Economics and Computation Ad Auctions and Other Stories Christopher - PowerPoint PPT Presentation

Economics and Computation Ad Auctions and Other Stories Christopher A. Wilkens UC Berkeley March 6, 2013 1

Why mix Economics and Theoretical Computer Science? 3

Alan Turing, 1936: Introduced the Turing Machine as a tool to understand the limits of Logic . Image Source: Wikipedia 4

Alan Turing, 1936: Introduced the Turing Machine as a tool to understand the limits of Logic . Looking back... The limits of Logic cannot be fully understood without computational ideas! Image Source: Wikipedia 4

Alan Turing, 1936: Introduced the Turing Machine as a tool to understand the limits of Logic . Looking back... The limits of Logic cannot be fully understood without computational ideas! Economics today: Many important questions about complex economic systems require a Image Source: Wikipedia computational perspective. 4

Today’s Talk Sponsored Search Auctions – First-Price Auctions: How can we design first-price auctions that perform well? – Coopetitive Ad Auctions: Recognizing complexity may be important for performance. Market Equilibria – Complexity Equilibria in Markets Computational complexity begets stability. 5

The Sponsored Search Auction 6

Sponsored Search — History A long time ago in a galaxy far, far away... Image Source: Computer History Museum 7

Sponsored Search — History Idea: Willingness to pay is a proxy for relevance and quality. 8

Sponsored Search — History 1996: GoTo.com introduces paid search. Image Source: Computer History Museum 9

Sponsored Search — History The GoTo.com Model: Intel bids $2 for “Intel Laptop” . When user searches for “Intel Laptop” ... – Results for “Intel Laptop” sorted by bid. – Intel pays $2 if user clicks on link to www.intel.com (pay-per-click, PPC) 10

Sponsored Search — History Today: Ads shown alongside organic results. 11

Sponsored Search — History GoTo.com launches first successful search engine 1996 with sponsored results. GoTo.com partners with Yahoo!, MSN, AOL, and others to show sponsored results alongside organic ones. Google launches Adwords platform 2000 GoTo.com changes name to Overture.com. 2001 Overture purchased by Yahoo! 2003 12

Aside — Single-Item Auctions Alice b = 5 Item Bob b = 10 Charlie b = 1 Dave b = 8 13

Aside — Single-Item Auctions Alice b = 5 e m Item s I t W i n 0 = 1 p y s p a Bob b = 10 First-Price Auction: Charlie Highest bid wins (Bob), b = 1 pays own bid ( p = 10). Dave b = 8 13

Aside — Single-Item Auctions Alice b = 5 e m Item s I t W i n 8 = p a y s p Bob b = 10 Second-Price Auction: Charlie Highest bid wins (Bob), pays b = 1 second highest bid ( p = 8). Auction is truthful, because Dave no bidder can gain by lying! b = 8 13

Aside — The VCG Auction Question: What about truthfulness in complex auctions? The Vickrey-Clarke-Groves (VCG) Auction: – Pick the socially optimal allocation of goods to bidders. – Payment is negative externality: p i = Welfare − i [excluding i ] − Welfare − i [including i ]   �  Welfare − i = [ j ’s value for chosen allocation]  j � = i 14

Aside — The VCG Auction Alice b = 5 Item Bob b = 10 Charlie b = 1 Dave b = 8 15

Aside — The VCG Auction Alice b = 5 e m Item s I t W i n Bob b = 10 Charlie With all bidders, Bob wins and b = 1 nobody else gets anything: Welfare − i [including i ] = 0 Dave b = 8 15

Aside — The VCG Auction Alice b = 5 e m Item s I t W i n Bob b = 10 m e t I s n i W Charlie Without Bob, Dave wins: b = 1 Welfare − i [excluding i ] = 8 Dave b = 8 15

Aside — The VCG Auction Alice b = 5 e m Item s I t W i n Bob b = 10 m e t I s n i W Charlie Bob pays: b = 1 p Bob = Welfare − i [excluding i ] − Welfare − i [including i ] = 8 − 0 = 8 Dave b = 8 15

Sponsored Search — History of the Auction GoTo.com uses generalized first-price auction (GFP) 1996 to sell sponsored results. Bids are unstable. Google’s Adwords program sells advertising through 2000 monthly contracts. Google introduces generalized second-price (GSP) auction for Adwords. Features: 2002 – Results “ranked by revenue.” – Payment is “next highest bid.” 16

Sponsored Search — History of the Auction Generalized Second-Price (GSP) Auction: Company i bids $ b i for query ... – Click-through-rate (CTR) is the likelihood a user clicks on i ’s ad when shown in slot j : c i , j = α j × β i – Expected revenue is � � R = c i , j ( i ) p i = α j ( i ) β i p i i i – Sort results by β i × b i . – Per-click payment p i is minimum bid required for current rank: p i = β i +1 b i +1 β i 17

Sponsored Search — GSP Example 18

Sponsored Search — GSP Example Slots � b BestBuy = $1 Best Buy α 1 = 0 . 5 β BestBuy = 0 . 045 � b Intel = $2 . 10 α 1 = 0 . 25 Intel β Intel = 0 . 1 � α 1 = 0 . 10 b Newegg = $1 Newegg β Newegg = 0 . 09 � b Samsung = $2 Samsung β Samsung = 0 . 09 19

