Background: Safewards is a complex psychosocial intervention designed to reduce conflict and containment on inpatient mental health units. There is mounting international evidence of the effectiveness and acceptability of Safewards. However, a significant challenge exists in promising interventions, such as Safewards, being translated into routine practice. The Consolidated Framework for Implementation Research (CFIR) provides a framework through which to understand implementation in complex health service environments. The aim was to inform more effective implementation of Safewards using the CFIR domains and constructs, capitalizing on developing an understanding of variations across wards. Method: Seven Safewards Leads completed the Training and Implementation Diary for 18 wards that opted in to a trial of Safewards. Fidelity Checklist scores were used to categorize low, medium and high implementers of Safewards at the end of the 12-week implementation period. Results: Qualitative data from the diaries were analyzed thematically and coded according to the five CFIR domains which included 39 constructs. Twenty-six constructs across the five domains were highlighted within the data to have acted as a barrier or enabler. Further analysis revealed that six constructs distinguished between low, medium, and high implementing wards. Discussion: Our findings suggest that for implementation of Safewards to succeed, particular attention needs to be paid to engagement of key staff including managers, making training a priority for all ward staff, adequate planning of the process of implementation and creating an environment on each inpatient unit that prioritize and enables Safewards interventions to be undertaken by staff regularly.