Over the past quarter-century, evaluation researchers have recognized the importance of documenting implementation practices of programs as they are transferred from controlled to real-world settings. As programs become widely disseminated in the general population, there is a tendency for practitioners to alter programs in a manner more conducive to their immediate needs, which may adversely affect program outcomes. The current paper uses findings from an ongoing evaluation of a school-based victimization prevention program to highlight some of the difficulties in maintaining a high degree of fidelity when providing prevention programming in a school-based setting. The results, based on observations of program delivery and program provider descriptions of implementation, allow for the examination of fidelity based on different data collection techniques. (Contains 3 tables and 11 notes.)