Building a predictive lead scoring model for contact prioritization : the case of HUUB

• Autor(a) principal: Pereira, Rita Mafalda Magalhães
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.14/34877
• topic: Marketing automation; Lead scoring; Machine learning; Contact prioritization; Automatização do marketing; Priorização de contactos; Domínio/Área Científica::Ciências Sociais::Economia e Gestão

In the last decades, machine learning has become quite popular for solving business problems, as it often delivers high-quality and efficient solutions. Moreover, the amount of data collected by companies has grown substantially, which has contributed to this trend. Companies do not have enough resources to contact every lead, so contact prioritization is essential. Lead scoring supports this task, by assigning a value to each lead based on his actions or characteristics. Even though it is expected that lead scoring contributes to higher conversion rates, there is still very few literature on how to use machine learning to automate this process. This dissertation shows how to combine historical data from Customer Relationship Management platforms and supervised learning to develop a lead scoring model for companies. The approach followed is based on the CRISP-DM method, where several tools were used, such as HubSpot, Microsoft Power BI and RStudio. The classification model proposed is a decision tree that predicts the leads’ conversion outcome (Won or Postpone), developed using the CART algorithm and data from a logistics company – HUUB. The main findings of this project conclude that machine learning can be used to develop a lead scoring model to perform contact prioritization. However, there are several factors, especially data-related, that should be taken into consideration, since they may impact the model’s performance. Lastly, a suggestion for future research is to develop an experiment to compare the results of manual and automated lead scoring, to assess if machine learning actually provides a superior alternative to the manual approach.