on revenue in the generalized second price auction
play

On Revenue in the Generalized Second Price Auction Brendan Renato - PowerPoint PPT Presentation

On Revenue in the Generalized Second Price Auction Brendan Renato va Lucier Paes Leme Tardos (MSR-NE) (Cornell) (Cornell) Sponsored Search Auctions Sponsored Search Auctions Two AdAuctions Mechanisms Generalized Second


  1. On Revenue in the Generalized Second Price Auction Brendan Renato Éva Lucier Paes Leme Tardos (MSR-NE) (Cornell) (Cornell)

  2. Sponsored Search Auctions

  3. Sponsored Search Auctions

  4. Two AdAuctions Mechanisms Generalized Second Vickrey-Clarke-Groves Price Auction Mechanism

  5. Two AdAuctions Mechanisms GSP VCG

  6. Two AdAuctions Mechanisms GSP VCG Main Question : How do those mechanisms compare from a game-theoretic viewpoint ?

  7. Two AdAuctions Mechanisms GSP VCG - sort by bid - sort by bid

  8. Two AdAuctions Mechanisms GSP VCG - sort by bid - sort by bid - complicated - pricing = next pricing rule highest bid

  9. Two AdAuctions Mechanisms GSP VCG - Simple - Complicated - Industrial - Optimal in standard theory

  10. With respect to which metric ?

  11. With respect to which metric ? Search Advertisers Engine Users

  12. With respect to which metric ? Search Advertisers Engine Clicks * (value/click – pay/click) Clicks * pay/click Users Usefulness of ads

  13. With respect to which metric ? Search Advertisers Engine Clicks * (value/click – pay/click) Clicks * pay/click Revenue Users Usefulness of ads

  14. With respect to which metric ? Search Advertisers Engine Clicks * (value/click – pay/click) Clicks * pay/click Revenue (our focus) Users Usefulness of ads

  15. With respect to which metric ? Search Advertisers Engine Clicks * (value/click – pay/click) Clicks * pay/click Social Welfare = ∑ Clicks . (value/click) Users Usefulness of ads

  16. With respect to which metric ? Search Advertisers Engine Clicks * (value/click – pay/click) Clicks * pay/click Social Welfare = ∑ Clicks . (value/click) ? Users Usefulness of ads

  17. With respect to which metric ? Search Advertisers Engine Clicks * (value/click – pay/click) Clicks * pay/click Social Welfare = ∑ Clicks . (value/click) [Varian], [Eldeman, Ostrovsky, Schwarz], [Paes Leme, Tardos], [Lucier, Paes Leme], [Caragiannis, Kaklamanis, Users Kanellopoulos, Kyropoulou], [CKKKLPLT], [Athey, Usefulness of ads Nekipelov], [Ellison, Athey ], …

  18. Warm-up: When we have one ad slot…

  19. Which ad to place and how much to charge?

  20. Which ad to place and how much to charge? $0.03 / click $0.01 / click $0.02 / click

  21. Which ad to place and how much to charge? $0.03 / click $0.01 / click $0.02 / click

  22. Which ad to place and how much to charge? $0.03 / click $0.01 / click $0.02 / click

  23. Which ad to place and how much to charge? Pays $0.02 / click (second highest bid)

  24. Selling Display Ads

  25. Selling Display Ads Single ad slot Second-price auction

  26. Selling Display Ads Single ad slot Second-price auction • simple • easy to bid (truthful) • maximizes welfare • maximizes revenue with optimal reserve r

  27. Selling Display Ads Single ad slot Second-price auction Theorem ( Myerson ’81) The revenue-optimal auction* for a single item is the second price auctions with a reserve price. * under suitable assumptions

  28. Usually we have many ad slots …

  29. … and not all of them are equal

  30. … and not all of them are equal Position Auctions auctioning positions on webpages How to place ads and how to charge for them with multiple positions.

  31. Our Results 1) Without uncertainty: GSP revenue is at least ½ of a natural VCG-like benchmark 2) With uncertainty (Bayesian setting): GSP with appropriate reserve price extracts constant fraction of the optimal revenue 3) Revenue and Welfare: What are the trade-offs?

