Water distribution network (WDN) design problems are typically solved using only modeller expertise/engineering judgment or some type of global optimization algorithm. While there are many example problems available where both of these approaches work well, there are other more complicated and high-dimensional problems where both of these approaches may fail. This work highlights one such complicated example where modelling expertise and global optimization are combined to solve a multi-objective WDN expansion problem (http://www.wdsa2012.com/btn.html) involving more than 500 potential decision variables. EPANET2 is used to model the WDN and the efficient Pareto-Archived Dynamically Dimensioned Search Algorithm (PA-DDS) is used for optimization during the multi-stage solution approach. Our first solution to this problem was generated by evaluating a total of 20,000 different designs. In comparison with solution approaches relying mainly global optimization and super computing networks, our solution was non-dominated (of equivalent quality) and utilized orders of magnitude less computation time. This talk will highlight the importance of injecting modeller expertise into the design process and describe ongoing experiments designed to demonstrate that only novice modelling expertise is required to greatly enhance PA-DDS performance. Overall, results demonstrate the enormous difficulty associated with accurately approximating the set of non-dominated solutions in multi-objective WDN design optimization.