On the robustness of R&D