Systems and methods for machine learning modeling of mixed media marketing

Patent metadata

Field	Value
Publication number	US-12579552-B1
Application number	US-202217902654-A
Country	US
Kind code	B1
Filing date	Sep 2, 2022
Priority date	Sep 2, 2022
Publication date	Mar 17, 2026
Grant date	Mar 17, 2026

Abstract

A system monitors impression data from a plurality of media channels including attributes of content presented on a respective media channel. The system inputs the impressions data and conversion data into a machine learning predictive model. The machine learning predictive model is trained by determining an impact of historical impression data on historical conversion data for each media channel. A machine learning predictive model incorporates an impressions/conversions predictive model that generates coefficients to measure attribution of different channel/campaign combinations to predicted conversions. The machine learning predictive model also incorporates an optimization model to effectively allocate marketing budget among various channels and campaigns in order to realize incremental conversions. The optimization model utilizes a constrained optimization framework based upon user-inputted budgets and budget constraints. Disclosed mixed media marketing predictive modeling provides a better understanding of budget allocation strategy for mixed media marketing factors.

First claim

What is claimed is:

1 . A computer-implemented method for improving multi-media content allocation comprising:
monitoring, by a computer, impression data input from a plurality of electronic communication channels into a machine learning predictive model, the impression data comprising one or more combinations of a plurality of attributes of impression content presented via one or more impression generators for each electronic communication channel of the plurality of electronic communication channels; inputting, by the computer, conversion data comprising one or more attributes of one or more transactions corresponding to the impression content presented via the plurality of electronic communication channels and impression data into the machine learning predictive model, wherein the machine learning predictive model is trained by:
determining, based on a transformation model, an impact one or more time-based carry-over effects of one or more combinations of historical impression data on historical conversion data for each electronic communication channel of the plurality of electronic communication channels; and
reducing overfitting of the machine learning predictive model by:
converting the impression data to one or more parameter features
corresponding to one or more attributes of the impression content; and
reducing, based on a corresponding significance of each of the one or more parameter features determined using a variance-covariance model corresponding to the one or more parameter features, a number of the one or more parameter features;
establishing, baseline values for one or more timing parameters corresponding to respective electronic communication channels, the timing parameters controlling a significance of impression data over time on predicted conversion tendencies;

executing, by the computer, the machine learning predictive model to predict model parameters for each electronic communication channel of the plurality of electronic communication channels, wherein the model parameters represent a conversion tendency for each electronic communication channel of the plurality of electronic communication channels;
updating, by the computer, the machine learning predictive model by adjusting one or more timing parameters in response to differences between the predicted conversion tendencies and actual conversions observed for corresponding recent impression content to refine weighting of prior impressions and improve accuracy of subsequent executions for allocating content selection across each electronic communication channel; and
allocating, by the computer, an adjusted electronic communication channel content to each electronic communication channel of the plurality of electronic communication channels based on an optimization model that models one or more combinations of the model parameters according to an objective function and one or more optimization constraints.

2 . The method of claim 1 , wherein each electronic communication channel of the plurality of electronic communication channels corresponds to a respective campaign.

3 . The method of claim 1 , further comprising inputting, by the computer, a plurality of factors into the machine learning predictive model, wherein the plurality of factors comprise at least one of: seasonality, customer value, agent score, awareness, consideration, affinity, consumer confidence index (CCI), consumer price index (CPI), gasoline price, or unemployment rate.

4 . The method of claim 1 , wherein modeling one or more combinations of the model parameters according to an objective function and one or more optimization constraints generates a maximum total predicted conversions.

5 . The method of claim 4 , wherein the one or more optimization constraints comprise a marketing expenditure budget that serves as equality constraint in Legrangian analysis.

6 . The method of claim 4 , wherein the one or more optimization constraints comprise one or both of upper bounds for total marketing expenditures associated with respective electronic communication channels, and lower bounds for the total marketing expenditures associated with the respective electronic communication channels.

7 . The method of claim 4 , further comprising determining an optimal mix of impressions to achieve the maximum total predicted conversions based upon the model parameters for each electronic communication channel of the plurality of electronic communication channels.

8 . The method of claim 7 , further comprising generating the maximum total predicted conversions during a given time period and determining the optimal mix of impressions to achieve the maximum total predicted conversions during the given time period.

9 . The method of claim 8 , further comprising converting the optimal mix of impressions to optimal expenditures associated with the optimal mix of impressions.

10 . A system for improving multi-media content allocation comprising:
an analytical engine server comprising:
a mixed marketing model configured for generating conversion parameters for a plurality of marketing channel-campaign combinations by inputting impression parameters for a plurality of impression channel-campaign combinations into a machine learning predictive model,
wherein each impression parameter represents a respective impression tendency attributed to a combination of a respective impression channel and a respective marketing campaign,
wherein each conversion parameter represents a respective conversion tendency attributed to a combination of a respective marketing channel and a respective marketing campaign,
wherein the machine learning predictive model is continually trained by:
inputting one or more combinations of training data representing historical impressions impression data and historical conversions conversion data of an enterprise;
determining, based on a transformation model, one or more time-based carry-over effects of the historical impression data on historical conversion data for each marketing channel-campaign combination of the plurality of marketing channel-campaign combinations; and
reducing overfitting of the machine learning predictive model by:
converting the impression data to one or more parameter features corresponding to one or more attributes of the impression content; and
reducing, based on a corresponding significance of each of the one or more parameter features determined using a variance-covariance model corresponding to the one or more parameter features, a number of the one or more parameter features;

establishing, baseline values for one or more timing parameters corresponding to respective electronic communication channels, the timing parameters controlling a significance of impression data over time on predicted conversion tendencies;

a marketing mix optimization model configured for calculating maximum total predicted conversions by inputting the conversion parameters for the plurality of marketing channel-campaign combinations received from the mixed marketing model, wherein the marketing mix optimization model applies an objective function the plurality of marketing channel-campaign combinations subject to one or more optimization constraints based on marketing expenditure; and further configured for determining an optimal mix of impressions to achieve the maximum total predicted conversions; wherein the machine learning predictive model is updated by adjusting one or more timing parameters in response to differences between the predicted conversion tendencies and actual conversions observed for corresponding recent impression content to refine weighting of prior impressions and improve accuracy of subsequent executions for allocating

Assignees

• Massachusetts Mutual Life Insurance Co

Inventors

• Ma Yuzhi
• House Katie
• D'Angelo Sean
• Peskin Bisakha
• Ahani Asieh
• Halbleib Jennifer
• Baldenko Alex
• Fox Adam
• Merritt Sears

Classifications

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G06Q30/0202 Primary

  Market predictions or forecasting for commercial activities · CPC title