Section: New Results

Network Economics

Participants : Bruno Tuffin, Pierre L'Ecuyer.

The general field of network economics, analyzing the relationships between all acts of the digital economy, has been an important subject for years in the team. The whole problem of network economics, from theory to practice, describing all issues and challenges, is described in our book [67] .

Among the topics we have particularly focused on, the network neutrality debate was a major concern in 2014. In the position paper [79] , Bruno Tuffin and his co-author Patrick Maillé discuss for a large audience the issues and challenges of network neutrality in response to the European parliament text voted in April 2014. A related (and often forgotten) issue, the recently raised search neutrality debate questions the ranking methods implemented by search engines: when a search is performed, do they (or should they) display the web pages ordered according to the quality-of-experience (relevance) of the content? In [22] , we analyze that question in a setting when content is offered for free, content providers making revenue through advertising. For content providers, determining the amount of advertising to add to their content is a crucial strategic decision. Modeling the trade-off between the revenue per visit and the attractiveness, we investigate the interactions among competing content providers as a non-cooperative game, and consider the equilibrium situations to compare the different ranking policies. Our results indicate that when the search engine is not involved with any high-quality content provider, then it is in its best interest to implement a neutral ranking, which also maximizes user perceived quality-of-experience and favors innovation. On the other hand, if the search engine controls some high-quality content, then favoring it in its ranking and adding more advertisement yields a larger revenue. This is not necessarily at the expense of user perceived quality, but drastically reduces the advertising revenues of the other content providers, hence reducing their chances to innovate.

But while ISPs and search engines are almost the only Internet actors being pointed out as potentially non neutral, we investigate the economic impact and strategies of Content Delivery Networks (CDNs), Internet actors that reduce the capacity needs in the backbone network and improve the quality perceived by users. The growing importance of Content Delivery Network (CDN) in the value chain of content delivery raises concerns about the neutrality of these players. We consider in [52] the so-called push and pull models where the traffic is paid by the sender or the receiver, respectively, as well as the situation where the CDN is (vertically) integrated to, i.e., owned by, an Internet Service Provider (ISP). We then discuss the implication of CDNs into the network neutrality debate, another issue forgotten by researchers and regulators. We also propose in [53] a model to analyze the impact of revenue-oriented CDN management policies on the fairness of the competition among two content providers that use CDN services to deliver contents. We show that there exists a unique optimal revenue maximizing policy for a CDN actor –the dimensioning and allocation of its storage capacity– that depends on prices for service/transport/storage, and on the distribution of content popularity. Using data from the analysis of traces from two major content providers (YouTube Live and justin.tv), we remark that a CDN remains a relatively neutral actor even when one of the content providers it serves tries to monopolize the CDN storage space by implementing an aggressive policy to harm its competitors.

Finally, when a customer searches for a keyword at a classified ads website, at an online retailer, or at a search engine (SE), the platform has exponentially many choices in how to sort the output to the query. The two extremes are (a) to consider a ranking based on relevance only, which attracts more customers in the long run because of perceived quality, and (b) to consider a ranking based on the expected revenue to be generated by immediate conversions, which maximizes short-term revenue. Typically, these two objectives are not perfectly positively correlated and hence the main question is what middle ground between them should be chosen. We introduce in [78] stochastic models and propose effective solution methods that can be used to optimize the ranking considering long-term revenues. A key feature of our model is that customers are quality-sensitive and are attracted to the platform or driven away depending on the average relevance of the output. The proposed methods are of crucial importance in e-business and encompass: (i) classified ad websites which can favor paid ads by ranking them higher, (ii) online retailers which can rank products they sell according to buyers' interests and/or the margins these products have, (iii) SEs which can position the content that they serve higher in the output page than third-party content to keep users in their platforms for longer and earn more. This goes in detriment of just offering rankings based on relevance only and is directly linked to the current search neutrality debate.