Next Best Action : A clustering framework for Customer Relationship Management

Abstract: For companies that supply a service to their customers, there will always be a difference in engagement from customer to customer. Some customers might have created an account, they might also have a paid subscription, or they neither have an account nor are a subscriber. Meaning that they only interact with the free content provided. Looking at this from the perspective of the company, a more engaged customer will generate more value than one less engaged. Therefore, it would be beneficial to try and understand why some customers are more engaged than others. This project purpose is to create a customer relationship management (CRM) pipeline for determining what it is that make some users more engaged with a service than other. The findings will be presented to Bonnier that they may use to try and increase customer engagement across the board. This is achieved by clustering users based on their recency, frequency, and volume (RFV) data with the help of two clustering algorithms. The clustering algorithms are evaluated to detriment which is more suited as well as to determine the optimal number of clusters. Lastly the clusters are analysed on different metrics connected to the users as well as to the articles they read. The results show some implications into what makes some users more engaged. It’s also concluded that more clustering and analysis would be needed to get even more insight into what leads to more customer engagement.