Abstract
Product stock-outs and the resultant purchase of substitutes by customer is common in retail. These stock-out lead to censored sales data; the observed demand for products with stock-outs could be lower that the true demand for the same while the observed demand for the substitutes could be inflated. If historical sales data is used to forecast future demand for a product without accounting for stock-outs, it could lead to errors in demand forecasting on account of misspecification. In this paper, we demonstrate the need and the benefits of data unconstraining and develop a data unconstraining method to address the dual issues of estimation of demand and substitution rates when there are unobserved substitutions among choice alternatives. We compare the effectiveness of the proposed technique by comparing it against the popular techniques available in extant literature.
