qa-aman/discovery-interview-prep

Discovery Interview Prep

Generate interview scripts grounded in Mom Test principles (Rob Fitzpatrick). Reads personas and current assumptions from specs, produces question banks and 1-page interview guides per persona type.

Workflow

1. Research goal — Ask user: "What do you want to learn from these interviews?"
2. Read personas from your persona files or user research directory
3. Read assumptions — If a spec exists, read the Risks/Assumptions section for assumptions to validate
4. Generate question bank organized by category
5. Generate interview guide — 1-page per persona type
6. Output — Print to conversation or write to file

Mom Test Rules

Core principles (Rob Fitzpatrick):

• Ask about their life, not your idea
• Ask about specifics in the past, not generics or hypotheticals
• Talk less, listen more
• Seek concrete facts, not compliments or opinions

Question Categories

Behavior — What they actually do

• "Walk me through the last time you [relevant activity]..."
• "What happened after [specific event]?"
• "How often does [situation] come up?"
• "Show me how you currently do [task]."

Context — Environment and constraints

• "Where were you when you were doing this?"
• "What else was going on around you?"
• "What tools/devices were you using?"
• "Who else was involved?"

Pain Points — Problems and frustrations

• "What's the hardest part about [activity]?"
• "Tell me about the last time that was frustrating..."
• "What have you tried to solve this? What happened?"
• "How much time/effort does this cost you?"

Workflow — Process and sequence

• "What do you do first when you [start activity]?"
• "After [step], what happens next?"
• "Where does this process break down?"
• "What workarounds have you developed?"

Questions to AVOID

These produce unreliable data:

• "Would you use a feature that does X?" (hypothetical)
• "Do you think X is a good idea?" (opinion, not behavior)
• "Would you pay for X?" (hypothetical)
• "How much would you pay?" (unreliable)
• "Is [problem] important to you?" (leading)
• Any question starting with "Would you..." or "Do you think..."
• "On a scale of 1-10..." (no context, no actionable data)

Output Format

# Interview Guide — {Feature/Research Area}

**Research Goal:** {what we want to learn}
**Date:** {DD-MM-YYYY}
**Duration:** {suggested minutes per interview}

## Assumptions to Validate

| # | Assumption | Source | Question to Ask |
|---|-----------|--------|----------------|
| 1 | {from spec} | {spec section} | {Mom Test question} |

---

## Interview Script: {Persona Type}

### Opening (2 min)
- Introduce yourself, explain purpose
- "We're here to learn from you, not to test you"
- "There are no right or wrong answers"

### Warm-up (3 min)
- {2-3 easy context-setting questions}

### Core Questions (15-20 min)

#### Behavior
1. {question}
2. {question}
3. {question}

#### Pain Points
1. {question}
2. {question}
3. {question}

#### Workflow
1. {question}
2. {question}

### Wrap-up (2 min)
- "Is there anything I didn't ask about that you think is important?"
- "Who else should I talk to about this?"

---

## Questions to Avoid
{list of anti-patterns specific to this research area}

## Observer Notes Template

| Time | Observation | Category | Follow-up? |
|------|------------|----------|-----------|
| | | Behavior/Pain/Context/Workflow | Y/N |

Anti-Patterns

• Don't write leading questions disguised as open ones
• Don't ask hypotheticals — they produce unreliable data
• Don't skip the warm-up — it builds rapport
• Don't combine behavior and opinion questions in one

Quality Checklist

• [ ] Every question is grounded in past behavior, not hypotheticals
• [ ] Questions are open-ended, not yes/no
• [ ] Each persona gets a tailored script
• [ ] Assumptions to validate are linked to spec sections
• [ ] Observer notes template is included
• [ ] No leading questions remain