Confidence and Data Quality

Definition

Confidence and data quality explain how much Blackridge knows and how much is missing.

Report Fields

Token-economics reports include:

• cost_coverage
• lineage_coverage
• workflow_coverage
• provider_pricing_coverage
• semantic_cluster_coverage
• cache_status_coverage
• confidence_score
• classification
• warnings

Attribution confidence can be exact, strong, inferred, weak, or unknown per dimension.

Why It Matters

Blackridge should not present false precision. Modeled or inferred savings use ranges. Missing pricing is unknown, not zero. Weak attribution is not enough to assign ownership confidently.

Example

If an import has usage tokens but no cost/pricing, the report should count the request as unpriced rather than silently reporting zero spend.

Common Mistakes

• Ignoring warnings in reports.
• Assigning weak or unknown attribution to a team as if exact.
• Comparing reports with different data-quality coverage as if they are equally complete.

Related Docs

• Glossary
• Event Schema
• Report Fields