Localized surface plasmon resonances (LSPRs) have been extendedly studied in the past decades, due to their excellent ability to condense the light into nanometric length-scale. However, LSPRs often suffer from spectral broadening, a crucial origin of which is fabrication inaccuracies. Unfortunately, there is yet an efficient way to evaluate this impact, rendering the researchers having to rely on rather intensive, time-consuming trial-and-error experimentations. We present in this talk an approach to evaluate the various fabrication contributors to the spectral broadening phenomenon, providing a possibility of efficient optimization in lieu of extended experiments. This will shed light to automated fabrication and optimization, integrated design and improved ab initio modeling.