Incrementality Models for Multi-Exposure Action Attribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
It is difficult to determine the relative contributions of various online content exposures to a user's subsequent performance of a specified target action, particularly when these exposures occur at different times and involve both organic and digital components.
Innovation Solution
A model is created to quantify the incremental effects of content exposures on user actions by distinguishing between organic and digital components, using ablation experiments and decay functions to attribute performance levels, enabling the modification of transmission criteria for digital components based on these effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used to evaluate the influence of previous online activities on subsequent actions, then the precision of measuring user behavior impact is improved, but the complexity of data processing and model creation increases
Solution Approach 1:
The patent segments the evaluation process into distinct components: creating a baseline model representing user behavior without digital component exposures, and then creating a second model that includes digital component exposures. This segmentation allows the system to isolate and measure the specific incremental impact of digital components while managing complexity through modular model creation and comparison.
2Measurement precision
If the system tracks and analyzes all online content exposures to determine their contribution to target actions, then the accuracy of attribution is improved, but the computational resources and time required increase
Solution Approach 1:
The patent applies preliminary action by creating a baseline model in advance that captures user behavior patterns without digital component exposures. This pre-computed baseline serves as a reference point that eliminates the need to re-analyze all historical data each time an attribution evaluation is needed, significantly reducing computational time while maintaining attribution accuracy through model comparison.
3Measurement precision
If the model distinguishes between organic exposures and digital component exposures, then the precision of incremental effect measurement is improved, but the difficulty of detecting and measuring increases
Solution Approach 1:
The patent extracts the incremental effect of digital component exposures by removing the baseline behavior (organic exposures only) from the total behavior (including digital components). This extraction is achieved through model comparison, where the difference between the second model (with digital components) and the baseline model (without digital components) isolates the specific contribution of digital components, thereby precisely measuring their incremental effect while simplifying the measurement process.
Data Source
AI summary
Methods, systems, and computer programs encoded on a computer storage medium, for training and using machine learning models are disclosed. Methods include creating a model that represents relationships between user attributes, content exposures, and performance levels for a target action using organic exposure data specifying one or more organic exposures experienced by a particular user over a specified time prior to performance of a target action by the particular user and third party exposure data specifying third party exposures of a specified type of digital component to the particular user over the specified time period. Using the model, an incremental performance level attributable to each of the third party exposures at an action time when the target action was performed by the particular user is determined. Transmission criteria for at least some digital components to which the particular user was exposed are modified based on the incremental performance.