This study deals with an automated data analysis strategy to pinpoint potential unknown compounds in full scan mass spectrometry (MS) experiments. Three examples of an untargeted metabolomics approach to contaminant analysis are given. By comparing a plant-oil based hormone cocktail to 90 plant oil samples ca. 25 compounds specific to the hormone cocktail could be detected. Five of these compounds were confirmed as steroid hormones. A comparison of a drink water sample from a farm to distillated water showed the presence of contaminants specific to this drink water sample. A grass sample, which was known to give a false positive result in a DR-CALUX bioassay, was unexpectedly shown to contain an abnormal level of chrysene, which was obviously not eliminated during clean-up.