Lineage: Tracing a Number Back to Its Sources

TL;DR

• Lineage shows where a number comes from.
• AI explanations rely on clear lineage.

The problem

• Users can’t tell which tables or transformations produced a number.
• AI answers lack auditability.

Why it matters

• Lineage builds trust and enables debugging.
• It supports governance and compliance.

Symptoms

• Analysts spend time tracing calculations manually.
• AI explanations are vague or untraceable.

Root causes

• No metadata on measure lineage.
• Complex transformations without documentation.

What good looks like

• Measures list source tables and columns.
• Lineage is visible in reports and AI responses.

How to fix

• Add lineage annotations to key measures.
• Document transformation steps.
• Expose lineage in answer summaries.

Pitfalls

• Relying on external documentation only.
• Skipping lineage for derived measures.

Checklist

• Lineage documented for top KPIs.
• Transformation steps captured.
• Lineage surfaced to users.

Framework placement

Primary ARF layer: Analytical Explainability. Diagnostic bridge: semantic-reliability, execution-reliability, change-reliability.