A proposed algorithm for identifying heuristic-optimal bundling strategies

The practice of project bundling (grouping multiple projects into a contract) has been demonstrated to yield significant cost savings compared to multiple single-project contracts, and highway agencies seek guidance regarding the feasibility and effectiveness of this practice. However, there is a lacuna regarding formal computational bundling policy frameworks that inform the design of project bundles that yield minimal costs and risks. This paper formulates project bundling as a combinatorial optimization problem and solves it using a greedy heuristic algorithm. The framework incorporates analytical models developed in earlier studies to quantify pay-item similarity and assess contractual performance consequences. Using real-world contracts awarded over an 11-year period, numerical experiments show that the proposed heuristic algorithm can outperform three existing state-of-practice bundling policies.