ML Conversion Prediction for Promotion Campaigns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online services face challenges in accurately reporting conversion counts for promotion campaigns, as they often cannot detect or measure conversions that occur off-site, leading to under-reporting due to limited access to users' online behaviors on external internet locations.
Innovation Solution
A machine learning model is trained to predict conversion counts for non-measurable subscribers based on their online behaviors on sites under the online service's control, with an estimated error rate determination, and these predictions are included in overall conversion counts if the error rate meets a predetermined threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If online services rely on traditional conversion tracking methods, then implementation is simple, but conversion count accuracy deteriorates due to inability to detect off-site conversions
Solution Approach 1:
The patent introduces an intermediary machine learning model that acts as a mediator between the online service and the conversion tracking process. This model predicts off-site conversions by analyzing user behavior patterns and promotion exposure data, enabling accurate conversion counting without requiring direct integration with external websites. The ML model processes input data about user interactions and promotion exposure to generate predicted conversion counts, resolving the contradiction by adding computational complexity to achieve measurement precision.
Solution Approach 2:
The patent replaces traditional mechanical/conventional tracking mechanisms with a machine learning-based predictive system. Instead of relying on direct technical integration or cookies to track off-site conversions, the system uses ML algorithms to infer conversions from behavioral patterns and promotion exposure data. This substitution enables accurate measurement of off-site conversions while managing system complexity through automated prediction rather than complex tracking infrastructure.
2Quantity of substance
If online services attempt to track all user behaviors including off-site actions, then conversion reporting completeness improves, but system complexity and data processing requirements worsen
Solution Approach 1:
The patent extracts only the necessary features and data elements required for conversion prediction from the vast amount of available user behavior data. By identifying and processing only the most relevant features related to promotion exposure and user interactions, the system achieves complete conversion reporting without processing all possible user behavior data. This extraction approach reduces data processing complexity while maintaining data completeness for conversion counting.
Solution Approach 2:
The patent segments the conversion tracking process into distinct components: data collection, feature extraction, model prediction, and conversion counting. This segmentation allows the system to handle large volumes of data by processing it through structured stages rather than attempting to analyze everything simultaneously. The segmentation reduces overall system complexity by dividing the complex task of tracking all user behaviors into manageable, sequential processing steps.
3Measurement precision
If online services use machine learning prediction for off-site conversions, then conversion count accuracy improves, but measurement and processing time increases
Solution Approach 1:
The patent applies preliminary action by pre-processing and storing user behavior data and promotion exposure information before conversion events occur. The system maintains historical data about user interactions and promotion exposures, enabling the ML model to make rapid predictions when conversion events are detected. This preliminary data preparation reduces the time required for real-time conversion measurement by having the necessary information already available for the prediction model to process quickly.
Data Source
AI summary
An online host for conducting a promotion campaign is presented. Online behaviors of subscribers exposed by the promotion campaign are tracked for determining conversion counts for the promotion campaign. For exposed subscribers whose online behaviors are not sufficiently available to the online host to determine conversion counts (non-measurable subscribers), a machine learning model is trained to predict conversion counts based on online behaviors that are conducted on internet locations under control of the online service, and further trained to determine an estimated error rate regarding the predicted conversion counts. Conversion counts for a promotion campaign are determined according to an analysis of the online behaviors of the exposed subscribers whose online behaviors are sufficiently available to the online host to determine conversion counts (measurable subscribers), and according to the predicted conversion counts of the non-measurable subscribers with an estimated error rate that meets or exceeds a predetermined threshold.