Optimal Bidding on Keyword Auctions

A b s t r a c t Pay per click (PPC) auctions are used to sell positions in search engines. These auctions have gained increasing commercial importance , and many companies offer software to bid on these auctions. We present a trading agent for PPC auctions that is the first in our knowledge to use an explicit profit objective function. The agent creates a look-ahead plan of its desired bids, which allows it to exhibit intelligent behaviour including the ability to hold back money during expensive periods. We tested the agent in the latter part of 2003 in a live Overture auction. The agent generated four times the number of visits as human managers, in addition to reducing cost and variability. OVERVIEW In this paper we present a trading agent for pay per click (PPC) auctions. The agent exhibits many intelligent characteristics, including being able to plan ahead, allocate resources between auctions, and automatically explore. The first part of the paper will describe how PPC auctions work, including current automated approaches for bidding on these auctions. The second part will describe the major features of the agent including the formulation of its objective function, the method for estimating statistical functions and exploration strategy. The final part will describe an experiment in which we compared the performance of the agent against human controls. A PPC auction is a continuous second-price auction for advertising space in search engines. For instance, if a user types in a search at Google, they will get back a set of listings. At the right hand side of the page there will be some sponsored listings. These sites have paid, on a PPC auction, to have their companies shown after the user has entered a search meeting their specification (see Figure 1). This method of advertising is similar to the telephone Yellow Pages. In the Yellow Pages, advertisers pay more to have larger, more visible advertisements. In PPC, advertisers pay more to be listed higher in the returned results from an online search engine. Each competitor enters a bid b k,t which is the amount they are willing to pay should a customer click on their advertisement in the search results for keyword k at time t. For instance, a company may be prepared to pay up to $11.03 for a customer who typed 'home loan California' and $5.02 for a customer who typed 'home loan'. …