A provenance-linked dataset of Cup of Excellence award-winning coffee lots from 2016 to 2025

Abstract Cup of Excellence award records describe selected high-quality coffees, but heterogeneous webpages complicate reproducible comparison. This resource covers 2,711 awarded lots from 102 observed events and 127 award groups in 16 countries and territories during 2016–2025. It retains selected factual labels and values, documented normalization, field-level provenance, missingness and conflicts, while withholding organizer webpage copies and narrative text. A separate common-factor layer distinguishes reported lot attributes from environmental grid context, external experiments, derived quantities and assumptions. Versioned corrections preserve source evidence and identify changes to derived values. Structural checks, source replay within documented access limits, and regeneration of stored computational products assess consistency and reproducibility. No human source-audit observations have been recorded. Reuse remains subject to award-selection bias, score-system differences, incomplete coverage and uncertain identity links. The resource supports inspection and reuse without manufacturing missing measurements or treating external study samples as awarded lots.