Customer Churn Detection and Marketing Retention Strategies in the Online Food Delivery Business

Abstract

The purpose of this thesis is to analyze the behavior of customers within the Online Food Delivery industry, through which it is proposed to develop a prediction model that allows detecting, based on valuable active customers, those who will leave the services of Alpha Corporation in the near future. Firstly, valuable customers are defined as those consumers who have made at least 8 orders in the last 12 months. In this way, considering the historical behavior of said users, as well as applying Feature Engineering techniques, a first approach is proposed based on the implementation of a Random Forest algorithm and, later, a boosting algorithm: XGBoost. Once the performance of each of the models developed is analyzed, and potential churners are identified, different marketing suggestions are proposed in order to retain said customers. Retention strategies will be based on how Alpha Corporation works, as well as on the output of the predictive model. Other development alternatives will also be discussed: a clustering model based on potential churners or an unstructured data model to analyze the emotions of those users according to the NPS surveys. The aim of these proposals is to complement the prediction to design more specific retention marketing strategies.