The robust design method seeks to minimize the sensitivity of performance to uncontrolled variation. Product development frequently uses numerical simulations, analytic models and experimental data. However, these underlying predictions may be inaccurate, and include errors that cause the design to be unacceptable and require a design change. This paper presents a method that analyzes the possibility of a design change based on prediction uncertainty; and then estimates possible changes and evaluates the product design flexibility. The results indicate that small changes in design variables may reduce the likelihood and cost of future design changes.