Prompt library
Define a metric before it defines you
Metrics rot at their edges, so the explicit edge rulings with rationale are the core of the document, not the formula. Pre-computing the gaming vectors and pairing each with a guardrail metric is what makes the metric safe to put in an incentive.
Last reviewed July 17, 2026
The prompt
Write a definition document for this metric.
Metric: {{metric}}
How it will be used: {{usage}}
Sections:
1. The sentence: "{{metric}} counts X, measured as Y, over window Z" with every term in X, Y, Z pinned (what counts as active? which timestamp? whose timezone? inclusive of what?).
2. The formula: numerator and denominator with exact filters, in prose and in pseudo-SQL.
3. Edge rulings: 6-10 boundary cases ruled explicitly (trial users? internal accounts? refunds? the day a user converts? re-activations?). Each ruling gets one line of rationale so future debates cite reasons, not power.
4. What it does NOT measure: 3 adjacent things people will assume it covers. Name the metric that covers each instead.
5. Gaming: if a team is targeted on this metric, the 2 cheapest ways to inflate it without creating value, and the guardrail metric that catches each.
6. Ownership and change process: who rules on new edge cases, and where changes are logged, because a metric that changes silently poisons every chart built on it.
The document is done when two people cannot compute different numbers from it in good faith.Run in idaptOpens a new chat with the prompt prefilled. Nothing sends until you press send.
Fill in the variables
| Variable | What it is | Example |
|---|---|---|
| {{metric}} | The metric | weekly active teams |
| {{usage}} | What it will drive | the company north star and a team target |