Forecasting sales is a key factor for efficient operational decisions in replenishment planning and inventory control in almost all supply chains. For demand forecasting, the model introduced by Bass (1969) has been used in several studies with different type of products in literature. In this study, we apply Bass diffusion model on a large-scale real-world sales dataset from the fashion industry. In estimating the parameters of the model, we use three common parameter estimation methods from the literature and introduce a new method called randomized line search. We analyze the forecasting performance of these methods in estimating the demand in a sale season using the entire dataset involving several different product categories. Based on the mean absolute percent error criterion, we check how good Bass diffusion model works and the performance of different methods for finding the parameters of the model, across 409 different products using a real data set from an apparel retail chain with 150 stores.