Chapter 3: The Craft of Interviewing

Interviewing looks easy. Two people talking. How hard can it be?

Very hard, as it turns out. A skilled interviewer can extract honest, detailed, actionable insight from a 30-minute conversation. An unskilled interviewer can spend the same 30 minutes collecting compliments, confirming biases, and leaving with a notebook full of noise. The difference is technique, and technique can be learned.

Interview Preparation

The List of Three

Fitzpatrick recommends walking into every conversation with exactly three things you most want to learn ^mom-test. Not three questions to ask — three learning goals. The distinction matters because a learning goal ("understand how they currently track expenses") generates flexible, adaptive questioning, while a scripted question ("Do you use Excel for tracking expenses?") narrows the conversation prematurely.

Three is the right number because it is small enough to hold in your head without a script, but specific enough to keep the conversation productive. If you cannot articulate three things you want to learn, you are not ready for the interview.

Field Guides

Lombardo and Bilgen describe the field guide as the interviewer's working document ^prr. It is not a script — it is a structured set of themes, prompts, and follow-up paths organized around the research questions. A good field guide includes:

• Research objectives (what you need to learn, not what you will ask)
• Opening prompt (the first question that sets the conversational frame)
• Core themes with 2-3 prompt variations each
• Follow-up probes for when a participant says something interesting but vague
• Transition phrases for moving between topics naturally
• Closing questions (often: "Is there anything else I should have asked?")

The field guide should fit on one page. If it requires flipping through multiple pages, it is too long and will make the interviewer seem disengaged.

Discussion Guides for Structured Contexts

Aulet emphasizes more structured discussion guides for Primary Market Research in entrepreneurial contexts ^de. These are particularly useful when multiple team members will be conducting interviews and consistency across sessions matters. The guide specifies the exact sequence of topics, approximate time per section, and specific questions — but still leaves room for follow-up and improvisation.

The key difference: a field guide is a compass; a discussion guide is a map. Use field guides for exploratory research where you want maximum flexibility. Use discussion guides when you need comparability across sessions.

See Interview Guide Templates for ready-to-adapt examples.

Opening the Conversation

The first two minutes of an interview determine whether you get honest data or polite fictions. The opening sets the frame: is this a sales pitch, a survey, or a genuine conversation?

The VFWPA Framework

Fitzpatrick offers a framework for structuring casual-but-purposeful meetings, especially when the interview is embedded in a broader relationship (a customer meeting, a networking conversation, a coffee chat) ^mom-test:

• V — Vision: Share a brief, genuine statement about what you are trying to figure out (not what you are building). "We're trying to understand how teams coordinate when they're working across time zones."
• F — Framing: Explain where you are in the process. "We're early — still figuring out if this is even a real problem worth solving."
• W — Weakness: Show vulnerability. "We don't have all the answers, which is why conversations like this are so valuable."
• P — Pedestal: Put them on a pedestal of expertise. "You've been managing distributed teams for years — you know way more about this than we do."
• A — Ask: Make a specific ask. "Could you walk me through how your team handled the last project that involved cross-timezone coordination?"

The VFWPA framework works because it establishes you as a learner, not a seller, and the participant as an expert, not an evaluator.

The Three-Box Structure

Hall describes a simpler three-part structure for more formal research interviews ^just-enough:

1. Warm-up (5 minutes): Easy, non-threatening questions about the person's background and context. This builds rapport and calibrates the conversational tone.
2. Core questions (20 minutes): The substantive exploration, organized around your learning goals. Start broad and narrow as themes emerge.
3. Wrap-up (5 minutes): Open-ended closing ("Anything else?"), logistics, and thanks. The wrap-up often produces the best data, because the participant has relaxed and may share things they were holding back.

Question Patterns That Work

Temporal Prompts

The most reliable technique for getting concrete data is the temporal prompt — any question that anchors the conversation in a specific moment in time ^{mom-test cdh}:

• "Tell me about the last time you..."
• "Walk me through what happened when..."
• "Think back to [specific event]. What did you do first?"
• "When was the most recent time you felt frustrated by...?"

