Convalescent plasma (CP) transfusion has been indicated as a promising therapy in the treatment for other emerging viral infections. However, the quality control of CP and individual variation in patients in different studies make it rather difficult to evaluate the efficacy and risk of CP therapy for coronavirus disease 2019 (COVID-19). We aimed to explore the potential efficacy of CP therapy, and to assess the possible factors associated with its efficacy. We enrolled eight critical or severe COVID-19 patients from four centers. Each patient was transfused with 200-400 mL of CP from seven recovered donors. The primary indicators for clinical efficacy assessment were the changes of clinical symptoms, laboratory parameters, and radiological image after CP transfusion. CP donors had a wide range of antibody levels measured by serology tests which were to some degree correlated with the neutralizing antibody (NAb) level. No adverse events were observed during and after CP transfusion. Following CP transfusion, six out of eight patients showed improved oxygen support status; chest CT indicated varying degrees of absorption of pulmonary lesions in six patients within 8 days; the viral load was decreased to a negative level in five patients who had the previous viremia; other laboratory parameters also tended to improve, including increased lymphocyte counts, decreased C-reactive protein, procalcitonin, and indicators for liver function. The clinical efficacy might be associated with CP transfusion time, transfused dose, and the NAb levels of CP. This study indicated that CP might be a potential therapy for severe patients with COVID-19.