Want to add Amazon.com style recommendations to your game's marketplace or user-generated content repository? This session will introduce the different approaches used to build item recommendations for a digital marketplace and discuss the tradeoffs involved in the different approaches. It will focus on the user-based, collaborative filtering algorithm being developed to recommend items for the EverQuest Landmark marketplace, which also includes a large amount of user-generated content. The talk will cover retrieval strategies, similarity measures, and evaluation of recommendation systems. Concrete examples will be provided using Apache Mahout, an open-source, machine learning package. Attendees will learn how SOE is prototyping different recommendation systems, evaluating the results, and scaling the system to a large player base.