Many pre-service music teachers can play a passage before they can explain, model, or teach it to someone else. This exploratory mixed-methods pilot examines that gap in a required group-piano course at a public normal university in mainland China. The course had been redesigned one year earlier, leaving an archive of checkpoint sheets, SMART practice logs, learning diaries, classroom interaction notes, microteaching plans, and end-of-course assessment forms. In Stage 1, two intact second-year cohorts of non-piano-major undergraduates (n = 20) worked through either a redesigned syllabus, organised around a Learning Outcomes-Activities-Evidence matrix and a weekly Listen-Read-Play-Teach cycle, or a conventional demonstration-and-practice routine. Stage 1 is reported descriptively because the archive did not retain medians, interquartile ranges, Mann-Whitney U values, or individual-level distributions, and the per-group sample was too small for responsible inference. Stage 2 re-read process materials from a wider explanatory sample (n = 40), using a codebook focused on listening, reading, playing, peer teaching, and difficulty turning performance knowledge into teachable language. The redesigned class cleared more threshold-approved checkpoint tasks, kept more target-linked practice logs, and wrote microteaching plans that named cues, broke demonstrations into steps, and anticipated likely beginner errors. The comparison class more often recognised problems but stopped before teachable explanation. The study offers a modest example of how visible criteria, recurring teach-back, and a revisitable evidence trail can help make practical performance knowledge more teachable.