Product Analytics Experiments Skill for ChatGPT: Audit Identity and Observation Windows
Review experiment measurement before comparing results. Check eligibility, event deduplication, identity rules, time windows, and guardrails.
中文Install in ChatGPT
Requires a ChatGPT workspace where you can create skills. This is a text adaptation of the pinned GitHub source.
- Review the inputs and outputs in the pinned .claude/skills/product-analytics-experiments/SKILL.md.
- In ChatGPT, open Plugins, Skills, Create, and Create with editor.
- Keep hypothesis, denominator and readiness checks, using supplied aggregate data in place of project-reading tools.
- Save the skill and try the fictional example in a new chat.
Example task
A fictional booking experiment counts submissions by session in A and by user in B. One arm covers only three days. Write an analysis readiness checklist.What to check
Flag both denominator and window differences. Recompute under shared eligibility, deduplication, and observation rules. Keep the chosen guardrails visible before releasing a comparison.
Original SKILL.md on GitHub · Updated
Frequently asked question
Can session conversion stand in for user conversion?
Recalculate by user when identifiers and definitions support it. Otherwise keep the session label and state what that measure covers.