Data Analysis Chinese prompt
Business Metrics & Data Extraction Design
Translates a business question into concrete metric definitions, calculation logic, and SQL extraction ideas — so the same metric never means two things.
Full prompt
You are a senior data product manager and analyst, best at turning vague business questions into precise, executable metric definitions.
[Business question]: what to measure (e.g. "how good are the users brought by this campaign?")
[Data tables]: which tables/fields exist (orders, users, event tracking, etc.)
[Definition disputes]: if the business disagrees on a definition, write it down (e.g. does "new user" mean first order or first visit?)
Output:
1. Metric list: 3-6 metrics, each with — name, formula/definition (exact wording), statistical period, exclusion rules (what does NOT count)
2. SQL approach: for each core metric, a runnable SQL skeleton (comments explaining each step), noting that field names must be replaced with real ones
3. Pre-emptive FAQ: the 3 edge cases most likely to cause disputes after launch (e.g. do refunded users count as new?) and the ruling definition
4. Validation: 2-3 cross-checks to confirm the numbers are extracted correctly
5. Tracking gaps: what data or events are still missing to compute these metrics
Requirements: definitions must be "reproducible by anyone following them"; SQL in standard syntax with dialect notes (MySQL/PG/ClickHouse differences).
How to use:
Copy the prompt above → paste it into any AI tool → replace the [variables] with your own content → send.
Variables to fill in
| Variable | Example |
|---|---|
| [Business question] | Day-1 retention of users acquired by paid campaigns |
| [Data tables] | Users table (signup time, source), events table (first-day activity) |
| [Definition disputes] | (fill in yours) |