Fluorescence spectroscopy has numerous applications to characterize natural and human-influenced water bodies regarding dissolved organic matter (DOM) and contamination. Analyzing samples in a timely manner is crucial to gaining valid and reproducible excitation-emission matrices (EEM) but often difficult, specifically in transnational projects with long transport distances. In this study, eight samples of different water sources (tap water, differently polluted rivers, and wastewater treatment plant (WWTP) effluents) were stored under standardized conditions for 59 days and analyzed regularly. With this data set, the sample and fluorescence spectra stability was evaluated. Established analysis methods such as peak picking and fluorescence metrics were compared over time and benchmarked against dissolved organic carbon (DOC) and a maximal change of 10% in terms of their variability. Additional high-performance liquid chromatography (HPLC) data to identify single organic compounds provides insights into these DOM alterations and allows for conclusions about the underlying biological processes. Our results corroborate in a systematic way that the higher the organic or microbial load, the faster the sample must be processed. For all water sources, considerable changes were found between days zero and one, indicating a potential systematic bias between in-situ and laboratory measurements. The absolute signals of individual peaks vary substantially after only a few days. In contrast, relative metrics are robust for a much longer time. For specific metrics, when filtered and stored under cool and dark conditions, tap water may be stored for up to 59 days, non-polluted river water for up to 31–59 days, and WWTP effluents for up to 14–59 days. The storability thus depends both on the specific water source and the analytical plan. By systematizing our understanding of how the specific water source and DOM concentration determine the stability of samples during storage, these conclusions facilitate efforts to establish a standardized protocol.