Marketing spend optimization

Optimization methods that work with sequential even data represent a major step forward compared with traditional marketing mix and channel attribution models that work with aggregated data. In this article, we explore how deep learning methods can be used to analyze sequences of customer interactions and how the insights gained from such analyses can be used for spend optimization. We develop a line-up of attribution models ranging from basic last-touch attribution to LSTM with attention and evaluate them using a public dataset published by Criteo. In the last section of the article, we show how these models can be used to evaluate different scenarios for the allocation of marketing budgets.

In this tutorial, we discuss how traditional targeting and personalization models such as look-alike and collaborative filtering can be combined with reinforcement learning to optimize multi-step marketing action policies (aka Next Best Action policies). This approach can be applied to a wide range of marketing optimization problems, including promotion optimization, optimization of special offers, data-driven budget reallocation across channels, and so on.