This paper outlines a new technique to address the paucity of data in determining fatigue life and performance based on reliability concepts. Two new randomized models are presented for estimating the safe life and pS-N curve, by using the standard procedure for statistical analysis and dealing with small sample numbers of incomplete data. The confidence level formulations for the safe and p-S-N curve are also given. The concepts are then applied for the determination of the safe life and p-S-N curve. Two sets of fatigue tests for the safe life and p-S-N curve are conducted to validate the presented method, demonstrating the practical use of the proposed technique.