Data-driven agent-based exploration of customer behavior

Authored by David Bell, Chidozie Mgbemena

Date Published: 2018

DOI: 10.1177/0037549717743106

Sponsors: No sponsors listed

Platforms: No platforms listed

Model Documentation: Other Narrative Flow charts

Model Code URLs: Model code not found

Abstract

Customer retention is a critical concern for mobile network operators because of the increasing competition in the mobile services sector. Such unease has driven companies to exploit data as an avenue to better understand changing customer behavior. Data-mining techniques such as clustering and classification have been widely adopted in the mobile services sector to better understand customer retention. However, the effectiveness of these techniques is debatable due to the constant change and increasing complexity of the mobile market itself. This design study proposes an application of agent-based modeling and simulation (ABMS) as a novel approach to understanding customer behavior through the combination of market and social factors that emerge from data. External forces at play and possible company interventions can then be added to data-derived models. A dataset provided by a mobile network operator is utilized to automate decision-tree analysis and subsequent building of agent-based models. Popular churn modeling techniques were adopted in order to automate the development of models, from decision trees, and subsequently explore possible customer churn scenarios. ABMS is used to understand the behavior of customers and detect reasons why customers churned or stayed with their respective mobile network operators. A CART decision-tree method is presented that identifies agents, selects important attributes, and uncovers customer behavior - easily identifying tenure, location, and choice of mobile devices as determinants for the churn-or-stay decision. Word of mouth between customers is also explored as a possible influence factor. Importantly, methods for automating data-driven agent-based simulation model generation will support faster exploration and experimentation - including with those determinants from a wider market or social context.

Tags

Simulation Agent-based modeling models Market Social Network Analysis Decision-Making Satisfaction Retention Decision trees Customer behavior Information-systems research Design science research Churn prediction