A Study On Online Reverse Auctions Of Perishable Commodities Based On Priceline

Electronic commerce has provided a wide field for the new business models. Meanwhile, all kinds of business models have enriched the electronic commerce's development. In the Business-to-Customer situation, one of them is reverse auctions represented by Priceline.com. As the unique business model of Name Your Own Price, Priceline has owned millions of registration users and rapid growth of business income. It mainly deals in perishable commodities and has been a famous international reverse auction website since it was operated. However, quantitative research on Priceline is relatively not rich. The dissertation aims to study customers' behavior and sellers' policies based on Priceline, mainly on customers' bidding strategies and sellers' pricing policies. The main contribution of the dissertation includes the following three aspects.First, the dissertation studies the reserve price policies for the sellers and the bidding strategies for the customers by static game theory under incomplete information. We show under some basic assumptions that the optimal reserve price is increasing in the cost, but less than the cost for the seller. We also show that customers' bidding prices are increasing in their valuation, and the optimal bidding prices are exclusive under some condition. The results have been illustrated by numerical analysis.Second, this dissertation studies a problem of how to dynamically set reserve prices for some goods of the seller in a period of time when customers arrive at Priceline one after another according to a non-stationary Poisson process, so that the seller can gain maximal expected revenue. We consider this problem in two aspects, one is the continuous time model, and the other is the multi-stage model. The dissertation discusses the continuous time model first. The optimal control model of the expected revenue for sellers is presented, and some theoretical results have been obtained. After that, we set up a multi-stage mathematical model with dynamic programming, and make analysis on this model. Some properties of optimal value and optimal price are obtained, which are quite similar to the ones of continuous time model. Numerical analysis are consistent with theoretical results, and the comparison of traditional revenue management with continuous time model illustrates that the closer the bidding price is to the valuation, the more dominant Priceline will be to revenue management.Finally, we study revenue management problems with either customer-pricing or seller-pricing. Here, a seller wants to sell a given amount of items in a fixed time, customers arrive according to an arbitrary renewal process. In the customer pricing, customers bid and then the seller decides to accept or reject their bids; while in the seller pricing, the seller decides the price and then customers determine to accept or reject the price. Markov decision processes models are presented and expressions for the optimal policies are obtained. These problems differ from the traditional revenue management problems where (1) the seller sets continuously a price, and (2) customers arrive according to a Poisson process. It is shown for the customer-pricing that there is no impact on the seller whether or not he knows the customers' private information, the optimal policy is monotone in the remaining items, and the optimal value is the concave function of the remaining items. Similar results of expected revenue and optimal price strategy have been obtained. Also, the expected revenues of the seller in the two pricing cases are compared, and the models and the results are generalized to the multiple demand case. Finally, the models and the results are illustrated by numerical analysis.