Unraveling Consumer Purchase Journey Using Neural Network Models

Abstract

This study utilizes an ensemble of feedforward neural network models to analyze large-volume and high-dimensional consumer touchpoints and their impact on purchase decisions. When applied to a proprietary dataset of consumer touchpoints and purchases from a global software service provider, the proposed approach demonstrates better predictive accuracy than both traditional models, such as logistic regression, naive Bayes, and k-nearest neighbors, as well as ensemble tree-based classifiers, such as bagging, random forest, AdaBoost, and gradient boosting. By calculating the Shapley values within this network, we provide nuanced insights into touchpoint effectiveness, as we not only assess the marginal impact of diverse touchpoint types but also offer a granular view of the impact distribution within a touchpoint type. Additionally, our model shows excellent adaptability and resilience with limited data resources. When the historical data is reduced from 40 to 1 month, our model shows only a modest 19% decrease in accuracy. This modeling framework can enable managers to more accurately and comprehensively evaluate consumer touchpoints, thereby enhancing the effectiveness and efficiency of their marketing campaigns.

Keywords: Consumer purchase journey, Multi-channel marketing, Neural networks, Shapley value, Short lookback window, Deep learning