Machine Learning Ad Targeting via User Feedback Loops
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital advertising technologies are ineffective in targeting the right prospects at the right time due to reliance on inferences from vast amounts of digital data, leading to high costs and privacy concerns, and regulatory challenges.
Innovation Solution
A machine learning-based system that identifies user intent to purchase by analyzing digital media communications, reducing reliance on demographic and social media data, and using feedback loops to update models for more accurate targeting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current digital ad targeting technologies use demographic, behavior, and social networking data to identify prospects, then targeting becomes more targeted than traditional methods, but the system still relies on inferences about likelihood of purchase rather than actual intent, leading to ineffective ad placement at the right time
Solution Approach 1:
The system implements feedback loops where users interact with ads and provide feedback on their interest levels. This feedback is used to train machine learning models to better predict actual purchase intent rather than relying solely on inferred data, thereby improving the reliability of intent prediction while maintaining targeting accuracy
Solution Approach 2:
The patent replaces traditional mechanical data processing systems with machine learning and artificial intelligence models that can process and interpret complex user behavior patterns, social networking data, and demographic information to accurately predict purchase intent without relying on simple inference rules
2Loss of information
If ad targeting systems process massive amounts of varied digital data including behavior data, social networking data, and demographic data, then more comprehensive targeting insights are obtained, but the technical difficulty and cost of processing and storing this data increases significantly
Solution Approach 1:
The system segments the data processing task by using distributed computing architecture where different components handle different types of data (behavior data, social networking data, demographic data) separately before integrating them through machine learning models, thereby reducing the complexity of processing massive varied datasets
Solution Approach 2:
The patent introduces machine learning models as intermediary components that simplify the processing of complex data relationships. These models act as mediators between raw data and advertising decisions, automatically extracting meaningful patterns without requiring complex manual processing systems
3Measurement precision
If complex machine learning and artificial intelligence systems are employed to make ad targeting decisions, then prediction accuracy improves, but computing costs increase significantly due to per-prediction or per-hour charge rates
Solution Approach 1:
The system performs preliminary actions by pre-training machine learning models offline using historical data and computational resources that are not charged during actual ad delivery. This allows the models to make predictions efficiently during runtime without incurring per-prediction costs, as the heavy computational work is completed beforehand
4Loss of information
If advertisers access and use demographic and behavior data from digital media platforms, then more comprehensive targeting is achieved, but user privacy concerns and regulatory restrictions limit data availability and access
Solution Approach 1:
The system uses feedback mechanisms where users directly indicate their interest in ads, providing explicit consent-based data points that bypass privacy restrictions. This feedback-driven approach allows the system to work with user-provided data rather than requiring access to restricted demographic and behavior data from platforms
Data Source
AI summary
A system and method for identifying prospects with a buying intent and connecting them with relevant businesses. An example method may comprise obtaining a training dataset; applying a first scoring algorithm to obtain a first score for each entry in the training dataset; receiving one or more scores from a user for one or more entries in the training dataset; rescoring the training dataset based on the one or more scores received from the user; creating a deep learning model based on the rescored dataset; obtaining digital media posts comprising data from one or more digital media platforms; scoring each received digital media post by using the deep learning model; providing certain scored digital media posts to the user; receiving a second score from the user; and updating the deep learning model based on the second score.