Mast cell tumor (MCT) is a frequent cutaneous neoplasm in dogs that is heterogeneous in clinical presentation and biological behavior, with a variable potential for recurrence and metastasis. Accurate prediction of clinical outcomes has been challenging. The study objective was to develop a system for classification of canine MCT according to the mortality risk based on individual assessment of clinical, histologic, immunohistochemical, and molecular features. The study included 149 dogs with a histologic diagnosis of cutaneous or subcutaneous MCT. By univariate analysis, MCT metastasis and related death was significantly associated with clinical stage ( P < .0001, r = -0.610), history of tumor recurrence ( P < .0001, r = -0.550), Patnaik ( P < .0001, r = -0.380) and Kiupel grades ( P < .0001, r = -0.500), predominant organization of neoplastic cells ( P < .0001, r = -0.452), mitotic count ( P < .0001, r = -0.325), Ki-67 labeling index ( P < .0001, r = -0.414), KITr pattern ( P = .02, r = 0.207), and c-KIT mutational status ( P < .0001, r = -0.356). By multivariate analysis with Cox proportional hazard model, only 2 features were independent predictors of overall survival: an amendment of the World Health Organization clinical staging system (hazard ratio [95% CI]: 1.824 [1.210-4.481]; P = .01) and a history of tumor recurrence (hazard ratio [95% CI]: 9.250 [2.158-23.268]; P < .001]. From these results, we propose an amendment of the WHO staging system, a method of risk analysis, and a suggested approach to clinical and laboratory evaluation of dogs with cutaneous MCT.