Sponsored Search — GSP Example Slots � b BestBuy = $1 Best Buy α 1 = 0 . 5 β BestBuy = 0 . 045 � b Intel = $2 . 10 α 1 = 0 . 25 Intel Rank Ads: β Intel = 0 . 1 Rank ads by β × b . � α 1 = 0 . 10 b Newegg = $1 Newegg β Newegg = 0 . 09 � b Samsung = $2 Samsung β Samsung = 0 . 09 19

Sponsored Search — GSP Example Slots � b BestBuy = $1 Best Buy α 1 = 0 . 5 Intel β BestBuy = 0 . 045 � b Intel = $2 . 10 α 1 = 0 . 25 Samsung Intel β Intel = 0 . 1 � α 1 = 0 . 10 Newegg b Newegg = $1 Newegg β Newegg = 0 . 09 Best Buy � b Samsung = $2 Samsung β Samsung = 0 . 09 19

Sponsored Search — GSP Example Slots � b BestBuy = $1 Best Buy α 1 = 0 . 5 Intel β BestBuy = 0 . 045 Compute Payments: � b Intel = $2 . 10 α 1 = 0 . 25 Samsung Intel pays minimum bid needed to beat Samsung: Intel β Intel = 0 . 1 p Intel = β Samsung × b Samsung = 0 . 09 0 . 10 × 2 = $1 . 80 β Intel � α 1 = 0 . 10 Newegg Samsung must beat Newegg, etc... b Newegg = $1 Newegg β Newegg = 0 . 09 Best Buy � b Samsung = $2 Samsung β Samsung = 0 . 09 19

Sponsored Search — GSP Example Slots � b BestBuy = $1 Intel Best Buy α 1 = 0 . 5 p Intel = $1 . 80 β BestBuy = 0 . 045 Samsung � b Intel = $2 . 10 α 1 = 0 . 25 Intel p Samsung = $1 β Intel = 0 . 1 Newegg � α 1 = 0 . 10 b Newegg = $1 p Newegg = $0 . 50 Newegg β Newegg = 0 . 09 Best Buy � b Samsung = $2 Samsung β Samsung = 0 . 09 19

The Sponsored Search Auction: First-Price Auctions work with Darrell Hoy and Kamal Jain 20

Sponsored Search — First-Price Auctions Problem: The GFP sponsored search auction is unstable and revenue suffers. 21

Sponsored Search — First-Price Auctions Problem: The GFP sponsored search auction is unstable and revenue suffers. Solution (Hoy, Jain, and W): Change the bidding language. Get: – Strong static performance. – Dynamic convergence. 21

Sponsored Search — First-Price Auctions Instability in GFP: a Bidding War Intel 0 . 12 β × Bid Samsung Newegg 0 . 10 0 5 10 15 20 When Intel passes β Intel × b Intel = $0 . 18, Samsung drops its bid... ...and Intel follows. 22

Sponsored Search — First-Price Auctions Lahaie 2006, Edelman and Ostrovsky 2007: GFP does not have a pure-strategy equilibrium. 23

Sponsored Search — GFP Example Slots � b BestBuy = $1 Best Buy α 1 = 0 . 5 Intel β BestBuy = 0 . 045 � b Intel = $2 . 10 α 1 = 0 . 25 Samsung Intel β Intel = 0 . 1 � α 1 = 0 . 10 Newegg b Newegg = $1 Newegg β Newegg = 0 . 09 Best Buy � b Samsung = $2 Samsung β Samsung = 0 . 09 24

Sponsored Search — GFP Example Slots � b BestBuy = $1 Intel Best Buy α 1 = 0 . 5 p Intel = $2 . 10 β BestBuy = 0 . 045 Compute Payments: Samsung � Intel pays its bid: b Intel = $2 . 10 α 1 = 0 . 25 Intel p Samsung = $2 β Intel = 0 . 1 p Intel = b Intel = $2 . 10 Newegg � α 1 = 0 . 10 Samsung also b Newegg = $1 p Newegg = $1 Newegg pays its bid, etc... β Newegg = 0 . 09 Best Buy � b Samsung = $2 Samsung β Samsung = 0 . 09 24

Sponsored Search — GFP Example Slots � b BestBuy = $1 Equilibria cannot exist: Best Buy α 1 = 0 . 5 Intel β BestBuy = 0 . 045 (a) Samsung must bid minimum to beat Samsung � Newegg. b Intel = $2 . 10 α 1 = 0 . 25 Intel p Samsung = $1 β Intel = 0 . 1 Newegg � α 1 = 0 . 10 b Newegg = $1 p Newegg = $1 Newegg β Newegg = 0 . 09 Best Buy � b Samsung = $2 Samsung β Samsung = 0 . 09 24

Sponsored Search — GFP Example Slots � b BestBuy = $1 Intel Equilibria cannot exist: Best Buy α 1 = 0 . 5 p Intel = $0 . 90 β BestBuy = 0 . 045 (a) Samsung must bid minimum to beat Samsung � Newegg. b Intel = $2 . 10 α 1 = 0 . 25 Intel p Samsung = $1 (b) Intel must bid β Intel = 0 . 1 minimum to beat Newegg Samsung. � α 1 = 0 . 10 b Newegg = $1 p Newegg = $1 Newegg β Newegg = 0 . 09 Best Buy � b Samsung = $2 Samsung β Samsung = 0 . 09 24