The optimization of nanostructures for a specific application is a rather complex task, often akin to shooting in the dark with the hope of hitting the right target, viz. the best performing nanostructure. This difficulty stems from the usually large number of experimental parameters that can be changed in a nanotechnology process. This is the case for example in the fabrication of plasmonic nanostructures for surface enhanced Raman scattering (SERS): the choice of metal, its thickness, the shape of the nanostructures as well as their distribution on the substrate are all parameters that determine the overall Raman enhancement. Even for a relatively simple design, exploring all possible combinations of parameters is a prohibitive task both in terms of nanofabrication time and costs. In that contribution, we focus on the large scale fabrication of plasmonic nanostructures using solid-state dewetting and demonstrate how a full factorial design technique can efficiently optimize the process. In that approach, a target response for optimization and two specific values for the n fabrication parameters are first chosen, leading to an initial set of 2n experiments [1]. Here, as experimental parameters we consider the gold film thickness, the annealing temperature and time, and the heating rate; with the objective of maximizing the intensity of the SERS signal, specifically the 1363 cm-1 vibrational band of Rhodamine 6G. The corresponding 16 experiments provide a model of the dependency of the Raman enhancement on the four experimental parameters considered here. In the classical optimization procedure by full factorial design, this dependency is chosen linear. The set of experimental parameters that is obtained in this first fit of linear model does not necessarily lead to the strongest Raman enhancement and another set of experiments may be required. However, the full factorial approach also provides the direction in the parameter space in which one needs to move these parameters to reach a better solution. After such a solution is found, the linear model may need to be refined using second order functions to clarify the optimum region. This is usually done with the central composite experimental design [2]. The optimization process for the fabricated samples is documented in detail by studying the morphology and optical properties of each sample by scanning electron microscopy and UV-Vis spectroscopy. The gold nanostructures we fabricated this way have typical dimensions in the 50 nm range, which is in good agreement with data reported in the literature [3]. Furthermore, the optimized plasmonic substrate can be used to identify non-resonant bioorganic macromolecules such as bovine serum albumin. This work reveals an effective approach to optimize nanofabrication processes for large-scale and cost-effective production of uniform and highly sensitive SERS substrates.