Site differences in implementation trial outcomes are common but often not examined. In a Hybrid Type 1 trial examining the effectiveness-implementation of a peer-led group life-style balance (PGLB) intervention for people with serious mental illness (SMI) in three supportive housing agencies, we found that PGLB recipients' physical health outcomes differed by study sites. The matrixed multiple case study methodology was used to explore how implementation outcomes and changes in context of usual care (UC) services contributed to these site differences. Two implementation outcomes (i.e. PGLB fidelity ratings and intervention recipients' acceptability of PGLB and UC) and changes in healthcare services integration at the study sites were examined. ANOVAs were used to examine site differences in fidelity ratings and client satisfaction. Directed content analysis was used to analyze leadership interviews to identify changes in the context of UC services. Site 3 showed a trend approaching significance (P = .05) towards higher fidelity ratings. High levels of satisfaction with PGLB were reported at all sites. Significant differences in PGLB recipients' satisfaction with UC were found, with Site 3 reporting the lowest levels of satisfaction. Agency leaders reported an increase in prioritizing client's health throughout the trial with sites differing in how these priorities were put into action. Differences in PGLB recipients' satisfaction with UC, and changes in healthcare service integration seemed to have contributed to the site differences in our trial. The matrixed multiple case study methodology is a useful approach to identify implementation outcomes contributing to the heterogeneity of multisite implementation trial results.