MARKETS, MARKET PHASES AND STRUCTURES

Algorithmic selection is creating new Internet-based markets and changing existing ones on a large scale. It can constitute the (economic) core function of Internet-based services, for example in the case of the general search services of Google or Microsoft, and/or it is applied as an ancillary function, for example in e-commerce applications for filtering/ recommendation purposes by Amazon, or for the automated selection of status messages displayed in online social media applications by Facebook.

Altogether, almost all of the most popular and economically successful Internet-based services rely heavily on algorithmic selection in one form or another. Table 19.2 shows the ten most visited websites worldwide in 2014 and their applications based on algorithmic selection as a core and/or ancillary service. Seven of these rely heavily on computational advertising, and four on general search engines. Three websites use algorithmic selection as an ancillary service only (Wikipedia and two online shopping platforms). Further, the dominance of US (seven) and Chinese (three) companies is striking.

Despite variations between different categories, market sizes tend to be high (e.g., search, computational advertising) and growth rates impressive (e.g., music and film streaming) for services and products based on algorithmic selection.

Markets pass through different phases in their life cycles: from experimental and expan­sion phases to maturity, stagnation and decline. Accordingly, they show different market structures, sizes and growth rates, and call for different business strategies and public policies. Based on a review of available market data, a rough appraisal of various types of algorithmic selection by market phases can be given.

Source: Latzer et al.

Table 19.3 Concentration of search engine markets in selected countries, Europe and worldwide (end of 2013)

Note: a.

Sources: ComScore (2013a, 2013b, 2013c, 2014) (Europe, ID, IN, MY, PH, SG, TH, US, VN), Pavliva (2013) (RU), CNZZ (2013) (CN), Schautzer (2013) (JP), AT Internet (2014) (DE, ES, FR, UK), Sullivan (2013) (worldwide).

dating. Altogether, the shares of the top two online dating companies amounted to 64 percent (VanderMey, 2013).

Concentration tendencies are a constituent feature of many of the Internet businesses that offer products and services that operate on algorithmic selection. Many of these can be described as two- or multi-sided platforms, operating on two- or multi-sided markets (Rochet and Tirole, 2003) - a characteristic that has important interrelated economic, business and policy implications. In such cases, for example, the platform acts as an intermediary, as a market maker, between (at least) two demand sides that are interlinked by indirect network effects, which may be one reason for concentration in these markets.

These concentration tendencies can be explained by various industrial economic characteristics such as cost structures, scale and scope economies, direct and indirect network effects. As with traditional media markets, cost structures for markets of serv­ices using algorithmic selection are characterized by considerable economies of scale, resulting from high fixed and sunk costs (e.g., R&D, hardware and software main­tenance), and extremely low marginal cost of additional selection processes (e.g., an additional music recommendation). Hence dominant players enjoy cost advantages, resulting in high market entry barriers due to cost disadvantages of new entrants. A large market size is often necessary to operate efficiently, an issue that is also evident when considering indirect network effects that arise when the number of participants on one side (positively or negatively) affects the number of participants on the other. Usually the participation of one group raises the value of participating for the other group.

This leads to another important characteristic: markets that rely on algorithmic selec­tion are predominantly characterized by quality and innovation competition and less by price competition. Many applications are free of charge for end customers. Hence the perceived quality of a service is particularly important for gaining competitive advan­tage. The quality of service depends, inter alia, on the quality of algorithms, hardware (e.g., server farms) and (input) data (Argenton and Prufer, 2012). Exclusive access to data by service suppliers who create data (e.g., social media companies) results in a strong competitive advantage. These data form an essential input for selection processes, and might lead to exclusive quality improvements on the input side, thus contributing to concentration tendencies.

Moreover, exclusive access to user and usage data of one’s own service results in a competitive advantage for established players and forms a market-entry barrier for newcomers, because they will not be able to offer services of a comparable quality. In contrast to traditional media markets, the quality of services - in essence, the quality of selections - increases with the growing use of a service. The reason is that the results of earlier selections feed back into future selection processes and thus increase their quality. The quality of selections depends, inter alia, on the number of earlier selections, which is why more users and usage result in quality improvements of services. This is true for individual users (by improved personalization/customization of products that also increases users’ switching costs) and all other users as well.

Finally, concentration and market entry barriers are facilitated by considerable economies of scope, resulting from the use of central resources for multiple purposes, in particular of technological know-how - especially on algorithms, of hardware infra­structures and databases. Accordingly, many big players such as Google, Microsoft and Amazon are diversifying and offering a range of different types of algorithmic selection services, thus exploiting economies of scope. Among other things, Google offers search, advertising, aggregation and recommendation, Microsoft is active in search, advertising, surveillance, prognosis and aggregation, IBM in prognosis and surveillance.

Moreover, there is a connection between market phases and market structures. Many algorithmic selection applications are still in the experimental phase or an early expansion phase. These phases are, in general, characterized by high concentration, by temporary monopolies of innovators and early movers. In these early phases, innovators (often US companies in the case of algorithmic selection) also find favorable conditions to export and dominate markets abroad (e.g., Netflix, which uses algorithmic selection as a key part of its business).

19.5