Discrete-Time Survival Modeling for Media Attribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional communication systems fail to accurately attribute the influence of each media channel on conversions, as existing methods ignore time-decaying effects, interactions between channels, and employ non-adaptable rules, leading to inaccurate attribution results.
Innovation Solution
The implementation of a discrete-time survival modeling-based system that employs an algorithmic attribution model to determine the influence of each interaction and identify the most effective media channels for conversions by discretizing event histories, generating training observations, and training a logistic regression model to assign weights to interactions based on their lag increments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional attribution methods (first touch, last touch, equal linear weight) are used, then the system is simple to implement, but the attribution accuracy deteriorates because these methods ignore time-decaying effects and interactions between channels
Solution Approach 1:
The patent transforms the attribution problem by changing the parameter representation from simple weight assignments to time-dependent survival probabilities. The discrete-time survival model uses lag increments as parameters to capture time-decaying effects, where the probability of conversion decreases as the time lag between media exposure and conversion increases. This parameter transformation enables accurate attribution while maintaining computational feasibility through standardized statistical methods.
Solution Approach 2:
The patent segments the attribution analysis into discrete time intervals (lag increments) rather than treating all conversions equally regardless of timing. By dividing the attribution window into sequential time periods and applying survival modeling to each segment, the system captures the temporal decay of media influence. This segmentation allows the model to handle complex time-dependent relationships while using established statistical techniques for each segment.
2Adaptability or versatility
If predetermined and non-adaptable rules are employed for attribution, then the system is easy to operate, but the adaptability to different media channels and user behaviors deteriorates
Solution Approach 1:
The discrete-time survival model operates autonomously by automatically processing user event histories and media channel data to generate attribution scores. The model self-adjusts to different media channels and user behaviors by fitting survival curves to the observed data without requiring manual configuration for each channel. This self-service capability enables high adaptability while maintaining ease of operation through automated model training and evaluation.
3Loss of information
If conventional systems list all media channels used, then the system provides complete information, but the ability to identify the true influence and effectiveness of each channel deteriorates
Solution Approach 1:
The patent introduces survival probability as an intermediary metric that mediates between the raw media channel exposure data and the final attribution scores. Instead of directly attributing conversions to media channels, the model uses survival analysis to quantify the probability that a conversion would occur at different time lags, thereby revealing the true influence of each channel. This intermediary approach preserves complete information about all media interactions while precisely measuring their individual contributions through time-decayed probability weights.
Data Source
AI summary
The present disclosure relates to a media attribution system that improves multi-channel media attribution by employing discrete-time survival modeling. In particular, the media attribution system uses event data (e.g., interactions and conversions) to generate positive and negative conversion paths, which the media attribution system uses to train an algorithmic attribution model. The media attribution system also uses the trained algorithmic attribution model to determine attribution scores for each interaction used in the conversion paths. Generally, the attribution score for an interaction indicates the effect the interaction has in influencing a user toward conversion.