Bayesian Dynamic Pricing with Unknown Customer Willingness-to-Pay and Limited Inventory

摘要: We consider a stylized dynamic pricing problem with an unknown distribution of customer willingness-to-pay (WTP) and limited inventory. The seller learns customers' WTP from their binary purchase decisions, where the posted price serves as either left- or right-censoring point customer's WTP. is formulated finite-horizon Bayesian program. find that better information always improves revenue performance, but learning may bring negative value when inventory level becomes low. A derivative approximation heuristic devised for numerically solving problem. further develop performance bound to compare our proposed other benchmark heuristics. Numerical experiments demonstrate consistently outperforms others.