Clinicians face everyday the complexity of depression. Available pharmacotherapies and psychotherapies improve patients suffering in a large part of subjects, however up to half of patients do not respond to treatment. Clinicians may forecast to a good extent if a given patient will respond or not, based on a number of data and sensations that emerge from face to face assessment. Conversely, clinical predictors of non response emerging from literature are largely unsatisfactory. Here we try to fill this gap, suggesting a comprehensive assessment of patients that may overcome the limitation of standardized assessments and detecting the factors that plausibly contribute to so marked differences in depressive disorders outcome. For this aim we present and discuss two clinical cases. Mr. A was an industrial manager who came to psychiatric evaluation with a severe depressive episode. His employment was demanding and the depressive episode undermined his capacity to manage it. Based on standardized assessment, Mr. A condition appeared severe and potentially dramatic. Mrs. B was a housewife who came to psychiatric evaluation with a moderate depressive episode. Literature predictors would suggest Mrs. B state as associated with a more favourable outcome. However the clinician impression was not converging with the standardized assessment and in fact the outcome will reverse the prediction based on the initial formal standard evaluation. Although the present report is based on two clinical cases and no generalizability is possible, a more detailed analysis of personality, temperament, defense mechanisms, self esteem, intelligence and social adjustment may allow to formalize the clinical impressions used by clinicians for biologic and pharmacologic studies.