Sand production is one of the main research topics in the petroleum industry. This problematic phenomenon is related to the mechanical and hydrodynamic behavior and reservoir specifications. Sand production often leads to equipment damage and significant losses. This is usually studied by experimental and numerical methods. This study first gives a brief overview of the methods used to predict sand production in terms of experimental tests, numerical simulation, and field data. The main objective of this research is to highlight the shortcomings of these methods. In addition, experimental hollow cylinder test data and widely used continues numerical simulation are used to investigate these shortcomings. This study is performed by scrutinizing the most important and basic parameters in numerical modeling including two constitutive models (Mohr-Coulomb elastic-perfectly plastic and Mohr-Coulomb cohesion softening/friction hardening failure criteria) and various element sizes and shapes. Since most studies use continuum approaches, it was decided to use a finite difference program. Further, to reduce the undesirable influencing factors the fracture energy regularization method was implemented to diminish mesh dependency related to energy dissipation. In addition, a mesh size sensitivity analysis was performed to show the effects of size, shape and pattern of mesh on the results and cumulative probability distribution versus absolute relative error diagrams were applied to compare the accuracy of the models. After all, the best model of prediction was selected to simulate a real sand production in an oil well with 50° inclination and 6 SPF in North Sea Reservoir. Review of experimental papers shows that these tests usually have several hypotheses for simplification. Rock strength, stress state, fluid flow properties, test duration, and sample dimensions are the most important parameters in sand production tests. Researchers usually focus on only one or two of these parameters, which may lead to many errors in contrast to reality. In addition, numerical methods have some deficiencies such as mesh size problems, lack of critical-state-based constitutive models, sanding criteria, and lack of sufficient research for the calibration procedure in the DEM model. A small change in both the numerical simulation input data and the experimental procedure leads to different results. Indeed, the experimental test of sand production with multiple hypotheses can qualitatively simulate the steps and shape of well or perforation failure not the exact sand production rate. Also, numerical simulation results are very sensitive due to uncertainties and errors of the model and input data. It has been shown that the results are highly dependent on every simple parameter such as the shape of the elements. The differences between experimental results and numerical modeling in only one perforation can be 2 to 10 times. Despite all errors and uncertainties in both the laboratory and modeling studies, the simulation of a real sand production in North Sea Reservoir with same sized element was relatively acceptable. This level of accuracy of the model can be very helpful in subsequent decision making and sand control management.