Steel-concrete composite walls have been widely used as seismic resistant members, and their ductile behavior under cyclic loads is of great concerns. The ductility factor is a key parameter for evaluating the ductile behavior of lateral resisting members, and it could be obtained from cyclic tests of the composite walls, in which vertical pre-compressive and horizontal cyclic loads are applied. In the conventional method of obtaining ductility factor, the P–Δ effect is commonly involved in the procedure of skeleton curve processing. However, in the structural design software, such as PKPM, YJK, MIDAS and ETABS, the P–Δ effect is regarded as an independent option when setting up the structural model. To avoid repeated consideration of the P–Δ effect, the ductility factor that is input into the structural design software should exclude the P–Δ effect. In this paper, a method for modifying the experimental ductility factor is proposed by considering a specific type of composite wall, i.e. the multi-celled CFST wall (MCFSTW). The proposed method includes two phases: (1) the modification of skeleton curve by excluding the P–Δ effect and rigid-body rotation of foundation; (2) the derivation of yield coefficient α. According to the method, seven skeleton curves measured from cyclic tests of MCFSTWs are modified, and the ductility factors are acquired to be more suitable for practical structural designs. Finally, a refined finite element (FE) model is established. After comparing the modified test curves with the numerical curves, the proposed method is proved to be reliable.