Getting Started¶
Start with the confusion you already have. Adopt only what helps.
Small business or single team (a week or two)¶
Typical trigger: products and services, a handful of reports, maybe an AI assistant.
- List 5–10 terms people argue about (gross margin, revenue, active customer…). Discovery
- For each, write down the perspectives that really exist, and who owns each one. Modeling, Authority
- Write a one-page Metric Contract for the numbers leadership looks at most (template), and a Definition Contract for any term people argue about that isn't a number (template). Measurement, Modeling
- Write the answer table: which perspective applies in which situation, and where the answer is "ask". Start from the answer table template. Resolution
- Ask your AI assistant 20 test questions and see whether it asks when it should. Assurance
Mid-size organization (a quarter)¶
Typical trigger: multiple BI tools, conflicting executive numbers, an AI initiative.
- Everything above, for all executive metrics
- Named owners per perspective and a monthly decision slot for conflicts
- A build record for each contract; retire shadow metrics
- Checks run whenever a definition changes
- Change notes with effective dates
Enterprise (ongoing program)¶
Typical trigger: many business areas, semantic layers, regulated reporting, AI agents in production.
- Delegated ownership by business area; a small enterprise forum for defaults and cross-functional conflicts
- Answer rules served to semantic layers and AI agents
- Automated checks in release processes; claim traces for board and regulatory numbers
- SMF practices hosted in existing tools (catalog, semantic layer, governance platform) and connected to existing frameworks. See works-with.md.
Signs it's working¶
- Leaders name the perspective when they quote a number
- AI tools ask instead of guess, and say which perspective they used
- "Which number is right?" meetings become "why is this number what it is?" meetings
- Definition changes stop breaking reports silently