  32. Our Results 1) Without uncertainty: GSP revenue is at least ½ of a natural VCG-like benchmark 2) With uncertainty (Bayesian setting): GSP with appropriate reserve price extracts constant fraction of the optimal revenue 3) Revenue and Welfare: What are the trade-offs?

  33. Related Work Position Auctions model due to [Edelman, Ostrovsky, Schwarz] and [Varian], who study revenue of GSP for a particular class of equilibria.

  34. Related Work Position Auctions model due to [Edelman, Ostrovsky, Schwarz] and [Varian], who study revenue of GSP for a particular class of equilibria. In contrast, we study revenue of GSP over all equilibria in settings with uncertainty.

  35. Position Auctions Model • n advertisers and n slots • slot i has click through rate • values per click i.i.d. • distributions are nice* • reserve price r * nice = regular, i.e., revenue curve is concave, e.g., uniform, exponential, normal, …

  36. Position Auctions Model • Strategy for i : a bidding function • Assume non-overbidding • Bidding functions form a Bayes-Nash equilibrium, i.e., no player benefits from changing bids even after learning their valuation

  37. Position Auctions Model α 1 α 2 α 3

  38. Position Auctions Model b 1 (v 1 ) α 1 b 2 (v 2 ) α 2 α 3 b 3 (v 3 )

  39. Position Auctions Model b 1 (v 1 ) α 1 b 2 (v 2 ) α 2 α 3 b 3 (v 3 )

  40. Position Auctions Model b 1 (v 1 ) α 1 b 2 (v 2 ) α 2 • Remove bids below r α 3 • Sort by bid b 3 (v 3 ) • Charge next highest bid or r [GSP] • Charge a complicated function [VCG]

  41. Our Main Result Thm : If are iid according to a regular distribution, then GSP with the appropriate reserve price r extracts at least a 1/6 fraction of the revenue of the optimal mechanism. is a Bayes-Nash equilibrium of GSP r then its revenue is at least 1/6 of the revenue of any mechanism and any equilibrium of this mechanism.

  42. Our Main Result Thm : If are iid according to a regular distribution, then GSP with the appropriate reserve price r extracts at least a 1/6 fraction of the revenue of the optimal mechanism. Why is this good ? 1/6 is an absolute worst-case theoretical guarantee. It assumes very little about distributions. The guarantee doesn’t depend on # of players, alphas, or characteristics of the distribution (besides regularity).

  43. Our Main Result Application : We give guidelines for setting reserve prices in GSP. Myerson’s Thm prescribes how to set reserves on VCG. [see Ostrovsky and Schwarz, EC’11] Ingredients of the proof : 1) study two cases: (i) revenue is distributed among many slots (ii) most of the revenue comes from one slot 2) passing to virtual values 3) prophet inequalities

  44. Revenue-Efficiency Trade offs Now in the non-Bayesian world, where are fixed: Opt Revenue ? How does the plot of all equilibria look like? Revenue Is there an equilibrium achieving maximum revenue and social welfare ? Welfare Opt Welfare

  45. Revenue-Efficiency Trade offs Observation: The revenue optimal equilibrium might be inefficient. 1.2 Cost of efficiency : ratio between revenue-optimal equilibrium and revenue- optimal efficient equilibrium. For vector : 1.0

  46. Revenue-Efficiency Trade offs Observation: The revenue optimal equilibrium might be inefficient. Thm 3: If CTR α are convex, then the revenue- optimal equilibrium is efficient. α 1 Revenue α 2 α 3 α 4 α 5 i Welfare

  47. Revenue-Efficiency Trade offs Thm 3: If CTR α are convex, then the revenue- optimal equilibrium is efficient. Proof Idea : - Structure of revenue in inefficient equilibria - Local improvement proof

  48. Conclusion • Bounds on revenue extracted by GSP – Full Information – Bayesian setting – Revenue x Welfare tradeoffs

  49. Conclusion • Bounds on revenue extracted by GSP – Full Information – Bayesian setting – Revenue x Welfare tradeoffs Thanks !

Recommend


More recommend