In Engineer to order (ETO) business, on-time project execution is a complex endeavour. This complexity is multiplied by customized engineering drawings, delay in customer approvals, heavy document and change management and mainly due to the participation of multiple stakeholders. This introduces high level of uncertainty during execution and impacts total throughput time (TTPT) and on time delivery (OTD) often resulting in lower inventory turnover ratio (ITR). The major factors contributing to uncertainty comprise type and rating of products, components used, order type, customer segments, resources involved, suppliers, production lines, contract terms and conditions, etc. Moreover, these factors have different impact across different phases of order management from marketing till dispatch. In this study, a Machine Learning based model is used considering more than 40 variables, both numeric and textual, spanning all the manufacturing phases. As a first step, the key driver analysis determined the likelihood of the delay with certain sales channel, component type, product variant, complexity of job, number of panels, and sales & project engineer levels. This is followed by developing a Supervised Learning Algorithms which identified the clusters based on order specifications and associated propensity of delay in servicing the orders. The accuracy of the models varied from 73 to 77%. These preliminary findings are not only promising to establish confirmation on some of the intuitive findings, but these also help in initiating operation excellence projects on many other important but non-intuitive findings.