§3.3.1 Personalisation

49. Personalised ranking of offers is a widely used feature. Although the ranking outcomes can hypothetically differ for each individual consumer, parameters used to personalise results (which can be multiple) normally apply in the same way to all the offers of goods or services by users on the services concerned. In other words, the parameters that personalise results are objectively determined beforehand even if their application results in one consumer seeing a different ranking compared to another consumer using the same online intermediation services or online search engine. 50. As explained in Recital 24, predictability entails that providers determine ranking in a non-arbitrary manner. Where this is a main parameter, an explanation of the use of personalisation, its key features as well as its impact on ranking, including the volatility (i.e. the extent of the difference in ranking for different consumers), (12) could help improve predictability and avoid the risk of ranking being determined in an arbitrary manner. Providers may therefore have to reflect on how and if so, to what extent personalisation affects ranking in the case of their specific online intermediation services or online search engines. 51. This could require that providers analyse the potentially very long list of factors that are used for this personalisation, such as consumers’ personal profiles, interests, search behaviour, their actual geographic location, the time of day the search takes place, their use of cookie blockers or other technical tools and more generally, the wealth of data held on the specific consumer as well as their use of default settings (i.e. their ability to undo default setting or to apply filtering mechanisms (see 3.3.4 below)). 52. Given that certain providers may try to capture unique audiences for unique purposes in order to differentiate their respective business models, (13) they could also consider the composition of the consumers using the specific services when considering the impact of personalisation on ranking. 53. Providers should consider whether and, if so, how consumers can and do use privacy protection settings on their services, given that the use of such settings can affect the possibility to personalise ranking. Illustrative example An online marketplace ‘personalises’ ranking by assigning up to 10 000 predefined ‘features’ to individual consumers. The provider finds that this can be highly effective in ordering its business users’ offers (i.e. in ranking), although the effectiveness of its personalisation mechanism is heavily affected by the privacy settings used by its consumers. Depending on the number, heterogeneity and privacy preferences of consumers active on the service concerned, ranking outcomes can therefore vary greatly. Also, the provider may have designed its ranking mechanism to weigh other factors, such as paid ranking more heavily in cases or at times where personalisation is less effective. In such a situation, the provider may need to include these factual insights in the description of its main parameters determining ranking and the reasons for their relative importance, so as to allow the business users to properly understand the ranking mechanism, without however overburdening or confusing the business users by listing thousands of individual features of the mechanism used.

← 3.3 · All articles · 3.3.2 →

Source: EUR-Lex (Cellar) · retrieved 2026-09-07