Can social media trends improve demand forecast ?

Abstract

Fast fashion relies on a firm’s ability to predict fashion trends in the immediate future. Fashion trends are highly volatile and inaccurate forecasts result in dead inventory or lost sales. Fashion trends are reflected in social media trends. We mine social media posts pertaining to product attributes and categories of our interest. For some categories and attributes we discovered that number social media posts that mention a product attribute or category are co-integrated with the sales of those products. Inclusion of number of social media posts as an exogenous variable in demand forecasting models, improved their accuracy by a margin of 40%, thus leading to better downstream inventory planning.