Previous studies have reached consens about the economic value of customer retention, that (1) the cost to win a new customer is much greater than the cost to retain customer (Dawes and Swailes.1999); (2) long-term customers buy more, and if satisfied may bring new customers (Ganesh et al.2000); (3) losing customers not only leads to loss of revenue, but also increase the cost for attracting new customers (Athanassopoulos.2000, Colgate and Danaher.20000). Therefore, churn prediction is a well-known application of machine learning and data mining in Customer Relationship Management (CRM), which allows a company to predict the possibility of its customer churning. The difficulty of Churn Prediction is to predict precisely and timely, so that company may have enough time to keep their customers.
Previous studies have investigated churn mostly in B2C contexts, only a few predictive models have been developed in B2B context. Researches in B2B context focus mostly on non-contractual settings in sectors like retailing, e-commerce, logistics, and rarely on contractual settings such as software maintenance contract. Among them the highest accuracy based on the area under the receiver operating characteristic curve (AUC) is 92% (Tamaddoni Jahromi et al, 2014). Typical churn prediction models usually consider only the variables that reflect customer behaviour. The incorporation of macroeconomic data has been so far considered only in B2C context by few of studies (Gür Ali et al, 2014; Van den PoelE et al, 2004). These studies, however, do not consider the development of both customer behaviour and macroeconomic variables over time. In other words, the models lack the dynamic aspect.
The core task of this thesis is the development of a churn prediction system that provide a global view of customer churn in B2B context dynamically, which not only considers of customer behaviour and also consider of the impact of economics factors.
Data Innovation Community
Zhuonan Zhang, Karlsruhe Institut für Technologie, email@example.com
Nov. 2018 – Apr. 2019