Webzing the economy & economizing the Web 4/4/2013 Michalis Vafopoulos vafopoulos.org Do you believe it? That a mathematician made an online call to solve important and difficult mathematical problems by coordinating many researchers through the Web? Public spending in Real time? It is the economy, stupid! How long takes to have 50 million users? • 38 years for telephone • 13 years for television • 4 years for Internet • 3 years for iPod • 2 years for Facebook • <1 year for Google + • the next: Linked Data company??? 5 Main issues ①Web economics & business ②Goods in the Web ③Users ④Consumption and Production in the Web ⑤The Web of Data emerging industry ⑥Student evaluation & how to work 6 the Web Science perspective ① Internet economics: the predecessor ② Partial analysis of the Web economy – e.g. network economics, digital goods etc. ③ Mainly focus on business implications ④ Issues: Auctions, e-commerce, search engines ⑤ Lately, net neutrality & excessive market power ⑥ Web science perspective – Standalone artifact – How the Web transforms economy and business 7 Web economics • Introduction • The Web Economy – – – – – – Economy after the Web – Existing theories and missing tools • Goods in the Web – – – – Information goods Digital goods Network goods Web goods Network Effects Peer & non-market production Market Structure Antitrust regulation Web-based development • The Web Business – – – – 8 E-commerce Business models Advertising & sponsored search User behavior analysis Goods in the Web ①Data, information, knowledge ②Information goods ③Knowledge goods ④Digital goods ⑤Web goods 9 Information goods: definitions Definition I the good, which main market value emanates from the information it contains. Definition II anything that can be digitized (Varian) 10 Information goods: characteristics • • • • • • high fixed cost of production low marginal cost of reproduction increasing returns to scale experience good public or a private good non-rival and sometimes non-excludable 11 Knowledge goods exogenous or endogenous inputs in production as: ①know-what (facts) ②know-why (scientific knowledge) ③know-how (skills) ④know- who (networks) 1, 2 easily reproducible 3, 4 not easily reproducible 4 more important in the Web era 12 Digital goods Bits with economic value, which are (Quah): ①nonrival ②infinitely expansible ③(Initially) discrete or indivisible ④aspatial ⑤recombinant 13 Externality ①analyzes the impact that individual decisionmaking has on the other agents ②comparison of how decision-making involves others without exchange ③Positive (i.e. education) or ④Negative (i.e. profiling) 14 Network externality Network externality: ①Some goods/services create more value when more users consume the same goods and services ②They have little or even no value if they are used in isolation (e.g. telephony) 15 Network externalities in the Web Source of externalities =linking ①Web 1.0: documents (demand) ②Web 2.0: Users (supply) ③Web 3.0: structured data (?) 16 Network externalities in the Web Linked Data • bidirectional and massively processable interconnections among online data • enabler for existing infrastructures 17 Network externalities in the Web Negative: • lack of trust • security, • identity theft • clickjacking • Spamndexing • spoofing • … 18 The basics of network modeling You see a product… 20 And some reviews 21 And some recommendations… 22 The Amazon co-purchasing network Item X co-purchased most frequently with products Y1, Y2,.. 23 Amazon: the book-based multi-store The Amazon copurchase network for all item categories 24 hidden complementarity saves MS MS (purple) & Apple (orange) communities are “mediated” by compatibility like VMware Fusion, Parallels Desktop and compatible products like Office for Mac. Triad analysis: Winners in Product wars Analysis shows that copurchase links not only manifest complementary consumption, but also reveal competitive relations among products that are perfect substitutes. 