matteotitta Customer interviews skill
Prepare structured interview guides for B2B SaaS customer research. Produces question banks, preparation frameworks, and synthesis templates that feed directly into ICP research and positioning work.
Inputs
| Input | Required | Source |
|---|---|---|
| Interview type | Yes | User specifies: discovery, win/loss, churn, expansion, or testimonial |
| Company context | Recommended | /company-context output or client CLAUDE.md |
| ICP profile | Recommended | /icp-research output |
| Specific research goals | Yes | User states what they're trying to learn |
Interview types
Full question banks (10-15 questions per type) live in the premium reference. Use this table to pick the right type, then pull 8-12 questions from the bank.
| Type | Purpose | When to use |
|---|---|---|
| 1. Discovery | Understand the prospect's world before they became a customer. Map pain, workflow, buying triggers, decision criteria. | Early-stage ICP research, entering new segments, validating positioning hypotheses |
| 2. Win/loss | Understand why deals closed (or didn't). Extract decision criteria, competitive dynamics, buying process details. | After closed-won/closed-lost deals, competitive positioning, sales enablement |
| 3. Churn | Understand why customers left. Identify product gaps, service failures, unmet expectations. | After churn events, retention strategy, product feedback loops |
| 4. Expansion | Understand what drives upsell, cross-sell, seat expansion. Map the internal champion's advocacy process. | Expansion playbook development, pricing research, PLG analysis |
| 5. Testimonial | Extract quotable results, transformation stories, social proof. | Case study production, social proof gathering, website copy refresh |
Preparation framework
Run this checklist before every interview.
1. Research the interviewee
- Role, title, tenure at company
- LinkedIn profile review (career path, posts, interests)
- Their company's stage, size, industry
- Any previous interactions with your team (sales notes, support tickets, NPS scores)
- Public content they've produced (blog posts, conference talks, comments)
2. Define research goals
- What are the 3 specific things you need to learn from this conversation?
- What hypotheses are you testing?
- What gaps in your current ICP/positioning/competitive data does this fill?
- How will you use the output? (Which downstream skill or deliverable?)
3. Customize the question bank
- Select 8-12 questions from the relevant interview type
- Reorder based on priority (most important questions early, in case you run short)
- Personalize with company-specific details ("I noticed you switched from [competitor]...")
- Prepare 2-3 follow-up probes for each key question
- Remove questions you already have answers to
4. Logistics
- Interview length: 30-45 minutes (never over 60)
- Recording permission: ask at the start, not mid-conversation
- Note-taker: assign someone or use transcription (Granola, Otter, etc.)
- Warm-up plan: 2-3 minutes of genuine rapport before questions
During the interview
Principles
- Listen more than you talk. Target 80/20 ratio (them/you).
- Follow the energy. When they lean into a topic, go deeper. Your script is a guide, not a cage.
- Ask "why" and "tell me more" liberally. The best insights come from the second and third layers.
- Use their words, not yours. Don't rephrase their pain into your product language.
- Sit in the silence. When they pause, don't fill the gap. They're often about to say the most honest thing.
- Capture exact quotes. Flag verbatim language in your notes with quotation marks. These are gold for messaging.
Follow-up prompts
Use these when an answer is surface-level:
- "Can you give me a specific example of that?"
- "What happened next?"
- "How did that make you feel?"
- "Who else was affected by that?"
- "What would have happened if you hadn't done anything?"
- "Help me understand the timeline on that."
- "You mentioned [X] -- say more about that?"
Red flags during interviews
- They're giving you the answers they think you want. Redirect with "I really want the honest version."
- They're speaking in generalities. Push for specifics: "When was the last time that happened?"
- They're selling you on their company. Gently steer back: "That's helpful context. Now tell me about the problem side."
- You're asking leading questions. Catch yourself. "Don't you think X is a problem?" becomes "How do you think about X?"
Synthesis framework
After the interview, map findings to these categories. Each one feeds directly into downstream skills.
| Category | Feeds into | What to capture |
|---|---|---|
| Pain points and triggers | icp-research, positioning | Specific problems described; what triggered the search; urgency / cost of inaction |
| Buying process | icp-research, win-loss-analysis | Who was involved; evaluation process; timeline and friction points |
| Decision criteria | positioning, product-messaging | What mattered most; deal-breakers; how alternatives were compared |
| Language and vocabulary | tone-of-voice, messaging | Exact phrases for the problem; how they describe the solution to peers; words they avoid |
| Results and impact | case-study, testimonial content | Quantifiable outcomes; qualitative changes; before/after contrast |
| Objections and concerns | sales-enablement, battlecards | What almost stopped them buying; concerns remaining; what they'd change |
Anti-patterns
Don't do these:
- Asking yes/no questions. Every question should be open-ended.
- Running through all 15 questions in 30 minutes. Pick 8-12 and go deep.
- Recording without permission. Always ask first, and respect a "no."
- Interviewing through a screen share of your product. This is research, not a demo.
- Combining interviews with sales pitches. Keep them separate or you'll poison the data.
- Interviewing only happy customers. Churned users and lost deals are where the real insights live.
- Skipping synthesis. Raw notes decay fast. Synthesize within 24 hours or the context fades.
- Treating one interview as truth. A single data point is an anecdote. You need 5-8 interviews per segment to see patterns.