Guiding Principles¶
Status: v0.5 working draft
These principles hold in any size organization and with any technology.
1. Two things can be true¶
A term can legitimately mean different things to different parts of the business. Product gross margin and service gross margin are both correct. The failure is not the difference; it's the difference staying invisible behind a shared label.
Authoritative does not mean globally singular. Represent each legitimate meaning as a perspective with its own owner, instead of forcing false consensus.
2. Context selects the perspective. When context can't, ask¶
People and AI should use the perspective that fits the situation: a product review, a services review, a board deck. When the situation doesn't settle it, the right answer is a question ("which gross margin?"), never a guess.
3. Meaning is separate from measurement¶
A concept says what something means. A Metric Contract says how a number about it is calculated. Keep them linked, never collapsed.
4. Every meaning has an owner¶
Someone accountable decides what a term means within their scope. Central governance handles cross-functional conflicts and enterprise defaults. Owners closest to the work decide the rest.
5. Similar formulas can hide different meanings¶
Two metrics can share a formula and still measure different things, because the inputs mean different things. Compare meanings, not just calculations.
6. Say what can and can't be compared¶
If two perspectives can't be meaningfully compared or combined, say so explicitly. Averaging product and service margin looks reasonable and usually isn't.
7. Meaning changes; history shouldn't break¶
Definitions change with reorganizations, acquisitions and regulation. Version them, date them, name replacements, and keep past numbers interpretable.
8. Trust, but check¶
Good definitions don't guarantee correct use. Check that reports and AI tools use the intended meaning, and be able to back up consequential numbers.
9. Govern in proportion¶
Not every term needs governance. Start where confusion, reuse, AI use, regulatory exposure or executive visibility make it matter.
10. Work with what exists¶
SMF is technology-independent. Catalogs, semantic layers, metric stores, knowledge graphs and governance platforms are ways to implement it. Other frameworks are partners, not competitors.
11. Adopt in pieces¶
Every practice can be adopted on its own, at any size, and combined as needs grow. No practice requires another, a fixed sequence or a central system.
Practitioner notes¶
- Add: a real (anonymized) moment where two leaders meant different things by the same metric.