Uncertainty exists in many aspects of building simulation. A deterministic hygrothermal analysis may not sufficiently give a reliable guidance if a number of input variables are subject to uncertainty. In this paper, a probabilistic-based method was developed to evaluate the hygrothermal performance of building components. The approach accounts for the uncertainties from model inputs and propagates them to the outputs through the simulation model, thus it provides a likelihood of performance risk. Latin hypercube sampling technique, incorporated with correlation structure among the inputs, was applied to generate the random samples that follows the intrinsic relations. The performance of an internally insulated masonry wall was evaluated by applying the proposed approach against different criteria. Thermal performance, condensation and mould growth potential of the renovated wall can overall satisfy the requirements stipulated in multifold standards. The most influential inputs were identified by the standardized regression sensitivity analysis and partial correlation technique. Both methods deliver the same key parameters for the single and time-dependent output variables in the case study. The probabilistic method can provide a comprehensive risk analysis and support the decision-maker and engineer in the design and optimization of building components.