Estimation of potential lost sales in retail networks of high-value substitutable products

Sales data reveal only partial information about demand due to stockout-based substitutions and lost sales. We develop a data-driven algorithm to estimate stockout-based lost sales and product demands in a distribution network of high-value substitutable products such as cars, using only past sales and inventory log data and product substitution ratios. The model considers the particular customer and retailer behaviors frequently observed in high-value product markets, such as visiting multiple stores by customers for a better match and exploiting on-demand inventory transshipments by retailers to satisfy the demand for out-of-stock products. It identifies unavailable products for which a retailer could not fulfill demand and estimates the potential lost sales and the probability distribution of product demands for the potential lost sales using sales data in retailers with similar sales profiles while considering retailers’ market sizes. We validate the results of our algorithm through field data collection, simulation, and a pilot project for a case of recreational vehicles. We also show the result of implementing our model to estimate lost sales across the large retail network of a leading vehicle manufacturer. Our case study shows sales data significantly underestimate the demand for most products.