A retrospective review was performed of all shoulder arthroplasties with patients grouped on the basis of transfusion protocol time period. Group 1 had transfusions if postoperative hematocrit was <30. Group 2 had transfusions based on symptomatic anemia. Bivariate analysis of transfusion factors and multivariate analysis of significant bivariate factors were performed. Protocol change decreased transfusion rates from 16% (group 1, 153 arthroplasties) to 8% (group 2, 149 arthroplasties). Reverse shoulder arthroplasty (RTSA) transfusion rate decreased dramatically (from 24% to 5%). Transfusion rates after total shoulder arthroplasty (TSA) were low (4%) and after revision arthroplasty were high (21% + 27%) in both groups. Age, gender, heart disease, preoperative hematocrit, diagnosis, and estimated blood loss (EBL) were risk factors on bivariate analysis. Failed arthroplasty and fracture diagnoses carried high transfusion rates (25% + 28%). Logistic regression showed that low preoperative hematocrit, increased EBL, revision arthroplasty, and heart disease were transfusion risk factors. Protocol based on symptomatic anemia results in low transfusion rates after primary TSA and RTSA.