Explainable risk, not a feed of headlines.

Processing chain

1. Incrementally ingest global event candidates.
2. Filter low-value noise before long-term storage.
3. Normalize, deduplicate and cluster related event records.
4. Resolve geography, entities and risk taxonomy.
5. Map relevant commodities, supply-chain nodes and companies.
6. Score event severity, confidence, exposure, recency and confirmation.
7. Compare the current state with the previous snapshot.
8. Publish the daily delta and the evidence behind it.

Risk scoring principle

Scores remain decomposable. The system favors explainable weighted components over a single opaque model output.

Correlation policy

Event studies may include correlation, lag correlation, regression and R². Every analytical view must disclose the window, sample count, source and confidence. Correlation is never presented as causation.