<p>Understanding how fruit traits capture provenance structure is essential for traceability and the effective use of native fruit trees in conservation and pre-breeding. We assessed 30 mother trees of <i>Spondias tuberosa</i> (umbuzeiro) from three provenances in Sergipe, Brazil; due to asynchronous fruiting and field-access constraints, ripe fruits were obtained for 23 trees within the sampling window. Basic shape descriptors were computed under a simple, standardized setup, and basic shape descriptors were computed (length, width, aspect ratio, circularity) and CIELAB color indices. Exploratory analyses revealed clear phenotypic structure by provenance, with color channels—especially lightness and the green–red axis—carrying much of the discriminating signal, while simple shape metrics added complementary information. Models based on a compact subset of descriptors distinguished provenances with moderate performance, indicating that low-cost, image-derived features are sufficient to recover geographic structure. Dendrometric traits and fruit-quality measures (total soluble solids, pH, titratable acidity, firmness) provided additional context but did not replace the information captured by endocarp images. We translate these results into a concise descriptor panel and a field-ready workflow that can be implemented in collections and nurseries to support seed-lot traceability, germplasm management, and the selection of mother trees. The findings demonstrate a practical path to phenotyping at scale for conservation and sustainable use of umbuzeiro.</p>