This paper uses the Design Structure Matrix (DSM) to model and simulate the performance of development processes. Though the simulation is a powerful tool for analyzing process performance, its ability is limited by the quality of input information used in the analysis. DSM simulation requires process data that is hard to assess or estimate directly from development participants. In this paper, we propose a methodology that allows a more practical estimation of an important simulation parameter: rework probabilities. Furthermore, we show how does this assessment method (combined with simulation) allow managers to evaluate process improvement plans based on two resulting process measures: reliability and robustness. The method is illustrated with a real application from the automotive industry.