Web user profile

Customer Profiling allows a company to gain knowledge regarding their customers, by providing a better understanding of their needs, expectations, and preferences. This process helps the organization in better targeting customers for marketing actions and, consequently, to gain customers’ loyalty on increasing their engagement with the company. In the telecommunications industry, customer profiling has been an important topic of discussion, particularly because it is a very useful tool when it comes to increasing an organization’s competitive advantage by improving its marketing strategy, or by increasing its customer retention. This project, developed within the scope of a telecommunication company aims to develop a customer profile, by the extraction and analysis of customers’ online navigation data on the company’s website, and to build a model capable to predict conversions, this is, to predict which customers are more likely to generate revenue within a certain period of time. The user profile was created through the extraction and analysis of data available in the digital channel of the company, Google Analytics. In the end, a web user profile with 249 variables was achieved, where 239 features were built from the digital data, being the remaining 10 pre-built variables that were already used by the Communication Service Provider (CSP) to characterize customers. Regarding the development of the predictive model, three different classifiers were tested (Logistic Regression, LightGBM, and XGBoost), where it was possible to conclude that XGBoost was the one that best performed on predicting conversions while dealing with an imbalanced target class.