by the bandit model responsive to being updated, a plurality of predicted responses corresponding to the plurality of email templates based on the one or more customer characteristics for each of the plurality of second customers; and selecting one of the plurality of email templates for each of the plurality of second customers based on the plurality of predicted responses.

4 . (canceled)

5 . The system of claim 1 , wherein the operations further comprise:
determining that an optimized email template has been identified based on response information received from the plurality of second customers responsive to the plurality of second emails being sent to the plurality of second customers and one or more customer characteristics of the plurality of second customers; and sending a plurality of fourth emails with the optimized email template to a plurality of fourth customers corresponding to the one or more customer characteristics of the plurality of second customers.

6 . The system of claim 5 , wherein the operations further comprise:
determining the optimized email template based on the response information received from the plurality of second customers and additional response information received after the number of consecutive cycles, wherein each cycle comprises sending emails, compiling response information, updating the bandit model, and revising an allocation of the plurality of email templates.

7 . (canceled)

8 . The system of claim 1 , wherein the operations further comprise:
building the bandit model by using at least one of: a contextual bandit algorithm, an extreme gradient boosting algorithm, or a regression algorithm.

9 . The system of claim 1 , wherein the initial allocation is initialized as a uniform allocation of the plurality of email templates.

10 . The system of claim 1 , wherein the response information received from the plurality of first customers comprises a click through rate of the plurality of first emails.

11 . A method for targeted email marketing, comprising:
generating a plurality of email templates respectively corresponding to different customer marketing communications; determining a first number of customers for receiving each email template of the plurality of email templates; sending a plurality of first emails to a plurality of first customers, wherein each of the plurality of first emails includes one of the plurality of email templates selected based on an initial allocation; compiling response information received from the plurality of first customers in response to the plurality of first emails; updating a bandit model based on the response information received from the plurality of first customers, wherein the bandit model is configured to determine an optimized allocation of the plurality of email templates to customers, wherein:
the response information includes (i) one or more customer characteristics corresponding to respective ones of the plurality of first customers, and (ii) one or more content characteristics corresponding to the plurality of email templates of the plurality of first emails received by the respective ones of the plurality of first customers, and
the bandit model determines a relationship between the one or more customer characteristics and the corresponding one or more content characteristics based on the response information and is configured to predict customer responses to different advertising elements associated with the plurality of email templates based on the relationship between the one or more customer characteristics and the corresponding one or more content characteristics;
determining whether the bandit model is to be re-trained based on a performance of the bandit model and an expected performance of the bandit model, the performance of the bandit model being determined based on (i) customer response estimation values for each of the plurality of email templates, and (ii) values associated with prior customer responses obtained from historical data, wherein:
if the performance of the bandit model is not below the expected performance of the bandit model for a number of conse

Assignees

• Capital One Services Llc

Inventors

• Truong Edmond
• Shen Jie
• Rao Phanindra Venkatesh
• Vasisht Sunil Subrahmanyam
• Manion Patrick James
• Sun Meng

Classifications

• G06Q30/0246 Primary

  Traffic · CPC title

• G06Q30/0244 Primary

  Optimization · CPC title

• G06F18/2148

  characterised by the process organisation or structure, e.g. boosting cascade · CPC title

• G06F18/23

  Clustering techniques · CPC title

• H04L51/42

  Mailbox-related aspects, e.g. synchronisation of mailboxes · CPC title