A Lightweight Metric Dictionary That Actually Gets Used

TL;DR

• Keep it short, in the model, and owned.
• If it’s hard to maintain, it won’t be used.

The problem

• Metric definitions are scattered across documents.
• Teams don’t trust or consult the source of truth.

Why it matters

• AI needs definitions stored in the model to retrieve them.
• A concise dictionary reduces metric sprawl.

Symptoms

• People ask “What does this mean?” repeatedly.
• New measures appear without documentation.

Root causes

• Dictionary lives outside the model.
• No ownership or review cycle.

What good looks like

• Metric dictionary stored as descriptions and annotations.
• Owners listed and review dates tracked.

How to fix

• Start with top 20 metrics.
• Add definitions to model metadata.
• Assign owners and review quarterly.

Pitfalls

• Over‑engineering the dictionary with too much detail.
• No maintenance schedule.

Checklist

• Definitions stored in model.
• Owners listed for top metrics.
• Quarterly review scheduled.

Framework placement

Primary ARF layer: Semantic Integrity. Diagnostic bridge: semantic-reliability, change-reliability.