This document presents a method for selectively acquiring contextual information in travel recommender systems to improve recommendations. It proposes acquiring only the most important contextual factors for each user-item pair, rather than all available factors. It describes an algorithm called Largest Deviation that calculates relevance scores for factors based on their impact on rating predictions. An evaluation on two datasets found Largest Deviation achieved better prediction accuracy and ranking quality compared to baseline methods, while acquiring conditions for fewer contextual factors. The selective context acquisition approach allows travel recommender systems to provide more personalized recommendations without needing all available contextual information.