A metric is a number plus the definition that produced it, and the definition is where the disagreement lives. Two competent analysts, handed the same customer-level revenue file, will report different net revenue retention. Neither made an error. One counted usage revenue and the other did not, or one treated a three-month billing gap as a pause and the other as a churn followed by a win-back. Both readings are defensible. What separates them is whether the choice was stated alongside the number.
This reference exists to make the definition explicit. Each term states one measure, the exact method behind it, the choices that move it, the ways it is commonly misread, and — the section most standards omit — what the measure cannot tell you no matter how carefully it is computed.
How to read a definition here
Every term carries the same spine:
- The standard — a single normative statement. If you read nothing else, read this.
- What it measures — the question the metric answers, and for whom.
- How it is computed — the exact method, in the order the computation happens.
- A worked example — small, arithmetic, checkable by hand.
- The choices that change the number — the forks, each named with the switch that governs it.
- How it is misread — the failure modes that show up in real diligence.
- What it cannot tell you — the honest boundary.
Wherever a definition forks, the standard names the fork with a switch and gives it a
default: reactivation_window_months = 3 means a billing gap of three months or
less is a pause; nrr.usage_inclusive = true means usage revenue counts in
retention. A spreadsheet, a BI tool, or a metrics product exposes the same switches under
the same names. Compute under the defaults and you can say “NRR under
SoF Standard Definitions v1.0” and a reader knows every choice you made. Change a
default, and you cite the version plus the switch you changed.