27 Switching to best sellers the case of Internet security market Ass: if products A, B & C are perfect substitutes (authority triad), then A has higher sales rank consumers who bought Internet security s/w, more often, also bought Norton Internet Security than related products 28 Paychecks… 29 b a n k r u p t c y … 30 Why networks? • To be self-contained (actor)—and to be thoroughly dependent (network)—is to say twice the same thing. (Actor-Network theory) • Easy to model and visualize relations • Easy to calculate major statistics • The study of the Web network help us to conclude that most of real networks are: – Self-similar (Scale-free) – Small worlds 31 Network theory and related fields Web Science Financial Network Analysis Social Network Analysis NETWORK THEORY Computer Science Graph & Matrix Theory Biological Network Analysis Initial source Soramaki how? • Define: 1. Node (e.g. person, business) 2. Link [directed or not] (e.g. friendship, commerce) And if necessary: 3. Evaluation of node (e.g. score, potential) 4. Evaluation of link (weight) 4 0.54 (e.g. trust) 5 33 Web Goods: definition Existing approaches fail to both capture the digital and the network dimension (aka virtualization) 34 Web Goods: definition sequences of binary digits that • are identified and communicated by an exclusively assigned URI and • affect the utility of or the payoff to some individual in the economy. 35 Web Goods Their market value stems from the digital information they are composed from and a specific part of it, the hyperlinks, which connect resources and facilitate navigation and editing over a network of Web Goods with minimum cost. 36 Web Goods: categories • Pure: basically exchanged and consumed in the Web and are not tightly connected to an ordinary good or a service (pre-) existing in the physical world. • Non pure (e.g. car’s photo in the Web) 37 Web Goods: categories • commercial (e.g. sponsored search results) • non-commercial (e.g. Wikipedia entries) ---- • public (e.g. Linked Open Data) • private (e.g. subscription to online magazine) – financial fee – “personal data” fee – “social” or “membership” fee 38 39 Web economy 40 Consumption in the Web • More energetic and connected consumption – search and review, collaborative filtering – what connected consumers create is not simply content (e.g. product reviews) but context. • Consumer coordination at large in the Web: the Amazon co-purchase network 41 Consumption in the Web • Personal data abuse and regulation challenges • Joint consumption of information and advertisements in massive scale • Moving the borders between production and consumption 42 Moving the borders between production and consumption 43 Production in the Web ① Inputs: information and knowledge reloaded ② Incentives: from property to commons ③ Peer Production: decentralized inter-creativity outside the classic market ④ From mass to networked media 44 Incentives: from property to commons Property rights can be further analyzed to 4 parts: i. The right to use economic resources. ii. The right to modify form and substance of resources. iii. The right to benefit from use of resources. iv. The right to transfer resources. • Traditional economy: 1st consumers, the rest producers • Web? • the 4th P: Property, Procurement, Patronage and Peer Production (commons) 45 Web business ①IT vs. Web economy ②Google model o The Sponsored-search market o Who’s the data? ③App vs. Web economy o patent’s war o Apple: the digital zombie o HTML5 effect ④ The Web of Data emerging industry ⑤ Discovering the market sentiment in the Web 46 Types of network externalities ①Direct (e.g. mobile phones) ②Indirect (e.g. mobile phone accessories) ③Two-sided network effects (or multi-sided platforms) (e.g. hardware-software platforms and the Google’s advertising platform) 47 Issues in network markets Network monopoly (e.g. Microsoft, Google) Possible regulatory policies: ①Divestiture of the monopoly into separate firms. ②Unbundling or wholesale access to incumbent’s facilities (e.g. Internet explorer). ③Licensing of proprietary interfaces to potentially competing platforms. 48 Web Goods vs. Digital Goods • restricts non-rivalry and infinite expansibility (concurrency capacity) • initially discrete and indivisible, but • Web 2.0: micro-chunks consumption • easily edit, interconnect, aggregate and comment • extends aspatiality and atemporality from local (e.g. personal HD) to global level (e.g. downloadable file link) 49 Web Goods as commodities information and knowledge: multiple and controversial definitions Web Goods: qualify as commodities (Debreu, 1959) • stable identity (URI) • completely specified physically • temporally and spatially (reside physically in a Web server during a specific period of time) 50 Consumption & Production in the Web Existing literature: the Web will lower prices because: ① lower search and fixed costs ② less product differentiation (e.g. location is less important) ③ “frictionless commerce” Actually: no much evidence The real transformation: More choices with less transaction costs in production and consumption. 