Negotiating Beyond the Bid Price

This paper discusses the limitations of current on-line auction systems and presents our guidelines for more dynamic attribute-based bidding. Our prototype application for airline flight bidding demonstrates our vision for interfaces that facilitate negotiation between buyers and sellers. INTRODUCTION Despite the growth in the number of on-line auction sites, there is still a need for a more dynamic, personalized bidding experience. Existing bidding and auction sites overemphasize bid price as the sole parameter determining the match of a buyer and a seller. We believe that for dynamic pricing systems to truly benefit the buyer and seller, the negotiation interaction needs to extend further than a simplistic bid price exchange. For example, today’s on-line auction systems, such as eBay [1] and Amazon Auctions [2], require a buyer to first locate the exact product they are seeking and then enter a committed bidding relationship with the seller. So while the system offers little assistance in locating a desired product, it also discourages the buyer from bidding on more than on item at a time by requiring the buyer to pay for all accepted bids. In this manner, the only undetermined factor of the purchase is price. Priceline.com [3], the predominant airline auction site on the Internet, follows a one-time-only bidding model with its own share of drawbacks. At Priceline.com, a buyer picks an origin, destination and date of travel, and then enters the price he/she is willing to pay for a matching flight. Priceline searches a database of available flights and if it finds a matching flight with a price lower than the asking bid price, the bid is accepted and the buyer gets the airline ticket for his/her asking price [11]. The Priceline system allows the buyer to specify only the barest of requirements and requires the buyer to purchase any accepted flight bids. Because this arrangement can lead to undesirable flights with poor travel times and multiple connections, a buyer is forced to always bid for the “worst case scenario.” This is both a disservice to the buyer and the seller -robbing the buyer of an opportunity to pay more for the “ideal” flight and preventing the airline from selling their more desirable flights at higher prices. Another drawback of the Priceline system is that it takes advantage of the uneducated bidder. Priceline buyers need to research available fares at alternative web sites and travel agencies before submitting a bid, or they can end up paying more for a less-desirable flight than they would have paid through traditional, fixed-price purchase methods. On-line auctions and Priceline’s airline bidding system violate several principles we believe are necessary for a bidding system to benefit both the buyer and the seller. These guidelines are: • Products should be evaluated and selected on multiple criteria, not just price. • A buyer should not be required to already know specific information about the product’s attributes, such as the going price or the details of a new feature. Today’s online auctions and Priceline work with the assumption that you have done prior research. • The negotiation should be a non-binding arrangement, allowing the buyer to make multiple bids on multiple products, increasing the chance of a successful match. • Sellers should have the tools to evaluate bids based on complex criteria, not just immediate revenue. To demonstrate these guidelines, we have built an alternative airline flight bidding system, called SARDINE (System for Airline Reservations, Demonstrating the Integration of Negotiation and Evaluation). This system is in the initial stages of development and the work we present in this paper demonstrates our current progress. We chose to focus on the domain of airline ticket purchases for our prototype because the airline industry is relatively sophisticated in on-line reservations systems and is a leader in the area of yield management. Airlines today already perform limited dynamic pricing and consumers are accustomed to the concept of different prices for different flights and changing prices over time. For these reasons, we feel it is a good domain to explore, and we plan to later apply our techniques more generally to other domains. Figure 1: Agent Interaction OVERVIEW OF SARDINE The SARDINE system presents a method for buyers to submit ticket bids to airlines and for airlines to respond to each bid. The system uses software agents to coordinate the preferences and interests of each party involved, as shown in Figure 1. The Buyer Agent collects the buyer’s preferences and correlates these parameters with available flights from a reservation database (Step 1). The user then tells the Buyer Agent how much to bid and the Airline Agents accept the ticket bids from the Buyer Agent (Step 2). Finally, the Airline Agents consider the individual bids based on flight yield management techniques and specific buyer information (Step 3). Step One To collect the buyer preferences, the Buyer Agent asks the buyer to indicate his/her preferred flight parameters and how flexible he/she is on each parameter (Figure 2). After specifying an “ideal” date of travel, time of day, airline, number of connections, etc, these preferences are submitted to the Buyer Agent for analysis. The flexibility rating of “very flexible,” “somewhat flexible,” and “not flexible” is used by the Buyer Agent to determine an acceptable range for each value. For example, with a departure time of 5pm, “not flexible” translates to a departure time between 4:45pm and 5:15pm, “somewhat flexible” means between 2:45pm and 7:15pm, and “very flexible” means between 12:45pm and 9:15pm. Additionally, this flexibility metric is used as an indicator of how important each parameter is to the user. When a user is very flexible on a certain flight attribute, this indicates it Figure 2: Travel Preferences is not as important as an attribute that the user is not flexible on. Figure 3 presents the distance function used to match the user’s preferences to the available flights. Weight and range for each parameter, i, is computed from the flexibility rating, as described above. The ideal value is the value entered by the buyer for the parameter and the actual value is the value for that parameter of a located available flight. Summing over all parameters computes the distance the available flight is from the user’s ideal flight. The Buyer Agent locates the best flights for the user by minimizing this distance function.