Conditional Intensity Attribution Model for Marketing Touchpoints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing attribution models, such as those using logistic link functions, fail to capture causal relationships between marketing touchpoints, leading to underestimation of credit for earlier touchpoints that drive subsequent conversions, and are computationally inefficient in redistributing attribution credit.
Innovation Solution
A causal-based attribution model using a conditional intensity model with baseline and causal parameters, trained with hyperparameters like sigma, to redistribute attribution credit back to earlier touchpoints that cause later events, capturing long-term interactions and time-decaying relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional attribution models (e.g., logistic link functions) are used, then computational simplicity is maintained, but causal relationships between marketing touchpoints are not captured and attribution credit is underestimated for earlier touchpoints
Solution Approach 1:
The patent transforms the attribution model by changing the functional form from logistic link functions to conditional intensity functions. This parameter change enables the model to capture causal relationships through time-decaying kernels and baseline intensity parameters, significantly improving attribution accuracy for earlier touchpoints while maintaining computational tractability through the structured functional form
Solution Approach 2:
The patent replaces the traditional statistical modeling approach (logistic regression with link functions) with a point process-based conditional intensity model. This substitution introduces causal reasoning capabilities through the intensity function formulation, allowing the model to propagate attribution credit backward in time based on causal relationships between events
2Measurement precision
If causal-based attribution models are implemented to capture long-term interactions, then attribution accuracy for earlier touchpoints improves, but computational efficiency decreases
Solution Approach 1:
The patent segments the attribution computation into distinct components: baseline intensity estimation, conditional intensity calculation for each event, and credit propagation through causal relationships. This segmentation allows efficient computation by processing events in sequence and reusing intermediate calculations, making long-term interaction modeling computationally feasible
Solution Approach 2:
The patent performs preliminary estimation of baseline intensity parameters and causal relationship structures before conducting full attribution analysis. This preliminary action pre-computes time-decaying kernels and causal graphs, significantly reducing the computational burden during actual attribution credit distribution across long event sequences
3Reliability
If hyperparameters (e.g., sigma) are determined through extensive training, then model fit improves, but training time and computational resources increase
Solution Approach 1:
The patent employs a pragmatic approach to hyperparameter selection by using reasonable default values for parameters like sigma (time decay rate) that can be set without extensive training. This disposable approach accepts approximate model fit in exchange for dramatically reduced training time and computational resources, suitable for many practical marketing attribution scenarios where perfect fit is not critical
Data Source
AI summary
Methods and systems are provided for facilitating generation and utilization of causal-based models. In embodiments described herein, a set of events comprising touchpoints resulting in a conversion are obtained. A direct attribution indicating credit for an event contribution to the conversion is determined. An adjusted attribution for the event based on the direct attribution for the event augmented with an indirect attribution for the event is determined. The indirect attribution can be identified based on the event causing a subsequent event of the set of events to result in the conversion. Thereafter, the adjusted attribution for the event is provided to indicate an extent of credit assigned to the event for causing the corresponding conversion.