Optimización de contactos telefónicos efectivos en gestión de cobranzas mediante un modelo de mejor horario de llamada, usando regresión multinomial

Call Centers represent worldwide a consolidated industry and one of its activities is the management of collections. The present work proposes a predictive statistical model to increase the probability of phone contactability in the collection management through the best call schedule. This leads directly to consider more than two possibilities, that is, we are faced with a multicategorical response problem, so a multinomial model is specified. The cross-sectional data used in the empirical analysis comes from a large-scale collection company located in Ecuador. The individuals, object of this analysis, are borrowers who were in arrears in products of consumer credit and microcredit. The study includes the analysis of approximately 6000 individuals and the treatment of 139 explanatory variables collected between January and September 2016. The results suggest that historical contact information, day of the week, characteristics of debtors in arrears and propensity to pay (given by the ratio between arrears in short-term and in long-term) are determinants of an effective contact by phone.