search engines: alexandre de corniere simon p. anderson university of virginia discussion

8
Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion

Upload: irma-howard

Post on 18-Dec-2015

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion

Search Engines: Alexandre de Corniere

Simon P. AndersonUniversity of Virginia

Discussion

Page 2: Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion

Discussion outline

• Very clean, elegantly done, clear• Several extensions – ready to ship

• Overview• Random serving up of surfers? (obfuscation)• Minor comments• Future directions

Page 3: Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion
Page 4: Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion

Model backdrop• Circle for products and consumers (continua)• Optimal consumer search as a stopping rule• Hence no point in advertiser paying a if won’t stop leads to (restricted) optimality result: search costs

minimized (one search) • Defines the advertiser width per consumer• Then find the (monopoly – all stop) product price (no

discrimination) over the disparate consumers

• Search engine serves up advertisers at random• But, (when) is uniform distribution optimal?

Page 5: Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion

Tensions in engine (platform) pricing

• Classic two-sided market balance: revenue per viewer times number of viewers

• Here though there is only one click per viewer (only one advertiser pays)

• Still deliver up eyeballs by rendering attractive the expected package

• Need here to temper Diamond Paradox by including some less desirable matches

• Still monopoly pricing (not price competition), but affecting demand elasticity

Page 6: Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion

Google’s secret sauce • What should be served up?• Not just the ideal match! Then price is reservation value of

highest valuation consumer. Zero surplus, so no-one clicks.• Instead of uniform, what probability distribution is optimal

for engine?• More density on ends? Reduces price, but worse matches• Bruestle (UVA): serving up the “wrong” eyeballs• So, find the algorithm! (not too hard?)• Reminiscent of threshold match (AR, AER ‘06) – tell enough,

but not too much

Page 7: Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion

Minor points

• A3 should be (1-F) log-concave • Caplin and Nalebuff (1989); Anderson, de

Palma, Nesterov (1985)

Page 8: Search Engines: Alexandre de Corniere Simon P. Anderson University of Virginia Discussion

Random serving assumption• Would the engine profit from a different algorithm?• Serve up the best? (what would happen? –

Diamond paradox?)• Maybe not …• Bruestle (UVA: “serving up the wrong eye-balls”)Similar ingredients; but serving up some “wrong”

surfers How it works: full-fledged 2-sided market logic: get

more surfers on-board (to deliver to discipline prices)