Data Driven Insights into Building Project Performance and Outcomes Through Advanced Data Analytics

Construction management is a highly competitive project-based field of complex specialized services creating or altering the built environment for a client. For construction projects to be successful and in turn construction firms to be successful, understanding the relationship of performance statistics as indicators of project outcomes such as cost, time, and profitability are essential. There have been a number of efforts to identify key performance indicators related to construction project success; however, due to lack of available data many questions remain and no clear means for evaluation is evident. Analyzing project statistics as indicators of project success similar to the way analytics have been used in sports to predict success is an opportunity for analysis that could prove promising. Construction firm project data for a portfolio of building projects was analyzed using three different methods for this study. The first was identifying correlated factors for completed building construction projects. The second method was a regression analysis to find the largest contributing factors to profitability in the portfolio. The third analysis involved leveraging Machine Learning (ML) in the form of Extreme Gradient Boosting (XGBoost) to analyze the data to detect patterns and develop a model to predict project success for future projects from incomplete information based on the data from a portfolio of completed projects. A highlight of the correlation analysis identified profit differential as demonstrating a strong relationship with the number of Requests for Information and Architects Supplemental Instructions on a project. A highlight of the regression analysis found that the number of Requests for Information and Architects Supplemental Instructions accounted for approximately 82% of the variance of actual profit within this portfolio. While in a ML multivariate analysis through XGBoost found that Budgeted Profit contributed to 70% of the variance in Actual Profit. This study highlights the transformative potential of ML in finding the influence of complex factor interactions that may not be present in univariate analysis for the construction sector and the use of XGBoost in emphasizing the practical application of findings to a portfolio of building construction projects among individual construction companies.