This paper provides a review of the modelling of geogrid ultimate pullout resistance. Several analytical models were found to disregard the effect of the geogrid stiffness as well as the particle size distribution of the soil. A sensitivity analysis was done on 289 pullout tests to evaluate the significance of these omitted variables as well as other key variables contributing to pullout resistance. It was found that ultimate pullout resistance was the most sensitive to normal stress followed by geogrid length, geogrid stiffness and the friction angle of the soil. By considering geogrid geometry, stiffness and soil grading descriptors, in addition to the standard variables of length, stress and friction angle, regression models of ultimate pullout capacity improved by 15%. In addition, an alternative model to the simplified linear models of FHWA and EBGEO which maintains dimensional consistency while incorporating geogrid stiffness and non-linearity in the model, was proposed. This alternative model performed 32% better than the simplified linear models.