Two-Echelon Open Location Routing Problem (2EOLRP) is a logistics problem model that combines two levels of distribution, namely the main distribution center (depot) and Secondary Distribution Centers (SDCs) which then serve customers. Vehicles available at the depot serve product requests from SDCs, while vehicles available at SDCs serve end customers. This paper presents a comparative analysis of two initial solution construction methods, Random Search (RS) and Nearest Neighbor Search (NNS), combined with a Simulated Annealing (SA) algorithm to solve the 2EOLRP. The quality of the starting solutions plays a key role in enhancing metaheuristic performance, including SA. We evaluate both methods based on their solution quality and computational efficiency. Benchmark datasets adapted from well-known Two-Echelon Location Routing Problem (2ELRP) instances are used for testing. The experimental results demonstrate that NNS generally provides better initial solutions leading to improved final results, while RS offers faster computational times. The findings offer valuable insights into the impact of initialization strategies on the overall performance of SA in two-echelon distribution systems.