Non-target screening via high performance liquid chromatography–mass spectrometry (HPLC–MS) has gained increasingly in importance for monitoring organic trace substances in water resources targeted for the production of drinking water. In this article a new approach for evaluating the data from non-target HPLC–MS screening in water is introduced and its advantages are demonstrated using the supply of drinking water as an example. The crucial difference between this and other approaches is the comparison of samples based on compounds (features) determined by their full scan data. In so doing, we take advantage of the temporal, spatial, or process-based relationships among the samples by applying the set operators, UNION, INTERSECT, and COMPLEMENT to the features of each sample. This approach regards all compounds, detectable by the used analytical method. That is the fundamental meaning of non-target screening, which includes all analytical information from the applied technique for further data evaluation. In the given example, in just one step, all detected features (1729) of a landfill leachate sample could be examined for their relevant influences on water purification respectively drinking water. This study shows that 1721 out of 1729 features were not relevant for the water purification. Only eight features could be determined in the untreated water and three of them were found in the final drinking water after ozonation. In so doing, it was possible to identify 1-adamantylamine as contamination of the landfill in the drinking water at a concentration in the range of 20 ng L−1. To support the identification of relevant compounds and their transformation products, the DAIOS database (Database-Assisted Identification of Organic Substances) was used. This database concept includes some functions such as product ion search to increase the efficiency of the database query after the screening. To identify related transformation products the database function “transformation tree” was used.