Improving Application Management Services through Optimal Clustering of Service Requests

Abstract

In the area of Application Management Services (AMS), good resource planning, efficient workload assignment, and effective skill planning are critical to success. Meeting these objectives would require systematic and repeatable approaches for determining the best way of forming resource pools, assigning the right service requests to the right people, and identifying who to train for what skills under a constrained budget. In this paper, we present a methodology developed for the Global Business Services (GBS) organization of IBM to help achieve the above goals. Specifically, given a collection of service request records, we propose to group service requests that require similar problem-solving skills, into a single cluster using a statistical clustering technique. Such clusters are then associated with service consultants along with their respective service handling experiences and confidence levels. Using real GBS account data, we conducted a queuing-based simulation which has shown that, by applying the resource sharing plan recommended by our clustering analysis, we are able to achieve an average 40% resource reduction for both within- and across-geography situations, while maintaining the same Service-Level Agreements (SLA) with the customer.