A correct description of active sites is a precondition toward reaction mechanism study and catalyst screening. In many cases, high-symmetry and ideally ordered crystalline surfaces are used to represent the reactive sites. However, reaction-induced surface segregation and restructuring weaken the rationality of simply employing the suggested most stable surfaces. This paper describes establishment of the scaling relation between segregation energy and adsorption preference for a fast prediction of segregation over bimetallic surfaces with low computational cost, which helps the construction of more accurate active sites, especially for segregation-prone alloys. Based on this methodology, mechanisms of representative hydrogenation reactions are computationally explored on two surface models of a typical bimetallic alloy, CoCu. Our experiments performed in parallel demonstrate the reactivity of CoCu catalyst to be determined by surface Co/Cu composition, reflecting the limitation of one-and-only surface modeling in the description of complex reactions.