§8.3 Experiments

126. If providers use experiments, including A/B testing of changes or evolving fraud fighting mechanisms, they need to consider on a case-by-case basis whether and if so, how this affects ranking on their services and, in particular, whether it requires a change to the description of the main ranking parameters required under Article 5. As a general rule, providers could in this context consider elements such as the frequency and impact of the experiments, as well as the size of the test audience or of the geographic region affected by the testing. 127. In cases where the experiments are ‘live’ (meaning that ‘test audiences’ cannot necessarily be distinguished from the ‘rest’ of the user base), it is possible that the experiments can affect the relative prominence of the goods and services offered by all users. Accordingly, as a first step, providers may have to reflect on whether or not their experimentation is conducted in a genuinely isolated testing environment. If individual experiments are not isolated, providers may have to consider in particular (a) whether the experiment itself has an impact on ranking and if so, whether the parameter being tested is a ‘main parameter’ that needs to be described and; (b) whether and when the full roll-out of the experiment will mean the parameter being tested will change their ranking mechanism and be a ‘main parameter’ that needs to be described.

← 8.2 · All articles · 9 →

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