Physical crash test dummies of one type, such as the ES-2 side impact dummy, are not perfectly reproducible due to production and certification tolerances. This means that similar crash tests with various ES2 dummies will have varying results, called experimental scatter. New car designs should be insensitive to this scatter, thus the design is robust. As robustness analyses are very cost effective using numerical simulations, there is a need for a numerical dummy model, which incorporates the experimental scatter data. Such model is called a stochastic dummy model. Therefore, the existing MADYMO ES-2 model was modified to represent a wide range of experimental data, thus creating a stochastic ES2 model. An example of a side impact application shows the use of the stochastic ES-2 model to evaluate design robustness.