The effectiveness of safety design must be justified in terms of real automobile accidents with human occupants. Since it is far too dangerous to use human subjects in full-scale collision experiments, substitutes are usually studied, that is, experiments performed with dummies and mathematical models. Then we still have to extrapolate the results from the substitutes to the reality. Full-scale collision experiments with human occupants could be performed by reconstructing real automobile accidents accurately. To retrieve the most needed data for further improvements in safety design and to translate actual accidents as controlled experiments, an engineered method was developed to reconstruct accidents. It synthesizes previously validated analytical and experimental knowledges pertinent to automobile collision mechanics. The method was applied to an actual accident, indicating that the methodology is sound and accurate.