Car manufacturers carrying out crash tests at only one speed and with only one occupant size run the risk of sub-optimizing their safety systems. This is discussed occasionally, but often in such all-embracing terms that a car designer is often left without any advice as to how the sub-optimization can be reduced. This risk will be illustrated through an assumed case‘ An existing belt system is compared with some new, hypothetical designs. Depending on which test strategy is chosen, the safety properties of one of the new designs can be found to be either better, or worse, than the existing system. This shows that the consequences of an inadequate test strategy for new safety systems can be, that instead of achieving a reduction in injuries, the result might be an increase in the number of injuries out in the real traffic environment. The illustration is done using a method whereby accident injury statistics can be correlated with dummy responses from crash tests in the laboratory and with dummy responses in the MADYMO Simulation Program. Different collision speeds and variations in occupant size should therefore be considered when test strategies for occupant protection systems are defined. Also, legislators and consumers should pay more attention to safety performance in different circumstances.