FRAMEWORK FOR PREDICTIVE AND CAUSAL MODELING OF MULTI-TOUCH ATTRIBUTION IN A DIGITAL ENVIRONMENT
Authors
Keywords
multi-touch attribution, causal inference, customer journey analysis, predictive modeling, marketing optimization
Summary
In the increasingly complex ecosystem of digital marketing, determining the incremental contribution of different marketing channels to consumer conversion remains a fundamentally unresolved challenge. The subject of this study is the process of multi-touch attribution in a digital environment, and its object is the predictive and causal approach to assessing the contribution of channels. The main goal is to develop a comprehensive framework for multi-touch attribution that integrates probabilistic modeling (Markov chains), machine learning (XGBoost), and causal methods (Shapley values, counterfactual analysis) to provide a more accurate and interpretable assessment of channel influence. As part of the study, a hybrid model system was built and tested on real user journey data, and its effectiveness was compared to that of traditional heuristic models. The results show significant differences from heuristic approaches and highlight the superior predictive accuracy of machine learning-based methods (overall accuracy = 0.984; AUC = 0.846), while the application of explainable AI techniques increases the transparency of attribution. Theoretically, the study combines causal models, reinforcement learning, and interpretability in attribution modeling, linking econometric theory with data analysis, and in management terms, it offers a reproducible framework for resource optimization and data-driven decision-making.
Pages: 1
Price: 3 Points
