Global evaluation of Overture Maps reveals source-driven disparities in urban geospatial data

Abstract Open geospatial platforms aggregating data from multiple providers support urban analytics worldwide, yet the quality implications of global-scale conflation remain poorly understood. We assess Overture Maps data quality across two releases (November 2024, April 2026) for 11,422 GHSL-defined urban areas. Point-of-interest coordinate duplication improved from a city-level median of 27.5% to 8.7% between releases, driven mainly by Meta’s internal deduplication. Building attribute completeness is and remains low globally: only 9% of buildings carry any semantic attribute, and height data is absent entirely for nearly three-quarters of cities. Binomial mixed-effects models of building completeness and precision across 320 sampled cities show that source composition is the strongest predictor of data quality. A greater OpenStreetMap share is associated with greater completeness, whereas greater Google and Microsoft shares are associated with lower completeness and precision. The effects are concentrated in the global South regions where machine learning-derived sources currently predominate. Data users should treat the contributing source, rather than release version, as the primary quality diagnostic when selecting Overture data for downstream applications.