§3.2 Selecting the main parameters

40. For some services, there may be a large number of ranking parameters. For others, it may be more limited. Especially where there is a large set of parameters, the provider will need to select the most relevant ones (cf. Recital 24). More specifically, pursuant to Article 5(1), relating to online intermediation services, the providers concerned have to set out the ‘main parameters determining ranking’, as well as ‘the reasons for the relative importance of those main parameters’. Pursuant to Article 5(2), relating to online search engines, the providers concerned have to set out the ‘main parameters, which individually or collectively are most significant in determining ranking’, as well as the ‘relative importance of those main parameters’. 41. To determine which set of parameters are ‘determining’ (Article 5(1)) or ‘most significant in determining’ (Article 5(2)) ranking, and are to be considered the ‘main’ ones, providers could consider what drove the design of the algorithm in the first place. For example, whether it was driven by a desire to ensure that consumers found goods or services that were local, cheap, of high quality, etc. This approach reflects one of the objectives of Article 5, which is explained in Recital 24: ‘help business users to improve the presentation of their goods and services, or some inherent characteristics of those goods or services’. As a best practice, this company-internal reflection could be combined with giving thought to what the provider considers is the top type of result on its service. Similarly, the provider could consider what would make the services’ consumers most satisfied. This may relate to what the provider does to serve the interests of the consumers on its service, for example by considering why it chose the filters that are offered to consumers to sort results on its services or other similar techniques. 42. Where a provider identifies a large number of ranking parameters, it could consider dividing them into several broad categories and then consider what are the subcategories or nuances that fall under them. If numerous different categories of parameters are concerned, providers could consider working backwards to eliminate those that only have a peripheral role in determining ranking. This would then leave only the main parameters, which would then need to be categorised and subcategorised as necessary. This could enable the provider to determine and explain what is the ’relative importance’ of the ‘main parameters’ identified. 43. A best practice that can help users understand how ranking functions is to examine the more unexpected elements that determine ranking. These could also be factors that a user may assume are irrelevant as they are unrelated to the quality of the good or service they offer through the service. 44. A best practice is to help users’ understanding by including a high-level description of the company-internal process conducted by the provider concerned that led it to conclude what the ‘main’ parameters are. This could also help providers meet the requirement in Article 5(5) of enabling users to obtain ’an adequate understanding‘ of certain particular factors relating to ranking (see Section 6.2 below). 45. In a similar way, in making the decision on what are the main parameters, providers could consider what would be most useful to know for the users concerned. For business users, the transparency in respect of ranking serves to improve predictability and to help them improve the presentation of their goods and services, or some inherent characteristics of those goods or services. For corporate website users this is to obtain an adequate understanding of whether, and if so how and to what extent, certain design characteristics of the website used, such as their optimisation for display on mobile telecommunications devices, is taken into account. 46. It is important to note that ‘relative importance’ does not require the exact weighting of the main parameters, or indeed the providers’ algorithms, to be disclosed, (cf. Article 5(6) and Recital 27). However, providers should capture to some extent the inherently dynamic nature of (typically algorithm-driven) ranking in a more static description. Temporary changes that may be foreseeable and regular, such as sales promotions or the influence of the weather, could accordingly, as a matter of principle, be described in a general manner, without necessarily triggering a need to adapt the description of ranking each time the temporary change kicks in. Major alterations to the ranking mechanisms leading the ‘main parameters’ to change, including ‘evolutionary’ changes that occur over time as a result of a degree of deep-learning, would trigger the need to adapt the description given (11). 47. At a more general level, the main parameters identified should be a true reflection of what determines ranking under the mechanism in question. What are described as the main parameters should be what is actually most important in determining ranking. Therefore, it is irrelevant whether or not the parameters are factors that can be influenced by users. Similarly, providers cannot just refer to the fact that ranking is determined by algorithms that use artificial intelligence. To understand further the relationship between machine learning and the obligation to disclose the main parameters, see 3.3.12 below.

← 3.1 · All articles · 3.3 →

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