Article Preview
TopIntroduction
Agent-based electronic commerce is one of promising techniques to enhance effectiveness and performance of trading. In this paper, we give an analysis of agent-based advertisement auction, which is displayed on a webpage.
Advertisements on the webpage provide good opportunity to get new customers. In recent years, a lot of webpages providing a search service have advertisements, which are related with searched word by user. Some searching company has a advertising space on his/her webpages and allocates it for some advertisers based on an advertising fee. As same as items trading in the Internet auctions, a displayed advertisement on web page is also based on the auction, called the Internet advertisement auction. Internet advertisement auction is one of important income source for some search engines such as Yahoo! and Google (Yahoo!; Google). When users search for some words on the search engine, an advertisement related with the searched keywords is displayed with result of search (Edelman, 2007). The order of advertisements to be displayed is determined based on bid value in an auction. Advertisers can set up the interval and period to display the advertisement as a time slot. The advertising fee is determined based on the Generalized Second Price Auction, which is known higher revenues than the Generalized Vickrey Auction (Vickrey, 1961). A winner in the auction gets a space to display their advertisement and the web page owner allocates time and position in the web page to show the advertisement. There are a lot of contributions about GSP(Generalized Second Price Auction) researches in electronic commerce research. In this auction, bidding and winner determination are conducted multiple time. Advertiser can change his/her bid value because the auction is continued with repetition. When advertisers try to bid in an auction, they bid on their strategy. However, GSP has an envy free equilibrium and webpage owner providing advertisement space can get larger benefit compared with VCG (Vickrey-Clark-Groves) Mechanism.
Figure 1 shows the interface of searched result of “travel” in Google. In right side of the interface, advertisements are shown. These advertisements are related with the searched words. When the advertisement is clicked by end-users, advertisement fee is paid from advertisement companies to the website owner (Google). Generally, possibility of click is high order of display. This means that the advertisement fee of top-displayed advertisement is more expensive than lower advertisements. Google earned about 5.2 million USD by this advertisement system in 2008.
Figure 1. Capture of Google adwords
In previous research, the value of advertisement is assumed as independent with each other. Otherwise, some of their researches do not refer the value of the advertisement. However, each advertisement has a certain value for users. It means some criterions of a name value of a company, effectiveness, an importance and an attribution, that is dependently each other. When same or similar item is sold in two e-commerce sites, the price on the advertisement is different from another one. If a buyer considers the price is important attribute to choose item, the advertisement selling items at low price has more value for the buyer. For example, a shop A gives an advertisement to sell an item for $100. When a shop B gives the advertisement to sell the same item for a shop $80, its value of the advertisement is higher than shop A's value if the condition of item and other situations between shop A and B. In this paper, we focus on such situation and simulate the revenue of advertisers. Also, we analyze a result of simulation of Internet advertisement auction with relationship between values of each advertisement. After the simulation, we reformulate our proposed model and mechanism based on the preliminary simulation. Concretely, we discuss about dynamical environment. It is more realistic situation of the advertisement market on the Internet.
The rest of this paper consists of the following four parts. First, we show preliminaries on several terms and concepts of auctions. Then, we propose our value-based GSP and describe a preliminary experiment in some conditions. And we reformulate our proposed model and mechanism for applying a dynamic environment based on the preliminary simulation. Also we discuss some applications of the proposed model, especially about the GrobalAd system. Finally, we present our concluding remarks and future work.