A significant investigation is needed into the multi-scale spatial differentiation characteristics of ecosystem service value (ESV) and the causes underlying these variations. This is crucial for enhancing our comprehension of the processes governing ecosystem service scale changes. In this study, the Mawei River basin in Guizhou, an important water conservation area in the upper reaches of the Yangtze River, is selected as the research object. We established four grid scale units, namely 500 m×500 m, 1000 m×1000 m, 1500 m×1500 m, and 2000 m×2000 m. We constructed four grid scale units: 500 m×500 m, 1000 m×1000 m, 1500 m×1500 m, and 2000 m×2000 m. A suite of analytical techniques was employed, including the ESV equivalent coefficient method, sensitivity analysis, spatial autocorrelation analysis, optimal parameter geographic detector, and spatial regression modeling. The ESV equivalent coefficient, sensitivity analysis, spatial autocorrelation analysis, optimal parameter geodetector, and spatial regression model were employed to elucidate the spatial differentiation characteristics of ESV in watersheds at varying grid scales and to ascertain the influence mechanism of multiple drivers on ESV. The findings demonstrate that: (1) At the 500-2000 m grid scale, the spatial distribution of ESV within the watershed exhibits both consistency and variation. The overall performance is higher in the west and east and lower in the south and north, forming a high ESV value belt along the watershed. Furthermore, ESV shows a significant positive correlation and spatial agglomeration. As the grid scale increases, the overall characteristics of ESV in the watershed become more prominent, although the fineness is reduced and its spatial correlation and aggregation effect gradually weaken. (2) The spatial heterogeneity of ESV in the watershed is influenced by both natural and socioeconomic factors. Among these, the anthropogenic influence index is consistently the dominant factor affecting the spatial differentiation of ESV in the watershed at different grid scales. Furthermore, the qvalue of interaction with other factors is more than 50%. In comparison to the effect of a single factor, the interaction between driving factors contributes to the spatial differentiation of ESV more prominently. At the 500 m grid scale, the interaction between the anthropogenic influence index and land use type demonstrated a markedly pronounced explanatory power of 78.7%. Conversely, at larger scales, its interaction with elevation exhibited a progressively dominant influence.(3) The influence of each driver on ESV exhibits scale-related differences, and the model demonstrates optimal fit at smaller scales. Across various grid scales, the anthropogenic influence index showed a significant negative effect on ESV. Additionally, the significance and direction of the effect of certain drivers on ESV exhibited some discrepancies. This study aims to furnish a scientific foundation for regulating various drivers to facilitate multi-level management decisions for the Mawei River basin ecosystem. Furthermore, it seeks to promote the precise management of the ecosystem in the upper reaches of the Yangtze River and the coordination of human-land coupling at multiple levels.