This repo contains the code I developed during the master’s thesis I worked on to finish the Master’s Degree in Statistics for Data Science at the University Carlos III of Madrid. The summary of the paper is presented below, and the full paper is available from Github.

Digital marketing has suffered an incredible growth in the last decade. Companies advertising their products target online users with specific ads and customized marketing campaigns. This has allowed them to increase the return of investment over traditional marketing campaigns, that include for instance TV or radio ads. The main benefit of digital marketing is that it can be tracked and measured through the use of cookies stored on the browsers of the users. Mathematical models have been developed with the aim of attributing the sales to the different marketing channels used to reach the users. In this work we present the models currently used by most of the advertisers, that are based on simple rules. We also present three data driven models that introduce more granularity, and that have been developed to improve the shortcomings of the simple rule based models. Finally, we apply the models to a dataset and compare the results, with the aim of understanding and characterizing the behavior and properties of the different models.