Pay Per Click
Bidding Agent Competition
ACM Conference on E-Commerce 2006
Sponsored Search Workshop
Congratuations to all teams who particiaptated in the 2006 Bidding Agent Competition! Bidding Agent Competition results are now available!
The 2006 Bidding Agent Competition
Are you interested in e-commerce / machine learning / operationd s research / data mining / artificial intelligence? Put your skills to the ultimate test. Enter an intelligent bidding agent in the 2006 Sponsored Search Bidding Agent Competition . Your intelligent agent will need to balance exploration versus exploitation in real-time (the classic k-armed bandit problem), and react intelligently to fickle and potentially explosive changes in customer behavior.
But we're not just talking about a mathematical exercise. You will have the opportunity to actually bid in the real-world, with real money, and with the chance of making or loosing your funds. Through a special arrangement with Microsoft AdCenter - who are providing the auction platform and are offering free funds to teams - and a retailer who is yet to be announced - we are offering you the chance to bid with real money on a real retailer. The winner will be the team that creates the greatest percentage increase in revenue. Are you ready for this challenge? Read on!
What are Pay Per Click Sponsored Search Auctions?
Pay for click (PPC) auctions are arguably the most successful new form of advertising to emerge on the internet. They attract over $4 billion in advertising dollars per year, and more advertising dollars than any other form of internet advertising including banner advertisements. Marco Ottaviani provides a good introduction to pay per click auctions, and so does Aradise.com.
Pay per click auctions are fascinating both because of their commercial success, their propensity towards relevance (Jensen, 2005), and because of the mathematical optimization problem which they pose. Bidders are faced with thousands of keywords that need to be priced, and these bidders need to change their prices to take advantage of changes in consumer behavior which changes by the minute.
Bidding is challenging because of the high risk inherent in maximizing revenue. If costs are incurred and revenue events do not follow due to noise or a fundamental shift in consumer behavior, then the bidders could literally go bankrupt. Sophisticated machine learning systems are therefore required to balance these risks and achieve maximal returns.
Purpose of Competition
The Second Sponsored Search workshop, with sponsorship from ACM Special Interest Group on E-Commerce, Microsoft corporation and Yahoo corporation , are pleased to announce the first ever Sponsored Search Bidding Agent Competition.
The Bidding Agent Competition will allow entrants to test the effectiveness of their bidding strategies in a fair and controlled setting. The winning team will generates the greatest percentage increase in profit within a set period of time.
The purpose of the competition is to:
(a) serve as an educational and fun event, which will encourage more interest in science and technology
(b) contribute to the scientific literature on these auctions
(c) allow teams to test different strategies and find ways of improving their monetization of these auctions, and ultimately to write about these strategies (for more details about disclosure, please see the registration process).