Detecting price and search discrimination on the Internet Jakub - PowerPoint PPT Presentation

Detecting price and search discrimination on the Internet Jakub Mikians*, László Gyarmati, Vijay Erramilli, Nikolaos Laoutaris Telefonica Research, *Universitat Politecnica de Catalunya 1 Telefónica Research

Customers buy the same product for different prices 2 Telefónica Research

We may not be aware that this could happen on the Internet as well 3 Telefónica Research

Price difference does not necessary equal price discrimination 4 Telefónica Research

Price discrimination practice of pricing identical goods to different people based on the highest price they are willing to pay (reservation price) 5 Telefónica Research

Why study price discrimination? 6 Telefónica Research

7 Telefónica Research

Market sizes $934B* $71B * according to Goldman Sachs, by 2013 8 Telefónica Research

Search Discrimination 9 Telefónica Research

Search Discrimination § e.g. Bobble: filter bubble due to search personalization @ GTech 10 Telefónica Research

Economic implications 11 Telefónica Research

12 Telefónica Research

How do we do it and what did we find? 13 Telefónica Research

Information vector: system No PD, no SD 14 Telefónica Research

Information vector: location 15 Telefónica Research

Information vector: location § 6 Locations: NY, LA, DE, SP, SK, BR § Everything same except IP address § NTP synchronized § NO discrimination.. except.. 16 Telefónica Research

Kindle e-books Difference: 21% to 166% 17 Telefónica Research

Steam Mean difference: 20% 18 Telefónica Research

Staples 19 Telefónica Research

Information vector: personal information Does your PI/interests, inferred via browsing information, cause PD? 20 Telefónica Research

We created two online personas Affluent Budget conscious 21 Telefónica Research

Personas based: Affluent ii) Enable tracking i) Visit sites that classify you as ‘affluent’ via AudienceScience Affluent 200 sites, 65 products 2 weeks 22 Telefónica Research

What do we see? § P r i c e d i s c r i m i n a t i o n : N O discrimination § Search: Some discrimination 23 Telefónica Research

Personas: Search Discrimination (cheaptickets) Mean difference ~ 15% 24 Telefónica Research

How would you do it? § Too much infrastructure needed § Use ad-networks? § Idea: Use origin/referer § Coming from a price aggregator site can out you as price sensitive 25 Telefónica Research

nextag -> shoplet Mean difference ~ 26% Can be due to special contracts 26 Telefónica Research

Disclaimers/Limitations § Preliminary study, 200 online vendors, 65 product categories § Fine scale temporal variations § We take measurements multiple times § Assume information vectors in isolation will trigger PD § Underestimating PD 27 Telefónica Research

Summary § Price discrimination is important tool to price § Developed a methodology to uncover PD § Initial results § Tool for price comparison, available for beta testing http://pdexperiment.cba.upc.edu 28 Telefónica Research