<p>Methods for the chemical characterization of rocks have been rapidly advancing, providing increasingly detailed analyses. However, certain situations pose significant challenges, particularly when there is limited access to material (e.g., non-outcropping units) or when the preservation of samples is required, making destructive analytical techniques unfeasible. The oil and gas reservoirs of the pre-salt interval in Brazil’s eastern margin basins represent a fitting example. To address these challenges, this study evaluates the use of multivariate analysis to portable XRF-derived chemical data, a non-destructive technique limited by lower accuracy in chemical quantification. The aim was to test and validate, through comparisons with geological data, the use of multivariate clustering analysis as a tool to mitigate quantitative limitations and emphasize geochemical trends in portable XRF data, thereby establishing a standard method for pre-processing and processing such data. Two wells (A and B) from the Tupi Field in the pre-salt of the Santos Basin were selected as case studies. Measurements were taken at regular 30&#xa0;cm intervals, generating two distinct datasets for each well: one for oxides and one for chemical elements. After extensive pre-processing, principal component analysis (PCA) and k-Means clustering were applied. The chemical significance of the clusters was interpreted using significant Pearson correlations, followed by geological validation. In well A, five oxide-based clusters (C1 to C5) revealed contrasts in silicification, dolomitization, and depositional chemistry, accurately identifying features such as magnesian clay minerals and volcanic fragments. The elemental data also yielded five clusters (D1 to D5), providing refined insights into depositional variations and detrital input. Well B exhibited similar patterns: five oxide-based clusters (E1 to E5) suggested less diagenetic overprint compared to well A, while four elemental clusters (F1 to F4) included one that uniquely distinguished coquinas from other pre-salt carbonate rocks. These findings confirm the robustness of the proposed method and its effectiveness in extracting geologically relevant information from chemical datasets with limited quantitative precision. Beyond the specific context of the Brazilian pre-salt, this approach offers a replicable, non-destructive, and cost-efficient workflow for geochemical characterization, with potential applications in a variety of studies.</p>