Summary: In an \(m\)-stage reliability growth test, to estimate the lower confidence limit of the reliability of a system in the final stage, randomization is introduced to improve the existing natural chain method. A simulation study shows that when the number of stages is one, the randomization method is ideal.