Reliability improvement problems have much importance in various engineering systems. Real-world engineering systems contain multiple levels. So, a redundancy allocation problem (RAP) is an efficient method to improve the reliability. Different from existing approaches that consider either hot or cold standby redundancy for a component, our approach considers a multi-level cold standby redundancy allocation problem (MLCSRAP). In other words, MLCSRAP is considered for the module and component levels, simultaneously. The units are duplicated by cold standby redundant units. An optimization solution methodology based on a customized memetic algorithm (MA) is presented for obtaining the optimal reliability of configuration multi-level series. The optimal multi-level redundancy allocation provides the improved reliability and lower life cycle cost. The proposed multi-level cold standby redundancy is compared with a multi-level active redundancy. Also, we allocate the cold standby redundancy at all levels using hierarchical genetic algorithm and compare two algorithms together. Finally, the numerical example illustrates how the theoretical results obtained in this paper can be applied.