51 information and knowledge reloaded the production of information is based on 3 inputs: (a) existing information (b) the mechanical means of conceiving, processing and communicating information and (c) the human communicative capacity (geography still matters in some sectors) 52 Peer Production: decentralized intercreativity outside the classic market ① virtuous cycle : productivity creates new knowledge, attracts new Users, increase productivity, creates new knowledge... ② A peer’s private productivity < his social productivity due to supply-side knowledge externalities. ③ Peer Production happens if Users do not take advantage of other’s knowledge sharing (free riding), but contribute to the total productivity of the community. ④ usually fails due to lack of critical mass of Editors and in cases where sharing costs are higher than the cost of atomization. 53 From mass to networked media ① de-massification of the media as a result of information overload and technological advancements (Toffler) ② In the mass media the profit-maximizing strategy is to attract attention and not to invest in production quality. ③ Networked media: Never before was possible to create, distribute, promote yourself and get feedback for your music, writings or any other online content 54 Projects Web Goods characteristics Compare with traditional good categories 55 What about the mathematician? 56 supplement 57 Main issue Identify purchasing patterns in Web retail using available public data 58 outline ① Before analysis ② Data ③ Innovations ④ Main results ⑤The Amazon co-purchasing network ⑥Amazon: the book-based multi-store ⑦Triad analysis: Winners in Product wars ⑧Switching to best sellers ⑨Hidden complementarity saves MS 59 Before analysis Why Amazon? the best proxy for Web retail Algorithms: item-based top-N recommendations Analysis? co-purchase directed graph related literature? Computer science (Collaborative filters, Karypis) Economics (market basket analysis, Oestreicher-Singer) Software? Gephi, iGraph, FANMOD, mysql, & xls (old habits!) Data 226,238 items 13,351,147 co-purchase connections The crawler was started with an initial set of 300 items, which were the top 10 selling products in each of the thirty categories. 61 recommendations per item (average) 2nd level added 17,204 items 3rd 208,740 items Innovations Cross-category analysis Broad graph (in dyads at least one from category) Strict (both items from the same category) All recommended items crawled (max 104) Triads analysis Community analysis Main results Amazon has evolved into a book-based multi-store with strong cross-category connections. Top selling products are important in the co-purchase network, acting as hubs, authorities & brokers. Co-purchase links not only manifest complementary consumption, but also switching among competitive products (e.g. Kaspersky -> Norton). competitive products consumed as complements because of the existence of compatibility and compatible products that facilitate their joint consumption. overview ①Amazon: the book-based multi-store ②Triad analysis: Winners in Product wars ③Switching to best sellers ④Hidden complementarity saves MS Questions? 64 Descriptive statistics 65 Amazon: the book-based multi-store Figure shows the interconnections among all different categories of products in the Amazon co-purchase network. A link from category A to category B is added if products from category B are present in the broad network of category A. The link intensity grows with the number of products that are present in that network and the size of a category denotes the number of products that belong to it. The stronger links are from Movies & TV to Books, from Kitchen & Dining to Books, from Toys & Games to Books and from Books back to Movies & TV. The Amazon copurchase network for all item categories 66 hidden complementarity saves MS Fig. 2 shows a part of the Strict software co-purchase network, where different colors indicate different community membership. Different product communities have been identified based on the spin glass community detection algorithm and has been computed by the R package iGraph. It is interesting to observe that seemingly competitive products of Apple and Microsoft are in reality consumed as if they were complementary. Microsoft (nodes with red color) and Apple (nodes with blue color) product communities are “mediated” by compatibility like VMware Fusion, Parallels Desktop and compatible products like Office for Mac.