Treatment fidelity assessment is critical to evaluating the extent to which interventions, such as the Good Behavior Game, are implemented as intended and impact student outcomes. The assessment methods by which treatment fidelity data are collected vary, with direct observation being the most popular and widely recommended. The current study presents an exploratory analysis of the differences between fidelity estimates collected through direct observation versus permanent product across three dimensions of treatment fidelity: adherence, quality, and exposure. Findings indicate that direct observation may allow for a more complete estimate of adherence when compared with permanent product, if implementation occurs within the observation window. Quality was only captured via direct observation. Direct observation may not produce adequate estimates of exposure depending on implementation timing. Implications for treatment fidelity assessment are reviewed.