Riverbank failure is an essential source of river sediment and a major cause of river meandering. The Bank Stability and Toe Erosion Model (BSTEM) is a widely used simulation tool owing to its simplicity and clear presentation of results. However, the three core issues in the simulation by the software, time‐step determination, top‐of‐toe (TT) selection, and the iterative algorithm, still lack an in‐depth discussion to ensure the rationality and stability of the simulation results. In this study, the Jing‐150 cross‐section of the Jingjiang River was chosen as an example to illustrate these three problems. For the first problem, which is the choice of the time step, the simulation results become stable and close to the measured value as the time‐step decreases. After determining a suitable time step, the difference between adopting and modifying the default TT of the BSTEM was analysed. The results show that adopting the default TT causes chaos in the distribution of profile nodes, leading to an overly large bank failure width, whereas the modified TT node distribution is normal and the bank failure width is close to the measured value. For the third problem, the original iteration routine (OR) may cause the output bank profile to be unstable and cause a deviation between the bank width and measured value. The negative impact of unstable bank profiles diminishes as the time‐step decreases. However, the workload for reducing the time step is enormous. Therefore, an improved routine (IR) was proposed. With IR, the relative error between the simulated and measured values of the failure width was within 15% when the SD of the water stage was reduced to 0.2478. A similar result could only be obtained when the SD of the water stage was less than 0.099 for OR. This means that to obtain the same simulation accuracy for the failure width, IR can use a larger time step than OR, thereby reducing the workload of the simulation.