Plan for Today Continue with auctions Sponsored Search The VCG - PowerPoint PPT Presentation

Plan for Today • Continue with auctions • Sponsored Search • The VCG auction • The FCC Incentive Auction

Last time A Iii c 17g I is highest bidder price auction i winner winner pays her bid 2nd price archeni winner is highest bidder winner pays 2ndhighest bid How to bid each bidder bids to maximize Model given beliefs about others utility expected bidders Vi Ineceivesitem utility of pi value payment Vickrey auction 2nd price truthful anchen also called proof strategy dominant strategy incentive compatible Dsic a dominant strategy to it is bid value your

individually rational also IR bidder's utility guaranteed to be nonnegative afar maximizes called also surplus happiness of all total participants sum of utilities IE viIiwmsien pi tiEpi utility of utility of bidder i auctioneer n wi Ii I wins F I

1st price auction dominant strategies no analyzed by assuming each player's distribution value drawn from was F Showed that if F 0 too uniform is then L par is a equilibrium far h 2 Nash Bayes F utility of bidding pCyj1ohenp6yp vidagJJzEfuteeygb.dd.gov otmgfYfbjds pay µ Thm RevenneEquivalence for any anchor fi F then Vin If to allocated item sit in equilibrium bidden uf value and payment of highest a Oiso cuh valve player has same expected revenue

Advertising – how it used to be Posters newspapers magazines Billboards television

Pay-per-impression Price depends on how many people your advertisement is shown to. (whether or not they look at it, or care about it) “Half the money I spend on advertising is wasted; the trouble is, I don’t know which half” Andrew Wanamaker, advertising pioneer

How is the price determined? Complicated negotiations with high monthly premiums, forms a barrier to entry for small advertisers.

Modern Advertising On the web, Many different kinds of ads …

Sponsored Search Ads

Price determined by auction – per keyword Slot One Slot Two Ad A Ad B Ad C “Most people don’t realize that all that money comes in pennies at a time” Hal Varian, Google Chief Economist

How are these ads different than the ads in the offline media? In many cases: Pay-pe In per-cl click ck auction per keyword Model a standing bid on have who Bidders advertisers keyword of slots 1 K offer some Auction notidentical Slots CTRchukthroughrat on the jh K l j slot Prob of aclick Cj click for value adverts's a vii are sated all qualities CTR rate 9i due to gualitygad l l d f

bidder i's ad instotj overall ipf put Prob of Cj'gi click placed ad his i f gets bidder to Value slot's in Cj vi exp pig v g utility exp GSP Price Second Generalized bid g loBi aH slot j payment click to a on bid next is highest

Model Value per click Click-through rates utility PPC values bids At 1 7 1 A 6 c 0.4 1 6 I 0.4 6 2 B cz No slot 0 C 1 Cz Slots Advertisers

bidder 3 I Example 2 0.05 values o Ijg E 0.99 in this example below 9 bids 1 Y bidder then his utility will was be higher than it bidding truthfully VCG auction Truthful anchor maximize social welfare outcome to chooses S Vy S V allocation an social welfare of EIvicspcis is assigned to slot i C Cz Cz allocation YIximizing her externality the bidder charge each on others Imposes presence IBidduliT bidder 1 in archer Whats be way w I i a

VCE payment for bidder I incur because of her others loss charge bidder Ktl O cute c ai i Ci c i 2 j bidder VCG payment for KH Vi a ci i in jts sponsored search CG auction for T ftp.bjccj i 5 ai pi Get a bid from each bidder w bn by b Relabel bids bit GiI so bitafinicity ihs lots to assignbidder i K for i I chargebidder i for each click h K for i KH PPC b Kj I c pi j itithm and welfare maximizing IR truthful VCG.is Progg tmtfeness denote jth highest let bj bids Fix other me bid except my value Ckbk slot k V 1 cable Ca be K S Ca Li i 1 to slot move from i A value as bi I c i l i cm 0 price bi if want to do this v i

otherbids bit bi S V b by P has the GSP auction CTRs valuations payments same allocation an equilibrium y VCG as VCG paying p PI PIE comparison between revenue no sweeping VCE GSP

Facebooky can bid on what bidders events Various format of ads size placement of ads evaluating quality vi Cj more appropriate model each event type value far bidder has CTRs to figure out for advertisers difficult high quality outcomes easy advertisers happy biddig CG fully general I participants n outcomes of auction finite set of has private valuation each agent i Fw c R v w vilwt fauegsjntjpr.e.int aemsw F VCI choose outcome argmax Ebilw C 5 1 w C I Mw Lettnis on 07

O j bjlwT P mEq view'T pi ui View t.F.bjw9 wmeoxfpbo.tw choose bin t.wo.GL J maximize X bjlw w t to vi wEqgrax User interface for bidding band events they bidders allow to specify budget bidders allow requirements Computational

Pa valve value others O total they get get if a matinee 5 wana 5 Pb 1 4 3