Beyond one-size-fits-all: A novel data-driven framework for quantifying bidding competition intensity in highway design-bid-build contracts

Maintaining adequate competition during the letting phase of design-bid-build highway projects is essential for optimal outcomes. However, contractors' bidding patterns vary with contract attributes, such as project location, size, and work-type, which in turn affect the expected degree of competition. The Federal Highway Administration also encourages state DOTs to consider these attributes during the bid review process. However, most DOTs overlook specific contract attributes or rely solely on a single attribute, such as work-type, when measuring competition adequacy. Recent studies have proposed network-theory-based approaches to model competition using win-loss frequencies; nonetheless, these approaches have overlooked contract attributes. This study enhances the existing body of knowledge by establishing thresholds for competition intensity based on variations in bidding patterns resulting from the collective effects of contract attributes that are essential in evaluating bidding competition. To this end, a two-layer bipartite multiplex graph is proposed, comprising sets of bidder and contract nodes, with contract attributes stored in the contract nodes. Edges between nodes and bidders are weighted by the bid-to-engineers' estimate ratio to represent the bidding patterns influenced by contract attributes. A Graph Neural Network is trained to learn these bidding patterns, and the output contract vectors are clustered to represent similar bidding patterns. The proposed approach was validated using historical bid records from the South Dakota DOT, revealing that contract clusters derived from the model substantially outperform the work-type groups in capturing similar bidding patterns. For each obtained cluster, competition intensity thresholds for low, normal, and high competition are established based on proposed centrality measures that demonstrate stronger correlations with bidding performance than previous network-theory models. The resulting competition thresholds are expected to enable DOTs to make more accurate, data-driven evaluations of competition adequacy tailored to specific contract characteristics, improving bid analysis and award decisions.