Coaching with live observation and immediate performance feedback is an effective means to train paraprofessionals, but might not always be feasible. We used a multiple baseline across participants design with six paraprofessionals who taught elementary students with severe disabilities to test the efficacy of two innovations designed to improve the feasibility of delivering feedback. We found a functional relation between delayed, video-based performance feedback and paraprofessional implementation fidelity of two systematic prompting strategies. Observing a colleague receive feedback did result in some improvement, but did not enable all paraprofessionals to meet the training criterion. These findings suggest that delayed, video-based feedback is an effective and feasible training tool, but only observing a colleague receive feedback might be insufficient. [This is the online first version of an article published in "Focus on Autism and Other Developmental Disabilities."]