Payments are crucial for construction firms, yet contractors often encounter uncertainties regarding the timing and reliability of payments. This paper introduces a novel method for assessing the reliability of owner payments by analyzing historical payment behaviors, particularly in response to contractor claims.
Owners are categorized based on their payment behaviors using a payment profile that tracks response times to claims. A total of 364 potential payment behaviors are modeled. Both supervised (regression trees) and unsupervised (k-means clustering) machine learning algorithms are applied, utilizing four key variables: final payment percentage, average delay after a claim, average payment intervals, and maximum payment amount. Data from 45 Iranian and 14 Australian construction projects are analyzed, including a case study of a dam project in Iran.
Project owners are categorized based on their payment behaviors, revealing distinct differences between complete and incomplete payment groups. Owners who make complete payments experience an average delay of about one week, with 60–80% of payments settled within three weeks. In contrast, those with incomplete payments see 70% of payments completed within eight weeks. Approximately 50% of owners with complete payment behavior were classified as having favorable payment behavior, while around 29% of owners with incomplete payment behavior demonstrated behavioral stability. The K-means model classifies all owners, while the decision tree model predicts payment behaviors with 88% accuracy. Combining both models increases predictive accuracy to 96%. Consistent payment patterns were observed among projects managed by the same owners, underscoring the need for further research into external factors influencing payment behaviors.
Most owners with complete payments experienced a one-week delay, with a payment period averaging three weeks, and made payments ranging from 60% to 80% within a single payment period, establishing them as the most favorable owners compared to other groups. Conversely, owners with incomplete payments typically encountered delays of one week, had a payment period averaging two weeks, made payments of 30%–60% within one payment period, and reached a maximum total project payment of 70% within eight weeks after submitting the claim. The decision tree achieved an 88% accuracy rate in predicting owner payment behavior, while the K-means model successfully predicted all owners’ payment behavior. A recommended approach combines both models to enhance accuracy and reliability in predicting project owners’ payment behavior.
This paper offers valuable insights into owners’ payment behavior based on their historical patterns, aiding in risk-based decision-making during the tendering process and improving financial forecasting. Additionally, it allows owners to compare their payment behavior with that of other owners in the construction industry.
This paper introduces a novel approach, significantly contributing to contractor pre-tender risk assessments and enhancing the analysis of claim-payment processes.
