Weld expulsion is one of the most common welding defects during resistance spot welding (RSW) process especially for high strength steels (HSS). In order to control and eventually eliminate weld expulsion in production, accurate assessment of the expulsion severity should be the first step and is urgently required. Among the existing methods, real-time monitoring of RSW-related process signals has become a promising approach to actualize the online evaluation of weld expulsion. However, the inherent correlation between the process signals and the expulsion intensity is still unclear. In this work, a commonly used process signal, namely the electrode displacement and its instantaneous behavior when expulsion occurs are systematically studied. Based upon experiments with various electrodes and workpieces, a nonlinear relation between the weight of expelled metal and the sudden displacement drop accompanied by the occurrence of weld expulsion is observed, which is mainly influenced by electrode tip geometry but not by material strength or sheet thickness. The intrinsic relationship between this specific signal feature and the magnitude of expulsion is further explored through geometrical analysis, and a modified analytical model for online expulsion evaluation is finally proposed. It is shown that the improved model could be applied to domed electrodes with different tip geometries and varying workpieces ranging from low carbon steel to HSS. The error of expulsion estimation could be limited within ±20.4 mg (±2σ) at a 95% confidence level. This study may contribute to the online control of weld expulsion to the minimum level.