Abstract: The present disclosure relates to a method and a prediction device for predicting timely delivery of telecom service orders in real time. In one embodiment, the method receives order data of historical time period and processes the order data to derive one or more variables and add missing values in the order data. Based on the processed order data, one or more models are generated and a model having least model generation error rate is identified. Using the model thus identified, prediction of timely delivery of telecom services is predicted in real time using real time data. By way of identifying factors that influence the timely delivery in each stage, helps to improve the probability of timely delivery by correcting the identified factors, thus improving customer experience and revenue realization to the telecom service providers.