This project aims to apply a data science approach to analyze a wide variety of sleep-aid audio content and user engagement metrics from YouTube / Soundcloud, in order to explore this unique soundscape and identify the patterns most strongly associated with effective sleep induction. The trained model can be embedded in an application (sleep-aid or sleep-tracker) to evaluate the Sleep-aid Score of an audio content or recommend effective potential tracks —even beyond this sound scope— may it be a rainy night, a western classical or a Python lecture.