The availability of multilingual data sets can be pivotal for learning analytics' research that is trying to investigate issues related to the use of multiple languages. Nevertheless, existing data sets store limited information about the linguistic environment in which users interact with TEL applications. This paper tries to identify which data variables and properties make sense in a multilingual analysis context, by examining the case of data coming from a learning portal. More specifically, it analyses the log files of a web portal for organic and sustainable agriculture education, trying to identify whether its linguistic profile (i.e. language of interface, metadata records and learning resources) may affect the number of the users that is attracted and their search behaviour. The paper also includes some generic recommendations related to the information that data sets could store to facilitate multilingual learning analytics.