Customer Acquisition Process Digitalization: A Case Study on the Use of Machine Learning in The Corporate Insurance Industry

Abstract: This thesis explores the application of machine learning 8ml9 techniques in customer classification and their intergration into customer relationship management (CRM) systems within the corporate insurance industry. The research aims to address the gap in the use of AI-CRM for the corporate insurance industry. It was conducted as a case study at a Swedish insurance broker company. The study leveraged external data sources to create a data seet on customer information. The feature selection process included Variance Influence Factor (VIF) to remove collinearity and then Mutual Class Info and Random Forest, which are methods used to find which independent variables affect the dependent variable the most. Also, Recursive Feature Testing was applied to find the best feature combinations. Four different binary classification models were implemented and compared - Decision Tree, Random Forest, Support Vector Machine, and Artificial Neural Network. Note that Random Forest can be used both for feature selection and classification. The models were tested on four different feature combinations and evaluated using Accuracy, Recall, Precision, F1-score, and ROC-AUC. The study further conducted interviews at the partner company to evaluate their current CRM system. The findings show that ML-based customer classification can be leveraged to effectivize the customer acquisition process for corporate insurance. The Support Vector Machine model achieved the highest accuracy, at 80%. Depending on the avaliable data and the use of metrics, different classifiers had the best performance. The study also found that when implementing classification into AI-CRM, the specific requirements at the company need to be examined. This study found it important to conersider the data procurement process, the current customer acquisition process, the risks associated with misclassification, and present bias. The findings of this study have theoretical implications for the implementation of AI-CRM for customer acqusition. It demonstrates the practical benefits of intergrating machine learning techniques into CRM systems, emphasizing the effectiveness of AI-CRM for customer classification. Further, by comparing different classification models and evaluating their performance, the study enhances the theoretical understanding of model selection for customer classification tasks in this specific domain. Additionally, the research provides insights into effective feature selection methods, aiding researchers and practitioners in extracting relevant variables for customer classification.