Same metric name.
Different decision.
Define one ratio metric before it reaches a dashboard or experiment. State the outcome it cannot replace, the exact entity and arithmetic, the time and maturity rules, and three hand-calculated examples. The tool checks internal consistency; it never runs a query or declares the metric useful.
Business and calculation metric contract
| Case | Numerator / denominator | Expected | Calculated | Edge condition |
|---|
Passing arithmetic examples only proves this written contract is self-consistent. It does not prove event quality, query implementation, causal validity, statistical fitness or alignment with the real user/business outcome.
Analytics boundary: all text and examples stay in this browser; the paid pack receives only the redacted specification and tiny numeric hand cases. Do not enter raw user events, personal data, warehouse credentials, customer records or confidential SQL. The tool does not connect to analytics/warehouse systems, execute SQL, validate event quality, calculate an experiment, certify causality, publish a metric or replace product/data approval.
$15/mo Metric definition review membership
Unlock an editable Markdown contract with the outcome/proxy boundary, calculation specification, hand-example reconciliation, version impact checklist and product/data signoff. Re-run it for later definitions; production execution remains elsewhere.
- business outcome and proxy-risk boundary
- entity, numerator, denominator and dedupe contract
- window, time zone, filters, null and maturity rules
- hand-calculated examples and mismatch queue
- downstream-use, version and product/data signoff
Mature metric-library substitutes and the pre-platform contract wedge
PMF pre-judgment 75/100 · definition errors are consequential, but Amplitude, Eppo and Statsig already govern executable metrics