This talk focuses on several service-level phenomena with duration of therapy (i.e., how much therapy is enough?) providing a starting point. It will show how the original dose-effect model of treatment duration underpinned the promotion of briefer forms of psychological therapy but also how subsequent routine practice-based data has challenged the dose-effect model of treatment duration, proposing an alternative in the form of the GEL (good enough level) model wherein patients leave therapy when they have reached a ‘good enough level’ of improvement. The presentation also shows the intersection of treatment duration and therapist effects – a phenomenon that randomised controlled trials are ill-equipped to accommodate – and the extent of therapist variability. An example is presented of how a single psychological therapies service attempted to reduce such therapist variability without lessening patient outcomes.  However, drawing on a separate large-scale practice-based study utilising data across multiple services, a cautionary tale is noted in the guise of clinical outcomes being differentiated across neighbourhoods and clinics, indicating the need for learning health systems to take account of socio-economic factors at a local level.