In this study we show that following a newly proposed protocol for stable Zn isotope ratio measurements on a ThermoFisher Scientific Neptune Plus MC-ICP-MS significantly improves the long-term external reproducibility down to < ± 0.025‰ 2SD on δ66/64ZnJMC-Lyon for a single static analysis with 90 measurement cycles of 4.2 s integration time each. This protocol includes: (i) stable isotope measurements in medium-mass resolution mode (R > 6500) to prevent the influence of any polyatomic interferences; (ii) the use of a double-spike, which in our case was a previously designed 64Zn-67Zn mixture; (iii) Zn purification with a two-column anion exchange method; (iv) thorough pre-run sample Zn purity checks and signal matching between sample and reference standard solutions; and, (v) pre-run adjustment of sample carrier and desolvating nebulizer system (DNS) gas flows such that the oxide formation rate determined on the 140Ce16O+/140Ce+ ratio remains below 1%. Following these routines, we obtain long-term external reproducibilities of 0.290 ± 0.017‰ 2SD for the IRMM-3702 (n = 145) certified Zn isotope amount reference material (RM) and − 9.974 ± 0.025‰ 2SD for the Alfa Aesar (n = 75) in-house Zn standard solution. The averages of sets of five single measurements of the IRMM-3702 RM and the Alfa Aesar standard solution narrow down the external reproducibility to ± 0.006‰ and ± 0.023‰ (both given as two standard deviations; for pooled data referred to as 2SD*), respectively. Alternatively, five-fold numbers of measurement cycles can reach the same external reproducibility for single measurements. The δ66/64ZnJMC-Lyon compositions of certified natural rock reference materials obtained by means of individual digests and analyses accurately reproduce the literature values, but our data also show that different batches of the BHVO-2 can be isotopically heterogeneous and clearly affect the reproducibility (BIR-1a: 0.247 ± 0.023‰, n = 17; BCR-2: 0.268 ± 0.025‰, n = 12; BHVO-2 #3323: 0.323 ± 0.024‰, n = 6; BHVO-2 #0535: 0.353 ± 0.015, n = 5, all 2SD). In cases where no effect of heterogeneity on reproducibility could be determined, we report the lowest uncertainty for replicate RM digests combined with multiple measurements (JB-2: 0.225 ± 0.012‰ 2SD, n = 3, i = 15). Mixtures between solution digests of the BIR-1a and BHVO-2 in combination with data pooling of five single measurements indeed demonstrate a measurement accuracy of < ± 0.010‰ 2SD*, emphasizing the analytical capability of the proposed method to depict small-scale Zn isotope variations. With the overall analytical improvement, we seek a better discrimination between variable, and potentially overlapping processes causing Zn isotope fractionation in various terrestrial reservoirs.