Summary Modelling fire performance of building fire barriers would allow optimising the design solutions before performing costly fire resistance tests and promote performance‐based fire safety engineering. Numerical heat conduction analysis is widely used for predicting the insulation capability of fire barriers. Heat conduction analysis uses material properties and boundary condition parameters as the input. The uncertainties in these input parameters result in a wide range of possible model outcomes. In this study, the output sensitivity of a heat conduction model to the uncertainties in the input parameters was investigated. The methodology was applied to stone wool core sandwich panels subjected to the ISO 834 standard fire resistance temperature/time curve. Realistic input parameter value distributions were applied based on material property measurements at site and data available in literature. A Monte Carlo approach and a functional analysis were used to analyse the results. Overall, the model is more sensitive to the boundary conditions than to the material thermal properties. Nevertheless, thermal conductivity can be identified as the most important individual input parameter.