Considering that the traditional method for controller performance assessment is mainly based on minimum variance, which is an unattainable boundary, a new vector-based approach is proposed that is driven by the historical data and is not very dependent on prior knowledge of the process unit. A performance vector is designed to represent the controller precision and its response speed. The length and angle of the vector are constructed using statistical theory and signal processing techniques, which is the key step in the proposed method. It works well for disturbance rejection situation in single-input–single-output (SISO) systems. In particular, an index Pf is introduced as a measurement of the controller performance to show how much potential the control loop will still have for improvement. The Wood–Berry model is used for simulation studies to demonstrate the effectiveness of the proposed method; then, the possibility of extending the method to a multiple-input–multiple-output (MIMO) system is also discussed here.