In this paper, we present an application of machine learning models with real-world data in order to enhance a company's performance. First, we present a brief literature review about the new trends in the field of supply chain predictive analytics and about the models used during this case study. We present then the industrial data-driven models to improve the overall supply chain performances in the scope of the Sales and Operations Planning (S&OP) process. We will present in our work, the considered assumptions and the methodology developed in order to implement a sales prediction engine, through several time series forecasting models. In this work, we use and mix classic time series models like ARIMA and Exponential Smoothing, and the recent Prophet algorithm. Finally, we present the results of the implementation of our methodology and the noted improvements on the predictions accuracy. Our discussions derive on how predicting customer demand for many products from different natures and on how the model parameters affected our sales forecasting process.