Customer segmentation is dividing of customers into groups of individuals thatrnhave common characteristics or traits. By segmenting customers based on theirrnusage behavior, telecom companies can better target and classify their customers,rnprovide the services that meet their expectations and increase proï¬tability. On therncontrary, companies with improper segmentation or luck of segmentation facingrnthe problem of providing the exact product or service to meet the actual customerrnneeds. Incorrect proï¬t prediction and wastage of resource utilization are the mainrnproblems of ethio telecom which results from poor customer segmentation.rnTo mitigate the segmentation problem this study focuses on segmenting telecomrncustomers based on their usage behavior using unsupervised clustering techniques.rnK-means algorithm was used to cluster the Call Detail Record (CDR) data.rnBefore clustering CDR data were collected, relevant attributes selected and preprocessingrntechniques such as data cleaning, data aggregation, data integration,rnand data formatting were performed. In addition, four datasets were formed byrnsummarizing the data on a monthly base.rnThe experimentation results in eight different clusters. These clusters were analyzedrnusing quantile score techniques. The clusters were ranked and mapped withrncustomer segmentation type. Among the clusters, the cluster with 236rnsubscribersrnwas scored the highest in terms of duration, frequency and money. As a result,rnthis cluster was chosen as a platinum customer type. They are highly proï¬tablerncustomers, vital to affect its revenue and need to serve well by the company.