A framework for combined bayesian analysis and optimization for services delivery

One of the key challenges facing the professional services delivery business is the issue of optimally balancing competing demands from multiple, concurrent engagements on a limited supply of skill resources. In this paper, we present a framework for combining causal Bayesian analysis and optimization to address this challenge. Our framework integrates the identification and modeling of the impact of various staffing factors on the delivery quality of individual engagements, and the optimization of the collective adjustments of these staffing factors, to maximize overall delivery quality for a pool of engagements. We describe a prototype system built using this framework and actual services delivery data from IBM's IT consulting business. System evaluation under realistic scenarios constructed using historical delivery records provides encouraging evidence that this framework can lead to significant delivery quality improvements. These initial results further open up exciting opportunities of additional future work in this area, including the integration of temporal relationships for causal learning and multi-period optimization to address more complex business scenarios.