Let the design process push and pull you constantly. It is healthy. Design is nonlinear. It goes backwards as much as forward and it curves around back on itself. That is okay. If you have issues with that, this may not be the best profession (although there are engineers that successfully do only computer modeling or code research, so there is still hope).
Above is a bridge truss, the top one may be familiar with, but the bottom one is new and novel. How does one do this improvement? You start with what you know and proceed. You ask questions “how can I improve this boring Pratt truss (the top truss A)?” Well maybe I can study the shear and moment diagrams and adjust the spacing of the verticals such that the axial force in the diagonals is always the same. The cost saving will be significant if the rods, gusset plates, welds, bolts etc are all the same. Then you can proceed from B to C and remove the middle diagonal – (zero shear). This is a new truss, a novel and beautiful truss - an economic, efficient, and elegant truss. One that is created.