Agent Skills: Discovery Interviews & Surveys

Discovery Interviews & Surveys

Table of Contents

• Workflow
• Common Patterns
• Guardrails
• Quick Reference

Workflow

Copy this checklist and track your progress:

Discovery Research Progress:
- [ ] Step 1: Define research objectives and hypotheses
- [ ] Step 2: Identify target participants
- [ ] Step 3: Choose research method (interviews, surveys, or both)
- [ ] Step 4: Design research instruments
- [ ] Step 5: Conduct research and collect data
- [ ] Step 6: Analyze findings and extract insights

Step 1: Define research objectives

Specify what you're trying to learn, key hypotheses to test, success criteria for research, and decision to be informed. See Common Patterns for typical objectives.

Step 2: Identify target participants

Define participant criteria (demographics, behaviors, firmographics), sample size needed, recruitment strategy, and screening questions. For sampling strategies, see resources/methodology.md.

Step 3: Choose research method

Based on objective and constraints:

• For deep problem discovery (5-15 participants) → Use resources/template.md for in-depth interviews
• For concept testing at scale (50-200+ participants) → Use resources/template.md for quantitative validation
• For JTBD research → Use resources/methodology.md for switch interviews
• For mixed methods → Interviews for discovery, surveys for validation

Step 4: Design research instruments

Create interview guide or survey with bias-avoidance techniques. Use resources/template.md for structure. Avoid leading questions, focus on past behavior, use "show me" requests. For advanced question design, see resources/methodology.md.

Step 5: Conduct research

Execute interviews (record with permission, take notes) or distribute surveys (pilot test first). Use proper techniques (active listening, follow-up probes, silence for thinking). See Guardrails for critical requirements.

Step 6: Analyze findings

For interviews: thematic coding, affinity mapping, quote extraction. For surveys: statistical analysis, cross-tabs, open-end coding. Create insights document with evidence. Self-assess using resources/evaluators/rubric_discovery_interviews_surveys.json. Minimum standard: Average score ≥ 3.5.

Common Patterns

Pattern 1: Problem Discovery Interviews

• Objective: Understand user pain points and current workflows
• Approach: 8-12 in-depth interviews, open-ended questions, focus on past behavior and actual solutions
• Key questions: "Tell me about the last time...", "Walk me through...", "What have you tried?", "How's that working?"
• Output: Problem themes, frequency estimates, current workarounds, willingness to change
• Example: B2B SaaS discovery—interview potential customers about current tools and pain points

Pattern 2: Jobs-to-be-Done Research

• Objective: Identify why users "hire" products and what triggers switching
• Approach: Switch interviews with recent adopters or switchers, focus on timeline and context
• Key questions: "What prompted you to look?", "What alternatives did you consider?", "What almost stopped you?", "What's different now?"
• Output: Hiring triggers, firing triggers, desired outcomes, anxieties, habits
• Example: SaaS churn research—interview recent churners about switch to competitor

Pattern 3: Concept Testing (Qualitative)

• Objective: Test product concepts, positioning, or messaging before launch
• Approach: 10-15 interviews showing concept (mockup, landing page, description), gather reactions
• Key questions: "In your own words, what is this?", "Who is this for?", "What would you use it for?", "How much would you expect to pay?"
• Output: Comprehension score, perceived value, target audience clarity, pricing anchors
• Example: Pre-launch validation—test landing page messaging with target audience

Pattern 4: Survey for Quantitative Validation

• Objective: Validate findings from interviews at scale or prioritize features
• Approach: 100-500 participants, mix of scaled questions (Likert, ranking) and open-ends
• Key questions: Satisfaction scores (CSAT, NPS), feature importance/satisfaction (Kano), usage frequency, demographics
• Output: Statistical significance, segmentation, prioritization (importance vs satisfaction matrix)
• Example: Product roadmap prioritization—survey 500 users on feature importance

Pattern 5: Continuous Discovery

• Objective: Ongoing learning, not one-time project
• Approach: Weekly customer conversations (15-30 min), rotating team members, shared notes
• Key questions: Varies by current focus (new features, onboarding, expansion, retention)
• Output: Continuous insight feed, early problem detection, relationship building
• Example: Product team does 3-5 customer calls weekly, logs insights in shared doc

Guardrails

Key requirements:

1. Avoid leading questions: Phrase questions neutrally rather than telegraphing the "right" answer. Instead of: "Don't you think our UI is confusing?" use: "Walk me through using this feature. What happened?"

2. Focus on past behavior, not hypotheticals: What people did reveals truth; what they say they'd do is often wrong. Instead of: "Would you use this feature?" use: "Tell me about the last time you needed to do X."

3. Use "show me" over "tell me": Actual behavior is more reliable than described behavior. Ask to screen-share, demonstrate current workflow, show artifacts (spreadsheets, tools).

4. Recruit right participants: Screen carefully. Wrong participants waste time. Define inclusion/exclusion criteria and use screening surveys.

5. Sample size appropriate for method: Interviews: 5-15 for themes to emerge. Surveys: 100+ for statistical significance, 30+ per segment if comparing.

6. Seek disconfirming evidence: Actively look for evidence against your hypothesis. If 9/10 interviews support the hypothesis, focus heavily on the 1 that does not.

7. Record and transcribe (with permission): Memory is unreliable. Record interviews, transcribe for analysis. Take notes as backup.

8. Analyze systematically: Use thematic coding, count themes, and present contradictory evidence rather than cherry-picking supportive quotes.

Common pitfalls:

• ❌ Asking "would you" questions: Hypotheticals are unreliable. Focus on "have you", "tell me about when", "show me"
• ❌ Small sample statistical claims: "80% of users want feature X" from 5 interviews is not valid. Interviews = themes, surveys = statistics
• ❌ Selection bias: Interviewing only enthusiasts or only detractors skews results. Recruit diverse sample
• ❌ Ignoring non-verbal cues: Hesitation, confusion, workarounds during "show me" reveal truth beyond words
• ❌ Stopping at surface answers: First answer is often rationalization. Follow up: "Tell me more", "Why did that matter?", "What else?"

Quick Reference

Key resources:

• resources/template.md: Interview guide template, survey template, JTBD question bank, screening questions
• resources/methodology.md: Advanced techniques (JTBD switch interviews, Kano analysis, thematic coding, statistical analysis, continuous discovery)
• resources/evaluators/rubric_discovery_interviews_surveys.json: Quality criteria for research design and execution

Typical workflow time:

• Interview guide design: 1-2 hours
• Conducting 10 interviews: 10-15 hours (including scheduling)
• Analysis and synthesis: 4-8 hours
• Survey design: 2-4 hours
• Survey distribution and collection: 1-2 weeks
• Survey analysis: 2-4 hours

When to escalate:

• Large-scale quantitative studies (1000+ participants)
• Statistical modeling or advanced segmentation
• Longitudinal studies (tracking over time)
• Ethnographic research (observing in natural setting) → Use resources/methodology.md or consider specialist researcher

Inputs required:

• Research objective: What you're trying to learn
• Hypotheses (optional): Specific beliefs to test
• Target persona: Who to interview/survey
• Job-to-be-done (optional): Specific JTBD focus

Outputs produced:

• discovery-interviews-surveys.md: Complete research plan with interview guide or survey, recruitment criteria, analysis plan, and insights template