The objective of this study is to establish reliable markers for mortality in children with refractory cardiogenic shock who underwent extracorporeal membrane oxygenation. A retrospective observational cohort study was performed at academic children’s hospital for forty-three consecutive pediatric patients who required veno-arterial extracorporeal membrane oxygenation (ECMO) support with refractory cardiogenic shock from January 2011 to October 2017. 30-day mortality in this cohort was 39.5% (17/43), and successful ECMO weaning rate was 69.8%. Blood lactate was elevated before ECMO implantation and the lactate peak concentration had significant differences between survivors and non-survivors, 8.4 ± 4.3 vs 13.9 ± 6.6 mmol/L. AUC to ROC curve analysis of lactate peak was 0.745 (p < 0.05), and the best cut-off value was 14.2 mmmol/L (sensitivity: 53%, specificity: 92%). The length of lactate level > 5 mmol/L was the most significant connection to 30-day mortality. Its AUC was 0.722 (p < 0.05), and the best cut-off value was 3.3 h (sensitivity: 67%, specificity: 80%). Non-survivors had significantly higher lactate levels during 0–6 h of ECMO support, compared to survivors, which also persisted at 7–12-h, 13–24-h, and 25–48-h ECMO. However, lactate clearance at 12 h, 24 h, 48 h revealed no significant differences between survivors and non-survivors based on 30-day mortality. Lactate peak and the duration of high lactate concentration before ECMO were reliable markers for 30-day mortality of pediatric patients with refractory cardiogenic shock. Static lactate values after ECMO implantation were associated with mortality while dynamic lactate value was not. Ensuring adequate ECMO support after cannulation and early diagnostic and intervention should be implemented to normalize the lactate level.