In France, the number of revisions for total shoulder arthroplasty (TSA) has increased by 29% between 2006 and 2010. Published studies have reported a revision rate of approximately 11% for hemi-arthroplasty and total anatomical implants, and 10% for reversed implants. The decision to revise or not revise a TSA requires that a rigorous, clinical, laboratory and imaging initial assessment be done in order to answer five questions. Is it infected? Is it unstable? Is it worn? Is it loosened? How is the rotator cuff? This assessment and an evaluation of the bone stock are required to decide whether or not to revise. If the problem is infection, the best solution is not always complete removal of the implant, which results in very poor shoulder function. In such a situation, a multidisciplinary consultation is essential in the decision-making. If the problem is instability, the cause must be identified and rectified. Instability is often caused by insufficient restoration of the humerus length. If the problem is loosening, the type of revision must take into account the patient's age, the rotator cuff status and the available bone stock. The possibilities to reimplant an anatomical glenoid are scarce, and only for cases with minor bone loss and an intact cuff. If a bone graft without reimplantation of a glenoid component is preferred, it should be a tricortical graft to resist wear and medialisation. In the other cases, a reversed shoulder implant with an autograft is preferable. Whether or not the humeral stem is loose, it must often be removed. However, its removal is very difficult, risky and it often causes complications, with humerus fracture being the most common. The possibility of reconstruction depends on the quality of the remaining bone stock. In all these risky situations, the patient should be duly informed and should take part in the decision-making process.