Implementación de un modelo de propensión para el cambio de línea prepago a postpago en una empresa de telecomunicaciones
Loading...
Código QR
Authors
Aroni Rios, Alfredo Junior
Contact Email
Abstract
Las empresas de telefonía móvil ubicadas en el Perú cuentan con el gran reto de lograr migrar a sus clientes con un plan prepago a postpago. Para ello utilizan canales como el call center quienes contactan a los clientes prepago y se les ofrece la migración de su plan a postpago. Debido a esto, este trabajo busca implementar un modelo de propensión que mejore la efectividad de la base seleccionada que se envía al call center para la campaña de migraciones mediante la identificación de la probabilidad a migrar. Se utilizó el algoritmo LightGBM para calcular la propensión, así mismo, se segmentó en deciles para que el negocio lo pueda utilizar como una alternativa a la selección tradicional donde solo se usan tres variables: recargas, tráfico de datos y tráfico de minutos. El modelo logro alcanzar un AUC de 0.68 en un nuevo mes, llamado periodo de validación. Así mismo, se logró probar el modelo durante 6 meses en un porcentaje de la base enviada al call center. Dando como resultados que el grupo seleccionado por el modelo obtenga una mayor tasa de conversión respecto a la selección tradicional. El incremento de la tasa de conversión de migraciones alcanza mas de 0.9% en promedio de manera mensual. Por lo que el modelo propuesto optimiza y simplifica la selección tradicional de la base enviada al call center para las migraciones a postpago. Además, se valida que contribuye a la mejora de la tasa de conversión y por ende a la cantidad de migraciones logradas por dicho canal.
Mobile phone companies located in Peru face the major challenge of migrating their customers from prepaid to postpaid plans. To do so, they use channels such as call centers, which contact prepaid customers and offer them the option of switching to a postpaid plan. This study therefore seeks to implement a propensity model that improves the effectiveness of the selected database sent to the call center for the migration campaign by identifying the probability of migration. The LightGBM algorithm was used to calculate propensity, which was then segmented into deciles so that the business could use it as an alternative to traditional selection, which only uses three variables: top-ups, data traffic, and minute traffic. The model achieved an AUC of 0.68 in a new month, called the validation period. Likewise, the model was tested for six months on a percentage of the database sent to the call center. The results showed that the group selected by the model obtained a higher conversion rate than the traditional selection. The increase in the migration conversion rate reaches more than 0.9% on average per month. Therefore, the proposed model optimizes and simplifies the traditional selection of the database sent to the call center for postpaid migrations. In addition, it has been validated that it contributes to improving the conversion rate and, therefore, the number of migrations achieved through this channel.
Mobile phone companies located in Peru face the major challenge of migrating their customers from prepaid to postpaid plans. To do so, they use channels such as call centers, which contact prepaid customers and offer them the option of switching to a postpaid plan. This study therefore seeks to implement a propensity model that improves the effectiveness of the selected database sent to the call center for the migration campaign by identifying the probability of migration. The LightGBM algorithm was used to calculate propensity, which was then segmented into deciles so that the business could use it as an alternative to traditional selection, which only uses three variables: top-ups, data traffic, and minute traffic. The model achieved an AUC of 0.68 in a new month, called the validation period. Likewise, the model was tested for six months on a percentage of the database sent to the call center. The results showed that the group selected by the model obtained a higher conversion rate than the traditional selection. The increase in the migration conversion rate reaches more than 0.9% on average per month. Therefore, the proposed model optimizes and simplifies the traditional selection of the database sent to the call center for postpaid migrations. In addition, it has been validated that it contributes to improving the conversion rate and, therefore, the number of migrations achieved through this channel.
Description
Universidad Nacional Agraria La Molina. Facultad de Economía y Planificación.
Departamento Académico de Estadística e Informática
Keywords
LightGBM; Modelización; Telecomunicaciones; Comportamiento; Consumidores; Informática
Citation
Date
2025
Collections
Seleccionar año de consulta:
Licencia de uso

Excepto si se señala otra cosa, la licencia del ítem se describe como info:eu-repo/semantics/openAccess

