Actionable KPI-driven segmentation

Patent metadata

Field	Value
Publication number	US-12505462-B2
Application number	US-202318132758-A
Country	US
Kind code	B2
Filing date	Apr 10, 2023
Priority date	Nov 14, 2018
Publication date	Dec 23, 2025
Grant date	Dec 23, 2025

Abstract

An improved analytics system generates actionable KPI-based customer segments. The analytics system determines predicted outcomes for a key performance indicator (KPI) of interest and a contribution value for each variable indicating an extent to which each variable contributes to predicted outcomes. Topics are generated by applying a topic model to the contribution values for the variables. Each topic comprises a group of variables with a contribution level for each variable that indicates the importance of each variable to the topic. User segments are generated by assigning each user to a topic based on attribution levels output by the topic model.

First claim

What is claimed is:

1 . A computer-implemented method, comprising:
receiving, by one or more processors, raw event data that includes user behavior, user attributes, and features from one or more data stores; converting, by the one or more processors, the raw event data to training data by at least assigning each user a topic related to the user attributes based at least in part on feature engineering; training, by the one or more processors using a machine learning model, a propensity model to:
predict a key performance indicator (KPI) of interest by combining the user behavior and the user attributes with known KPI outcomes on a per-user level, and
determine a likelihood of an outcome for an existing customer based on the combining of the user behavior and the user attributes and a predicted KPI outcome;
predicting, by the one or more processors using the trained propensity model, an outcome for the KPI of interest for each user of a plurality users based on the user behavior and the user attributes associated to at least one topic corresponding to a contribution value; determining, by the one or more processors using the trained propensity model, a likelihood of an outcome for an existing user of the plurality of users based on the user behavior and user attributes correlated with at least one historical outcome of interest; determining, by the one or more processors, whether the predicted outcome for at least a first user matches an actual outcome within a predefined threshold level by comparing the predicted outcome for at least the first user with the at least one historical outcome of interest; updating, by the one or more processors using the machine learning model, the propensity model with additional training data when the predicted outcome does not match, within the predefined threshold, the at least one historical outcome of interest; based on the updating of the propensity model, determining an impact of each reason code, of a plurality of reason codes, on the predicted outcome from the propensity model by at least encoding a contribution score into a weight value, each reason code representing a variable used by the propensity model to predict the outcome for the KPI of interest; in response to the determining of the impact, determining a contribution level for each of at least a portion of the plurality of reason codes indicating a rank of importance of each reason code in defining topic, of a plurality of topics; clustering user data of each user, of the plurality of users, to one or more clusters according to the contribution level from each of at least a portion of the plurality of reason codes; and based on the clustering, generating, by the one or more processors, a plurality of user segments, the generating of the plurality of user segments being indicative of assigning each of user, of the plurality of users, to at least one topic, of the plurality of topics.

2 . The computer-implemented method of claim 1 , wherein each contribution level comprises an integer weight determined by comparing raw contribution scores output by the propensity model.

3 . The computer-implemented method of claim 1 , wherein the determining of the impact is based on generating a user reason-code matrix that includes each reason code, of the plurality of reason codes.

4 . The computer-implemented method of claim 3 , wherein the determining of the contribution level is based on using the user-reason-code matrix as input and using a topic model to generate a topic matrix, the topic matrix includes the plurality of topics.

5 . The computer-implemented method of claim 4 , the method further comprising:
generating a descriptive segmentation results table based on the topic matrix and a user-topic matrix by assigning labels to each user segment.

6 . The computer-implemented method of claim 1 , wherein each contribution level comprises a raw contribution score output by the propensity model.

7 . The computer-implemented method of claim 1 , wherein the updating of the propensity model is based on the predicted outcome for the first user failing to match the actual outcome for the current user.

8 . The computer-implemented method of claim 1 , the method further comprising:
generating a targeted campaign for a particular user associated with a first customer segment based on a subset of reason codes associated with the first customer segment.

9 . The computer-implemented method of claim 1 , wherein the plurality of users are selected from a larger set of users based on the plurality of users having predicted outcomes within a predefined threshold level for the KPI of interest determined using the propensity model.

10 . One or more computer storage media storing computer-useable instructions that, when used by one or more computing devices, cause the one or more computing devices to perform operations comprising:
receiving, by one or more processors, raw event data that includes user behavior and user attributes; converting the raw event data to training data by at least assigning each user a topic related to the user attributes based at least in part on feature engineering; training, using a machine learning model, a propensity model to predict a key performance indicator (KPI) of interest by combining the user behavior and/or the user attributes with known KPI outcomes on a per-user level; training, using the machine learning model, the propensity model to determine a likelihood of an outcome for an existing customer based on the combined user behavior and user attributes and a predicted KPI outcome; predicting, using the trained propensity model, an outcome for the KPI of interest for each user, of a plurality of users based on the user behavior and the user attributes associated to at least one topic corresponding to a contribution value; determining, using the trained propensity model, a likelihood of an outcome for an existing user of the plurality of users based on the user behavior and user attributes correlated with at least one historical outcome of interest; determining whether the predicted outcome for at least a first user matches an actual outcome within a predefined threshold level by comparing the predicted outcome for at least the first user with the at least one historical outcome of interest associated with at least first user; updating, using the machine learning model, the propensity model with additional training data when the predicted outcome does not match, within the predefined threshold, the at least one historical outcome of interest; determining an impact of each reason code, of a plurality of reason codes, on the predicted outcome from the propensity model, each reason code representing a variable used by the propensity model to predict the outcome for the KPI of interest; in response to the determining of the impact, determining a contribution level for each of at least a portion of the plurality of reason codes indicating a rank of importance of each reason code in defining a topic, of a plurality of topics; clustering user data of each user, of the plurality of users, to one or more clusters according to the contribution level for each of at least the portion of the plurality of reason codes; and based on the clustering, generating, by the one or more processors, a plurality of user segments, the generating of the plurality of user segments being indicative of assigning each user, of the plurality of users, to at least one topic, of the plurality of topics.

11 . The computer storage media of claim 10 , wherein each contribution level comprises an integer weight determined by comparing raw contribution scores output by the propensity

Assignees

• Adobe Inc

Inventors

• Wu Xiang
• Yan Zhenyu
• Kuo Yi-Hong
• Dai Wuyang
• Bartik Polina
• Pani Abhishek

Classifications

• G06Q10/067

  Enterprise or organisation modelling · CPC title

• G06F17/16

  Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)} · CPC title

• G06N20/00

  Machine learning · CPC title

• G06Q10/06393

  Score-carding, benchmarking or key performance indicator [KPI] analysis · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title