The variance-based sensitivity analysis can be used as a statistical procedure to evaluate different individual measures to influence the thermal behaviour of machine tools. It provides specific ratios for parameters affecting the variance of the output parameters that are not comparable in any other way. This method is implemented by means of large samples used in a simulation model which reduces computing time and is oriented towards physical objects. Furthermore, conditional and total variances are estimated. The article describes the principles of modelling with discrete elements to integrate all variants into one simulation model, the definition of input and output parameters and sample generation considering the parameter-specific distribution functions and parameter dependences. To evaluate the applied measures, temperatures, temperature differences, heat flow and time required until achieving thermal equilibrium are examined as such, or in relation to the deployed energy. The implementation is demonstrated by means of two examples. One refers to the design, and the other one deals with the operation phase of existing machine tools.