Buffer sizing in critical chain project management by network decomposition

Abstract

Project management organizes about 30% of the world’s economy. Many recent projects apply critical chain project management (CCPM) methodology, which requires the design of project and feeding buffers. Accurate sizing of these buffers is essential, because too small buffers result in emergency procedures to prevent late delivery, whereas too large buffers result in uncompetitive bids and lost contracts. Previous buffer sizing research, focused predominantly on the critical chain, typically results in excessive buffers, and in critical chains being challenged by feeding buffers during planning. This work also performs inconsistently, for example in makespan estimation, at execution. We propose a new procedure for buffer sizing based on network decomposition, which offers logical advantages over previous ones. First, the size of a feeding buffer is determined from all associated noncritical chains. Second, the project buffer incorporates safety margins outside the critical chain by comparing feeding chains with their parallel critical counterparts. Computational testing on a case study of a real project and extensive simulated data shows that our procedure delivers much greater accuracy in estimating project makespan, and smaller feeding buffers. Furthermore, the resulting critical chain is never challenged. Additional benefits include delayed expenditure, and reductions in work-in-process, rework, and multitasking.