Abstract: In recent decades, numerous deep excavation projects for metro lines and transportation tunnels have been executed in soft clay in urban areas of coastal cities. Collapses of these deep excavations in soft clay are reported from time to time, including the infamous collapses of Nicoll highway excavation for metro circle line in Singapore (2004) and Xianghu excavation for a subway station in Hangzhou (2008). In routine practice, the stability or deformation of an excavation is calculated using the separated approaches, i.e., the limit equilibrium method and the finite element method (FEM), respectively. It is well recognized that the former usually does not consider the effect of excavation width, while the latter usually involves very sophisticated soil models and additional challenges posed by determination of model parameters. These limitations have led to the development of an upper bound method entitled mobilizable strength design (MSD) method by Prof. Bolton in Cambridge University, for predicting stability and deformation of excavations in soft clay in a unified yet simple manner. The authors (Wang & Long, 2014) have recently proposed an improved MSD method (i.e., MMSD method), where a more realistic plastic deformation mechanism is implemented for analyzing the stability of excavations in soft clay. The capability of MMSD for predicting deformation of excavations in soft clay is later verified against the field data of eight case histories (Wang et al., 2018). This study aims to examine the capability of MMSD to predict the collapse of Nicoll highway excavation and Xianghu excavation. It is shown that the MMSD method offers more accurately the prediction for the occurrence of the collapses of the two case histories than the existing limit equilibrium methods (standard method, Hsieh et al's. method and Su et al's. method) and finite element methods, as it accounts for a more realistic deformation mechanism for narrow deep excavations and the strength anisotropy of soft clay.