Demand estimation and inventory optimization under stock-out-based substitution and censoring

Participant(s)

Activity date	14 Jul 2018 → 16 Jul 2018
Degree of recognition	International

Description

Retail customers often have a favorite product in mind, but may switch to a substitute if their first choice is out of stock. Estimating demand from sales data then becomes difficult, as stock-outs censor demand for the primary product, and recapture demand to substitutes. We propose two methods for estimating first- choice demand and the substitution probability for Poisson demand and stochastic substitution between products: (1)Maximum likelihood estimation(MLE), (2)Bayesian updating. These two approaches assume that the exact timing of sales is known, while the latter requires only that sales are recorded at discrete checkpoints. The Bayesian updating estimation is easy to apply in practice, and exploits exact sales timings for better accuracy. We also introduce ordering policies to maximize the daily profit based on the estimated demand model.

Research Unit / Event Journal/Book Series