Language managers in their different forms (language planners, terminologists, professional neologists…) have long tried to intervene in the lexical usage of speakers, with various degrees of success: Some of their lexical items penetrate language use, others do not. To address this issue, previous researchers have often sought to gain a better understanding of lexical implantation phenomena. In this thesis, I assert that the problem language managers face is a typical applied science problem: The goal of their research should not be to know but to do. I suggest that language managers should continuously gather information from the speech community to adapt their strategies on the go and to enhance the effectiveness of their interventions. Taking advantage of the Esperanto speech community, I empirically demonstrate, through a focus group study and a corpus study, how monitoring data on opinionated autonyms could help language managers absorb uncertainty in their deliberate lexical interventions.