Steam-assisted gravity drainage (SAGD) is a challenging enhanced oil recovery technique that requires proper parameterization in terms of well geometry and steam injection parameters as these parameters are highly dependent on the reservoir geology and fluid characteristics (e.g., net pay variation, water saturation). Injection rate and pressure, temperature, steam quality, lean zone, gas zone, water saturation, and wellbore hydraulics are among the most important engineering parameters of SAGD. This work focuses in the sensitivity analysis under geological uncertainty of some operational parameters, such as: injection rate, temperature, steam quality and pressure. To reach this objective we build a realistic reservoir model with multiple geological settings and descriptions. Moreover, two complementary approaches were studied: i) a deterministic sensitivity analysis relying on a factorial design matrix and ii) a stochastic analysis using particle swarm optimization. There are several key geological uncertainties in SAGD, such as net pay variation, inclined heterolithic stratification (IHS) presence, high water saturation in reservoirs, bottom water issues, and gas cap problems. This work focuses on the role of porosity and permeability in oil production, as these impact these geological properties. To reach this objective, two specific groups of ten different porosity and permeability pairs of geostatistical realizations were generated using Direct Sequence Simulation and Co-simulation. Group 1 used the same variogram model for 10 random seeds; and Group 2 used a combination of values that adopted a simultaneous and a random variation, which were assigned to the seed and to the variogram. We opted by a reservoir model that mimics an anticlinal reservoir, as it proved to be a determining factor for the low levels of Net Present Value (NPV) when compared to other geological scenarios. Within the four most commonly studied parameters in SAGD projects, the association of deterministic versus stochastic multi-objective analysis allowed to accurately validate injection rate as the most relevant parameter. Regarding the geological uncertainties, Group 1 showed that for a certain porosity distribution and permeability orientation, the random origin of these distribution maps eventually attributed a higher NPV uncertain. On the other hand, in Group 2, the simultaneous spatial and seed variation provided less uncertainty associated to NPV. • Sensitivity analysis under geological uncertainty of SAGD operational parameters. • Impact of small-scale variability of petrophysical properties in SAGD performance. • Deeper understanding of relevant injection parameters that influence the NPV and CSOR. • Comparison between deterministic statistical analysis and stochastic optimization.