Most of the real world problems pose practical challenges for making decisions primarily due to availability of limited data. Quantification of risk and assessment of structural reliability becomes difficult in such scenarios. Techniques for performing safety analysis for such problems are discussed in this paper.While complete characterization of a system behavior may be difficult with limited data of its response, statistical models based on extreme value theory provide the basis for making decisions with reasonable confidence. The same may not be true, however, for such structures early in their design cycle due to limited experience of their performance. In such cases response surface methodology can be very useful in determination of risk and suitably making modifications to the design to improve the reliability of the component or system.Applications of these methods for some real world scenarios are demonstrated.