A series of compressor and cascade test recordings are studied to investigate the off-design loss prediction method. The blade design performance is used to predict the off-design loss changing rate at all operating conditions through analytical derivations and statistical correlation studies. The linear correlation between the incidence and a non-dimensional blade loading factor is the foundation of the prediction method. The off-design incidence is normalized using the off-design blade loading factor for different series of blade designs. An analytical method is introduced to predict the off-design blade loading factor based on design parameters and linear correlation. The changing rate of the off-design loss against the blade loading factor is empirically given through statistical analysis. In application, the prediction method can be used to demonstrate the design space of the off-design incidence for a blade series. The modification of the endwall and the rotor tip loss is recommended to give a more accurate prediction in those regions.