Aims Currently, there is little information about the need for peri-operative blood transfusion in patients undergoing shoulder arthroplasty. The purpose of this study was to identify the rate of transfusion and its predisposing factors, and to establish a blood conservation strategy. Methods We identified all patients who had undergone shoulder arthroplasty at our hospital between 1 January 2011 and 31 December 2013. The rate of transfusion was determined from the patient's records. While there were exceptions, patients typically underwent transfusion if they had a level of haemoglobin of < 7.5 g/dl if asymptomatic, < 9.0 g/dl if they had a significant cardiac history or symptoms of dizziness or light headedness. Multivariable regression analysis was undertaken to identify predictors of transfusion. High-and low-risk cohorts for transfusion were identified from a receiver operating characteristic (ROC) curve. Results Of 1174 shoulder arthroplasties performed on 1081 patients, 53 cases (4.5%) required transfusion post-operatively. Predictors of blood transfusion were a lower pre-operative haematocrit (p < 0.001) and shoulder arthroplasty undertaken for post-traumatic arthritis (p < 0.001). ROC analysis identified pre-operative haematocrit of 39.6% as a 90% sensitivity cut-off for transfusion. In total 48 of the 436 (11%) shoulder arthroplasties with a preoperative haematocrit < 39.6% needed transfusion compared with five of the 738 (0.70%) shoulder arthroplasties with a haematocrit above this level. Discussion We found that transfusion was needed less frequently than previously described for shoulder arthroplasty. Patients with a pre-operative haematocrit < 39.6% should be advised that there is an increased risk for blood transfusion, while those with a haematocrit above this level are unlikely to require transfusion.