Optimising distribution networks is challenging because of the complex relationships between entities in the network and their conflicting objectives. It is an NP-hard combinatorial optimisation problem, and a simheuristic approach has been proposed to solve these problems. Simheuristics combines simulation with metaheu¬ristics to optimise a problem. However, in real-world networks, metaheuristics and simulation require long computation time. This paper describes correlation as domain-specific information to guide the search of a metaheuristic algorithm to construct promising solutions in order to reduce the number of simulation runs. The approach was tested on a distribution network of an international trading company. The utilisation of a correlation definition randomly biased the selection of solution elements to construct promising solutions.