# Design Product Experiment

> Turn a product uncertainty into a pre-registered experiment protocol with an estimand, assignment, guardrails, and a decision rule. Use when the result must be able to change a decision. A staged rollout is not automatically an experiment. Do not use to estimate an already-identified observational effect or to run a demo without assignment.

- Skill: `sylphxai/design-product-experiment` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add sylphxai/design-product-experiment`
- Raw SKILL.md: https://api.skillmd.com/api/skills/sylphxai/design-product-experiment/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: SylphxAI (https://skillmd.com/u/sylphxai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/sylphxai/design-product-experiment

---


# Design Product Experiment

A staged rollout is not an experiment. Seal the decision metric and rule before movement. Analyze the assigned cohort. Inconclusive is a legal result; peeking a winner is not.

Open [product experiment systems](references/product-experiment-systems.md) for assignment, guardrails, and readout. Open [counterfactual evaluation](references/counterfactual-evaluation.md) for logged-policy evaluation.

Use `prototype-product` when the cheapest discriminating probe is not a randomized assignment. Use `analyze-causal-inference` when the assignment already happened and must be identified after the fact.