---
name: usertold-study-design
description: "Use when turning a product question into a focused UserTold interview: choose a public script, adapt it to the product and participants, validate the Study JSON, and keep activation behind explicit approval."
---

# Design a UserTold Study

Use this skill when a builder, founder, product manager, or coding agent needs a reusable interview script for a concrete product decision.

## Choose the Script

Read https://usertold.ai/guides/research-script-templates and route the question to exactly one primary template:

- customer discovery: https://usertold.ai/guides/customer-discovery-interview-questions
- onboarding or activation usability: https://usertold.ai/guides/find-where-users-get-stuck
- churn or switching: https://usertold.ai/guides/learn-why-users-switch
- pricing comprehension: https://usertold.ai/use-cases/pricing-page-research
- concept testing: https://usertold.ai/guides/concept-testing-interview-questions
- feature requests: https://usertold.ai/guides/identify-user-needs

Ask for the product decision, eligible participants, recent situation, and whether the participant can use the real product. Do not combine several workflows into one Study.

## Adapt the Script

- Preserve the selected template's research mode and evidence boundary.
- Replace examples with the user's real product, task, terminology, and safe stopping point.
- Keep participant instructions outcome-based; do not reveal the correct path.
- Ask about recent behavior before opinions or hypothetical future use.
- For observed tasks, keep Speak → Observe → Talk → Speak and retain a bounded observation plus pre-debrief analysis.
- Never infer participant behavior that was not captured.

Show the complete proposed Study JSON. Validate it with the current UserTold MCP, CLI, or REST contract before saving it.

## Approval Boundary

Drafting, reading, and validation are read-only. Explain material choices and wait for explicit approval before creating or changing a Study. Activating a Study changes participant capture and always requires a separate explicit decision.

## Review Boundary

After interviews complete, inspect the recording, transcript, timeline, and source-linked Evidence. Keep what the participant said, what was observed, and what the analysis inferred distinct. A script does not validate a conclusion by itself.