Biventricular resynchronisation is an additional therapeutic option in the management of refractory heart failure, with a functional and haemodynamic benefit as well as an improved morbidity and mortality. However, the rate of non-responsive patients has prompted a re-think about the presumed mechanisms of action for this procedure. This study aims to identify candidates more successfully. Based on five years experience in this centre, our work confirmed a medium and long term clinical benefit with multisite pacing. Nevertheless, there was evidence of a relative discordance between the functional benefit and the haemodynamic impact in terms of ejection fraction achieved with resynchronisation. While QRS narrowing appears to be a predictive factor for a successful procedure, the ECG alone is not sufficient to select 'unsynchronised' candidates. Statistical analysis reveals that before implantation the independent predictive factors to identify non-responsive patients include the presence of a complication of myocardial infarction and a low grade mitral leak. The limits of the ECG suggest a more mechanical than electrical approach to understanding the mechanisms of action for resynchronisation. Its effectiveness in cases of right bundle branch block confirm the hypothesis of left intra-ventricular conduction defects, not apparent on the surface ECG but accessible through new imaging techniques. Based on the hypothesis of delayed movement of the ventricular walls, the principle of resynchronisation aims to restore homogenous contraction. Echocardiography allows observation of electromechanical delay and opens new perspectives in the future for selecting patients for pacing. Ar