We collected major element, trace element and Nd isotopes of cumulate plagioclase and clinopyroxene in lower crustal gabbros from Hess Deep oceanic crust (~2°15'N, 101°30'W) to investigate the Nd isotopic heterogeneity of melts delivered to a complete section of Hess Deep oceanic crust, accreted at the fast-spreading (133 mm/yr) East Pacific Rise (EPR). These data are presented in Cooper et al. (2025) (https://doi.org/10.1130/G52872.1). Elemental maps of 58 samples were initially obtained prior to selecting a subset of 25 samples for in-situ microanalysis. We targeted the Nd isotope record of cumulate plagioclase and clinopyroxene from lower crustal gabbro samples, representing early crystallisation products of melts delivered to the crust. These samples were collected in several expeditions: Ocean Drilling Program (ODP) Leg 147; RSS James Cook cruise JC21; Integrated Ocean Drilling Program (IODP) Expedition 345 (Site U). Combined, these studies provide the most complete composite section of fast-spreading EPR crust to date (stratigraphic depth of 4350 m to 25 m). In our study, we selected 25 samples for in situ Nd isotope microanalysis, covering the range of mineralogy and textural diversity, and over the full stratigraphic depth. For a comparison to local MORB compositions, we selected a set of 13 upper-crustal sheeted dikes collected on the RSS James Cook cruise JC21. Our data reveal that the mantle is heterogeneous at the scale of melt extraction, and the crystal record from the lower crust shows greater 143Nd/144Nd heterogeneity than the overlying MORB. Hence, Pacific MORBs do not reflect the full heterogeneity of their mantle source, and some aggregation of melts occurs within the crust. Data was collected between 2020 and 2023 by George Cooper, Johan Lissenberg and Max Jansen at Cardiff University, UK, as part of NERC Grant NE/T000317/1:HiDe: A Highly Heterogeneous Depleted Upper Mantle? Mineral isotopic analyses were performed on a Thermo Scientific TRITON Plus at the Vrije Universiteit in Amsterdam. The long-term average and reproducibility (2019–2022) for the JNdi-1 standard is 0.512094 ± 0.000011 2 SD (standard deviation; n = 28) with 1011Ω resistors (used for clinopyroxene) and 0.512105 ± 0.000044 2 SD (n = 45) with four 1013Ω resistors (used for plagioclase). Full methodology can be found within the supplemental Material of Cooper et al. (2025) at https://doi.org/10.1130/GEOL.S.28485770 The DOI is a supplement to https://doi.org/10.1130/G52872.1 Methodology: https://gsapubs.figshare.com/articles/journal_contribution/Supplemental_Material_Crustal_versus_mantle-level_aggregation_of_heterogeneous_melts_at_mid-ocean_ridges/28485770?file=52665137