This article describes a collaborative and data-based reflective practice study carried out by two language teachers in a Japanese tertiary education context. This research utilized conversational analysis (CA) of audio-recorded classroom data which was subsequently discussed as part of a critical friendship. Both the inclusion of CA data and collegial discussion were in response to recent calls for increasingly data-led and dialogic reflective practice methods. In this study, the two teacher-researchers identified salient issues that they saw as incongruent with their stated teaching beliefs. They then collaboratively determined interventions to deal with the inconsistencies between their beliefs and their actual behavior and repeated the CA data collection and dialogic analysis phases. The two teachers found that the CA data and the opportunity to collaboratively interpret it provided valuable insights into their classroom practice while also offsetting the tendency for excessive self-criticism. This study provides a concrete example of a reflective practice approach that is both theoretically grounded and tied to bottom-up issues that are of interest to teachers.