Depth-averaged method (DAM) is one of the widely used numerical methods to back analyse the post-failure deposits of submarine landslides due to its high efficiency. However, its simplifications of the velocities along the thickness of the slide cannot capture complex behaviours such as shear band propagation. A novel non-averaged method, material point method (MPM), is used to validate the DAM analysis. The runout distances and morphologies of viscous debris flows predicted by the DAM and MPM are compared with those predicted by experiments and computational fluid dynamics analyses. The ranges of the shear strength, viscosity and sensitivity parameters are investigated to determine the feasibility of the DAM. The conventional DAM algorithm specialised for no-slip bases is enhanced to reproduce the phenomenon of block sliding of slides on frictional bases by considering the stability of the front and rear faces. Then, a spreading of horsts and grabens due to shear band propagation is presented with the MPM analysis. Two real cases of submarine landslides, Southern Mediterranean slide and Finneidfjord slide, were back-analysed with the DAM and MPM.