OBJECTIVE Faculty coaching is recognized as an essential element for effective use of portfolios in undergraduate medical education, yet best practices for training these coaches are uncertain. METHODS New portfolio coaches participated in a multifaceted training program that included orienting modules, a 7.5-hr training workshop featuring analysis of reflective writing, an Observed Structured Teaching Exercise (OSTE), and subsequent longitudinal coaches' meetings for timely task training. Four desired coaching skills were emphasized in the initial training: creating a safe environment, explicitly using performance data, asking questions that elicit reflection, and guiding the student to develop future goals and plans. We collected and analyzed several outcomes: (a) coaches' self-assessment at key intervals, (b) open-ended written responses to three coaching vignettes, (c) video recordings of the OSTE, and (d) subsequent student evaluation of the coach. In an attempt to capture learning from the workshop, both the responses to written vignettes and the video-recorded encounters were coded for presence or absence of the four desired skills. BACKGROUND Our portfolio and coaching program was instituted as part of a major undergraduate medical education reform. A new cohort of 25 coaches is enrolled with each matriculating student class, and each coach is assigned to work individually with 8-10 students, forming a coaching relationship that continues over 4 years. Coaches are compensated at 5% full-time equivalent. RESULTS On coach self-assessment, the majority of coaches reported significant improvement in their perceived ability to assess a student's level of reflection, enhance reflection, use performance data, and guide a student to develop goals and plans. After two semesters, coach perception of improved abilities persisted. Students rated coaches as excellent (82%), reporting that coaches created safe environments (99%), promoted insight (92%), and aided in goal setting (97%). Written responses to vignettes before the OSTE found that several coaches omitted desired behaviors; however, posttraining responses showed no discernable pattern of learning. Coding of the OSTE, in contrast, documented that all coaches demonstrated all four of the desired skills. CONCLUSIONS Although coaches reported learning related to key skills, learning was not apparent when responses to written vignettes were examined. In contrast, skills were demonstrated in the OSTE, perhaps due to the added structured tasks as well as anticipation of feedback. In conclusion, this portfolio coach training program achieved its desired aim of providing students with portfolio coaches who demonstrated the desired skills, as reported by both coaches and students.