Web-based interventions for depression have burgeoned over the past 10 years as researchers and health professionals aim to harness the reach and cost-effectiveness that the internet promises. Despite strong clinical evidence of their effectiveness and policy support, web-based interventions have not become widely used in practice. We explored this translation gap by conducting an implementation pilot of MindBalance, a web-based intervention for depression built on the SilverCloud platform, in three IAPT services. We posed three questions: (1) Who chooses to use MindBalance? (2) Is MindBalance effective for these clients? (3) How do clients use MindBalance? Our results for questions (1) and (2) are commensurate with the positive findings in the literature on patient acceptability and clinical effectiveness for such interventions. Client usage, captured in adherence data as well as usage case-studies, was diverse and differed markedly from face-to-face sessions. The most surprising result, however, concerned the small number of people who were offered the intervention. We reflect upon why this was and discuss implementation issues that primary mental health services should consider when adding a web-based intervention to their services.