Advancing Customer Journeys Through The Identification of The Next Best Channel In Digital Marketing

Understanding how the customer interacts with the digital communication over time is primary to achieving engagement and conversion. Much research discusses how to design a perfect customer journey, but few identify and integrate the best time and next best channel, as it may change over time. This paper proposes an initial exploration of the topic that synthetically connects to find the next best channel to maximise the client journey, optimize time, engagement, and consequently lead to conversion. For this purpose, algorithms for handling sequential data are explored, such as Long Short-Term Memory (LSTM), and classification algorithms, such as RF, XGB, XT, LightGBM, and CAT, are employed. A study of the influence of the campaign’s topics is made, and their influence is evaluated in predicting the next best channel. Au-toGluon for hyperparameter tuning was explored. The results show a great agreement, with prediction accuracy of 0.76 (without campaign information) and 0.94 (with campaign information added) for the ensemble model and 0.9764 for LSTM. The word "summer" and "sale" leads to more conversions, while "new" leads to more opens and clicks. Future work will explore reinforcement learning for self-improvement in the model.