This paper focuses on an accident reconstruction methodology by estimating the errors introduced into reconstruction analysis as a result of assumptions made due to lack of data availability and other uncertainties. Mathematical models are used to show the sensitivity of their results, i.e., occupant kinematics, injury predictions, etc., to changes in these assumptions. For demonstration purposes, a real world crash involving an occupant with “no brain injury” was selected from NHTSA’s Crash Injury Research and Engineering Network (CIREN) database and reconstruction was carried out using the information available from the crash. The crash pulse for the case was obtained using Human-Vehicle- Environment (HVE) software and then applied to a MADYMO (Mathematical DYnamic MOdel) occupant simulation model of the case vehicle and occupant. Head acceleration output from the model subsequently served as an input into the NHTSAdeveloped SIMon (Simulated Injury Monitor) finite element (FE) head model and used to compute probabilities of various brain injuries. The results of the SIMon predictions were then compared to the brain injuries reported in CIREN. Sensitivity analysis was carried out at each step with respect to various assumed parameters starting with generation of the collision pulse in HVE and ending with SIMon brain injury predictors. Important parameters required for better injury predictions were also identified, and some observations that may be relevant to the CIREN accident investigation team are made. This paper shows that a “no injury” case can become an “injury” case due to the introduction of variability in reconstruction parameters. This paper thus shows the methodology, including important details to be taken into account as well as the additional information that needs to be collected from the real world crashes for better accident reconstruction analysis.