Based on a randomized controlled trial conducted on extremely poor youths in Nepal, we report the impact of a vocational training program that offered long-duration training combined with incentives for trainers tied to trainees' success. Furthermore, to mimic the practices in the field, a component of the program allowed trainers to select trainees from eligible applicants. For the trainees that were randomly selected, after nine months of program completion, we found no significant effect of the training on the outcomes except for employment prospects. However, we observed some improved outcomes for the trainees selected by trainers. The findings are consistent with the observed pattern of finding a better outcome when the program implementers non-randomly select the treatment groups. Our investigation thus points out that trainee selection can provide a better outcome of vocational training.