A mathematical model for pricing perishable goods for quick-commerce applications

Pricing of goods, especially for perishable items is a challenge for the rapidly growing quick-commerce (q-commerce) sector. The relatively short shelf life coupled with large sales volume makes correct pricing challenging. If the seller over-estimates the selling price, goods remain unsold and the seller has to bear a cost for disposal. On the other hand if the seller sets the price too low, it risks losing out on potential revenue. This paper attempts to formalize this problem by modeling the task as a finite-horizon stochastic dynamic problem. Our primary contribution is a basket based multinomial logit (MNL) choice model that attempts to model item substitutability and complementarity using historical orders. We demonstrate that these interactions are determined by the covariance of item co-purchase probabilities as derived from historical orders. Finally, we develop an approximate solution using Approximate Dynamic Programming (ADP) with Value Function Approximation (VFA) and compare the approaches using simulation techniques