Customer Segmentation for Value and Retention Using Data Mining: In the Case of Ethio-Telecom Mobile Service
No Thumbnail Available
Date
2021-09
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Market competition is becoming intense among telecom service providers around the world. To gain a competitive advantage in the industry, service providers must address and meet their customers' needs and demands. To survive in today's competitive market, companies must analyze and interpret their customers' usage behavior, as well as plan related market strategies to retain customers, be profitable, and build long-term relationships.
One of the most effective ways to engage with each customer is through segmentation. Customer segmentation can help organizations identify more effective marketing strategies for each segment by leveraging the power of data mining clustering technology.
In this paper, the unsupervised clustering technique with k-means algorithm is applied and customer detail record (CDR) and customer information data are used to segment Ethio-telecom customers for the purpose of retain existing customers and increase customer value by treating each customer segment according to their usage behavior.
The collected data is cleaned and preprocessed in an Oracle database and then the aggregated data is used to calculate the optimal cluster number using the k-means and elbow methods. Based on the selected attributes the dataset is segmented into five groups by Weka knowledge discovery tool.
Each cluster segment is scored and mapped with the type of customer segmentation based on three-month average usage data, frequency, longevity, and service interruption time. Clusters 2 and 4 account for 5% of the total customer size but cover 72% of the company revenue, whereas Clusters 1 and 5 account for 76% of the total customer size but contribute significantly less than the others. Finally, based on the analysis result, a marketing strategy for each segment is proposed.
Description
Keywords
Data Mining, Clustering, CDR, elbow method, k-means