dataUpdated 2026-10-24
Cohort Analysis Prompt
Group by signup period, channel or first action to tell whether retention changed because of product or user mix.
#同期群#留存#增长分析ClaudeGPTDeepSeek
Prompt
You are a growth analyst. Design a cohort analysis for {{product}} to answer {{business question}}.
Output:
1. Grouping dimensions: at least three (signup month, acquisition channel, first-week key action) and what each controls for.
2. Metric definitions: retention numerator/denominator, activation, time window, deduplication.
3. SQL approach: cohort aggregation steps; dialect-specific syntax is optional.
4. Reading rules: how to distinguish product improvement, channel quality change and seasonality.
5. Common traps: survivorship bias, definition changes, small samples, insufficient observation windows.
6. Conclusion template: observation → possible cause → next experiment.
Do not jump to conclusions; provide the analysis plan first and state required fields if data is missing.
Output: analysis plan, SQL pseudocode, reading checklist and conclusion template.Variables to replace
{{产品}}{{业务问题}}
Example output
Groups: signup month × channel × whether the user completed a first project in week one. If retention rises across channels in the new version, it may be product improvement; if only one channel rises, check channel quality first.
How to use it
Cohort analysis is about controlling variables. Ask what else changed at the same time before interpreting the retention curve.