Pick-a-bundle: a novel bundling strategy for selling multiple items within online auctions

In this paper, we consider the design of an agent that is able to autonomously make optimal bundling decisions when selling multiple heterogeneous items within existing online auctions. We show that while bundling the items together into a single lot is effective at reducing listing costs, it also results in a loss in auction revenue. To address this loss we introduce a novel bundling strategy, that we call pick-a-bundle, that can be implemented within any existing auction format. We show, mainly using simulations, that this new bundling strategy generates greater expected revenue than the complete bundle of all items, and, by inducing additional competition between bidders, it usually generates greater expected revenue than using separate auctions for each item. In order for our agent to accurately and efficiently calculate its expected revenue when using our new strategy, we derive a novel polynomial time algorithm for calculating the probability distributions of the sum of the top...

• Type: Conference
• Year: 2009
• Where: ATAL
• Authors: Ioannis A. Vetsikas, Alex Rogers, Nicholas R. Jennings
• Keywords: Artificial Intelligence, ATAL 2009, Bundling Strategy, Greater Expected Revenue, Optimal Bundling Decisions