Predicting product sales in fashion retailing: a data analytics approach

Abstract

In the retail context, an erroneous determination of the amounts to buy of each article from the suppliers, either by excess or defect, can result in unnecessary costs of storage or lost sales, respectively. Both situations should be avoided by companies, which promotes the need to determine purchase quantities efficiently. Currently companies collect huge amounts of data referring to their sales and products' features. In the past, that information was seldom analyzed and integrated in the decision making process. However, the increase of the information processing capacity has promoted the use of data analytics as a means to obtain knowledge and support decision makers in achieving better business outcomes. Therefore, the development of models which use the different factors which influences sales and produces precise predictions of future sales represents a very promising strategy. The results obtained could be very valuable to the companies, as they enable companies to align the amount to buy from the suppliers with the potential sales. This project aims at exploring the use of data mining techniques to optimize the amounts to buy of each product sold by a fashion retail company. The project results in the development of a model that uses past sales data of the products with similar characteristics to predict the quantity the company will potentially sell from the new products. The project will use as a case study a Portuguese fashion retail company. To validate the model it will be used several linear regression measures to quantify model quality.