Purpose: This study focused on unpacking the instructional coaching process, addressing key questions about what happens during a coaching session and what coaching elements predict teacher outcomes. Design/methodology/approach: Using coaching observational data, the research examined critical coaching processes described in the literature: coaching practices (observation, feedback, reflective discussion and planning), the coach-teacher relationship, coaching strategies and coaching duration. The study also developed a path model documenting how coaching behaviors predicted teacher instruction. Findings: Results showed that the coach talked more than the teacher and that most coaching time was spent in reflective discussion. The coach-teacher relationship was promoted by building rapport and reciprocal trust, with use of "we" language demonstrating that coach and teacher were working as a partnership. Most common coaching strategies were clarifying and the coach prompting the teacher to attend to teacher or student behaviors. Path model analysis showed that (a) the coach-teacher relationship quality predicted the level of teacher engagement in coaching and their instructional reflection and (b) the quality of coaching strategies predicted the overall quality of the classroom instruction. Originality/value: The study provides empirical evidence about the active ingredients of coaching -- those underlying processes that impact and improve teacher practice.