Background An international six-centre randomised controlled trial comparing individual placement and support (IPS) with usual vocational rehabilitation for people with serious mental illness found IPS to be more effective for all vocational outcomes. Aims To determine which patients with severe mental illness do well in vocational services and which process and service factors are associated with better outcomes. Method Patient characteristics and early process variables were tested as predictors of employment outcomes. Service characteristics were explored as predictors of the effectiveness of IPS. Results Patients with previous work history, fewer met social needs and better relationships with their vocational workers were more likely to obtain employment and work for longer. Remission and swifter service uptake were associated with working more. Having an IPS service closer to the original IPS model was the only service characteristic associated with greater effectiveness. Conclusions The IPS service was found to be more effective for all vocational outcomes. In addition, maintaining high IPS fidelity and targeting relational skills would be a valuable focus for all vocational interventions, leading to improved employment outcomes. Motivation to find work may be decreased by satisfaction with current life circumstances.