AI and Human Validation

Is price data reviewed by a person before it is published?

Yes. Every assessment passes expert human review before it reaches subscribers — automated processing is never the last step.

Validation runs in two layers. Automated cross-referencing checks each figure against multiple official sources and flags inconsistencies. Market experts then review the model outputs and decide whether they are fit to publish; values that do not pass that review are not published. This is the final quality gate before your data is updated each month.

What each layer checks, how the reviewed sample is selected, and what expert human review adds beyond the automated checks are all set out in Two-Layer Validation on our methodology site.

How does automation affect data quality?

Automation does two things for quality. It applies the same validated rules to every assessment every month, which removes the variability and handling errors that manual processing would introduce at scale. And it makes the process reproducible and auditable: the same source data, processed under the same published methodology, produces the same result regardless of who runs it.

That reproducibility is part of what backs our independence — alongside maintaining no commercial relationships with the producers, traders, and other firms active in the markets we cover. Automation and human expertise are not in opposition here: automated pipelines handle extraction and modeling, and expert reviewers validate the outputs before publication.

The full set of independence commitments is documented in What independence commitments back the analysis?, and the validation layers in Two-Layer Validation, both on our methodology site.