The current evidence on ESG is shaped by a range of persistent limitations that hinder both causal inference and generalizability. A primary obstacle is the incomplete and uneven observation of operational data at the granular level required for robust analysis. Ideally, assessing the real impact of ESG adoption would rely on detailed microdata—such as plant-level records of throughput, defect rates, downtime, maintenance activities, workforce composition, and time-stamped incident logs. In practice, such information remains scarce, inconsistently collected, and often locked behind confidentiality barriers. Where available, these datasets typically lack standardization in collection methods and vary widely in quality. Without consistent access to operational data at the line or site level, researchers must depend on aggregated firm-level indicators that conflate multiple processes, technologies, and managerial practices. This aggregation attenuates estimated effects and impedes precise mapping of ESG process improvements to outcomes like yield, learning-curve gains, bottleneck relief, or recovery from disruption. The challenges are compounded in supply networks, where information on upstream risk exposure, input substitutability, switching frictions, and shock propagation is rarely accessible beyond direct suppliers. Multi-tiered supply networks remain largely opaque, forcing researchers to impute network dynamics from partial or public data, which raises concerns about the external validity and causal interpretation of resilience models.