A new methodology for tracking the performance of subcontractors in the construction industry

Abstract

Forecasting the performance of subcontractors is fraught with difficulty. Any particular measurement will have both information about the subcontractors performance and also noise due to random effects. This thesis aims to present a methodology for separating the underlying performance from the noise in the measurements. A secondary aim is to examine how this affects the optimum time for updating historical records of subcontractors. A case study has been adopted to test this methodology using data collected from Saudi Arabia. Data was collected in two phases. The first phase involved interviewing experts in assessing subcontractors to explore the importance of tracking the performance of subcontractors, the most important performance factors, and the frequency of updating historical records. The second phase involved collecting data about subcontractors’ performance from historical records. The results of the interviews show that different organisations focus on different factors, but they have strong agreement that work quality and safety are important. They also have different frequencies for updating their historical records, ranging from 1 to 5 years. The performance questions were classified also in two groups using factor analysis: management questions and technical questions. The expected change over time in subcontractors’ performance was studied by using Markov chains. The noise content of these measurements was studied by comparing with hidden Markov models using the Baum Welch algorithm. A methodology was also provided that enables tracking of the loss of accuracy over time based on entropy. The results of the case study show that subcontractors improve over time in technical performance faster than in management performance. The updating time of the historical records based on the case study is recommended to be annually for technical questions and every two years for management questions. This research demonstrates that hidden Markov models provide a new strategy for forecasting subcontractors’ performance and reducing the effect of randomness to be annually for technical questions and every two years for management questions. This research demonstrates that hidden Markov models provide a new strategy for forecasting subcontractors’ performance and reducing the effect of randomness to increase accuracy. A limitation of this work is that it is based on a single case study.