160 likes | 291 Views
Online Ad Auctions. Hal R Varian February 16, 2009 Presented by : Sanket Sabnis. Key Concepts. The paper basically describes how Google and other search engines sell their internet ads.
E N D
Online Ad Auctions Hal R Varian February 16, 2009 Presented by : SanketSabnis
Key Concepts The paper basically describes how Google and other search engines sell their internet ads. When a user sends a query to the search engine, the system finds a set of keywords that match the query and determines which ads to show and where to show them. The expected revenue received by the search engine is the price per click times the expected number of clicks.
Importance of Ad Quality The interaction between the Advertiser , User and Google is shown. The search engine wants to improve User as well as Advertiser experience so that both will come back to the search engine thereby increasing revenue. U A G
Auction Rules Each advertiser a chooses a bid The advertisers are ordered by bid times predicted click through rate The price that advertiser a pays for a click is the minimum necessary to retain its position. If there are fewer bidders than slots, the last bidder pays a reserve price. Simple Auction :
Components of Ad Quality Score • Users vote with their clicks, which ads are more popular. • Relevancy determines which ads are relevant to which keyword. • High quality landing page leads to help the users find the information they are looking for and should have low popups and transparent in working
Ad Rank • Introducing the Quality of the Ad into the Ranking mechanism. • Ad Rank = Bid * Quality • Advertisers B, C, D are chosen to be displayed even though A had the highest bid.
How much does Advertiser actually have to pay ? (Cost Per Click)Adv = (Ad Rank)Adv below / (Quality Score)Adv What is the Value of Quality ? If Advertiser increases quality , CPC is reduced.
Auction Methodologies • Equilibrium Methodologies • All advertisers have same ad quality. • VCG ( Vickrey-Clarke-Groves) Mechanism • Rank ads in same way • Charge each advertiser cost that he imposes on other advertisers. • Turns out that the optimal bid is true value, no matter what others are bidding. • One problem with this algorithm is that it is vulnerable to click fraud since an advertiser benefits from clicks on lower positions
Bidding Behavior • Advertisers choose a single bid that will apply to many auctions. • We suppose that there is some reasonably stable relationship between an advertiser’s bid and the number of clicks that it receives during some time period. • Once an advertiser knows the cost-per-click and bid per-click function, it can determine its optimal behavior.
Undersold Auctions • Two slots • X1 = 100 clicks • X2 = 80 clicks • Value = .50 • Reserve price = 0.05 • Solve equation • P1 100 = .50 x 20 + 0.05 x 80 • P1 = 14 cents P2 = 5 cents • Total Revenue = 0.14 x 100 + 0.05 x 80 = $18
Oversold Auctions • Each bidder has to be indifferent between having his slot not being shown : • So ( value – prices)xs= 0 prices= value • For the previous 2 – slot example , with 3 bidders prices = 50 cents • Total Revenue = .50 x 180 = $90 • Revenue takes a big jump when advertisers have to compete for slots • That is why Google’s 98% of revenue is from Advertisement.
Advertiser Surplus Suppose that advertiser A is choosing some number of clicks Xaat a cost of Caand that it decides to change its bid so that it receives some smaller number of clicks, ˆXaand pay a smaller cost ˆCa. Determine what position, how much it would pay and how many clicks would it receive in the ad auction with this lower bid. A50 percent reduction in bids only results in a 12 percent reduction in costs.
Advertiser Surplus cont... For example, an Advertiser moving from position 4 to position 3 tends on average to increase the number of clicks by 20 percent. Then if we observe that a particular advertiser initially receives 100 clicks in position 4, we would estimate that it would receive 120 clicks if it bid enough to move to position 3. Ad configurations that are “fully sold”—have all the slots occupied—tend to have lower surplus than those that are “undersold.” This is because advertisers must compete for slots in the fully sold case, which tends to push their surplus down.
Number of Ads shown • Show more ads • Pushes revenue up, particularly moving from undersold to oversold, however • The Relevancy reduces thereby users click less in the future and revenue decreases • Optimal choice • Depends on balancing short run profit against long run goals from current clicks to long run loss from fewer clicks
Conclusion & Future Scope The total value enjoyed by advertisers is between 2 and 2.3 times their total expenditure on advertisement. The auctions have an interesting theoretical structure as well as significant practical importance. As more economic transactions take place online we will likely see other novel pricing mechanisms arise. Availability of real time data allows for fine tuning and constant improvement.