IT projects normally face time and cost overruns challenges. Predicting which projects will not be completed by the expected end date (time overrun) and within the allotted number of hours (man-hours overrun) helps manage company’s employee utilization and avoids overscheduling. This study applied a data analytics framework on IT project management dataset to predict the project completion time and overruns (in man-hours). We used linear regression and classification models to predict projects’ performance and analyze the underlying factors causing project delay. Our predictive analysis used 131 variables which included 536 tasks, 138 resources, 69371 employee hours, 72 contractors that were assigned to 434 projects. We also calculated two new variables closeness and betweenness among project team members. Results showed that Decision Tree outperformed SVM, ANN, LDA, and logistic regression in predicting man-hours overrun. In addition, preliminary Social Network Analysis (SNA) indicates that Avg-closeness and Avg-betweenness did not improve prediction on the overall amount of time and man-hours overrun but improve the prediction on time overrun, but task, resource and contractor assignments variables were significant at p-value of .01. The models we used helps identify key predictors of project performance and provide insights into the company’s resource management.