Temporal prompts work because they shift the conversation from abstract beliefs ("I think project management is hard") to concrete episodes ("Last Thursday I spent 45 minutes trying to figure out who was supposed to review the design"). The episode contains details — tools, workarounds, emotions, outcomes — that abstract statements never do.

Story Excavation

Torres describes a deeper technique: once a participant begins telling a story, you excavate it by asking for more granularity at each step ^cdh:

Participant: "I was trying to update the project timeline and it was really frustrating."

Poor follow-up: "Why was it frustrating?" (Invites the left-brain interpreter to rationalize.)

Good follow-up sequence:

• "What tool were you in when that happened?"
• "What were you specifically trying to change?"
• "What happened when you tried to make the change?"
• "What did you do next?"
• "How long did the whole thing take?"

Each question asks for a concrete behavioral detail. The frustration reveals itself through the story, not through the participant's interpretation of the story.

Other Effective Patterns

• "How are you dealing with that today?" — Reveals current workarounds, which are the strongest signal of unmet needs.
• "What else have you tried?" — Maps the solution landscape from the customer's perspective.
• "What was the hardest part?" — Identifies the peak moment of friction.
• "If you could have solved this instantly, what would you have done with the time you saved?" — Reveals the real stakes and the customer's priorities.

Question Patterns to Avoid

Leading Questions

"Don't you think it would be easier if...?" "Wouldn't it be nice to have...?" Leading questions telegraph the expected answer. The participant will almost always agree.

Future Hypotheticals

"Would you use a product that...?" "How much would you pay for...?" Future-tense questions about behavior are speculation, not data ^mom-test. People cannot reliably predict what they will do, and they dramatically overestimate their willingness to adopt new behaviors.

Opinion-Seeking

"What do you think about...?" "How do you feel about...?" Opinions are cheap, unstable, and heavily influenced by context. A person's opinion about productivity tools varies depending on whether they just had a productive morning or a frustrating one. Behavior is more stable and more honest.

"Would You" Questions

"Would you buy this?" is the most common question in customer discovery and the least useful. The answer is always yes, because saying yes costs nothing and feels supportive. Replace it with: "You mentioned you spent $200 on X last month. What made that worth the money?"

Handling Emotional Signals

Emotions are data. When a participant's voice changes — when they get animated, frustrated, embarrassed, or excited — you have found something important ^mom-test.

Anger or frustration: "It sounds like that was really frustrating. Can you tell me more about what was happening in that moment?" Anger signals a problem that the person cares deeply about. Dig in.

Embarrassment: "A lot of people tell us they do the same thing — it's really common." Normalize the behavior, then ask for details. Embarrassment often surrounds workarounds or failures that reveal the most important unmet needs.

Excitement: "You seem really energized about that. What would it mean for you if that problem was solved?" Excitement about a problem (not about your solution) is one of the strongest buying signals.

Indifference: Equally valuable. If someone describes a problem you expected them to care about and their tone is flat, the problem may not be as acute as you assumed. Do not try to convince them it should matter.

Handling Feature Requests

When a customer says "You should add X", resist two urges: the urge to agree and add it to your roadmap, and the urge to explain why you cannot.

Instead, excavate the request ^mom-test:

1. "Why do you want that?" — Surfaces the underlying need.
2. "What would that let you do that you can't do today?" — Reveals the job the feature would be hired for.
3. "How are you dealing with this right now?" — Uncovers current workarounds.
4. "How often does this come up?" — Calibrates urgency.
5. "What happens when you can't do it?" — Measures the cost of the status quo.

The answers to these questions are worth ten feature requests, because they reveal the problem — and problems are durable, while feature ideas are not.

Deflecting Compliments and Recovering from Pitch Mode

If you accidentally slip into pitching your idea and the compliments start flowing, you need to reset the conversation ^mom-test.

Recognize the signs: The participant is nodding, saying "That's great," asking when it launches, telling you how much they love it. These are all forms of social support, not purchase intent.

Reset the frame: "I appreciate that — but honestly, I'm more interested in understanding your current situation. Can I ask you about how you handled [specific relevant situation] recently?"

Prevent recurrence: If you notice yourself pitching, stop. Take a breath. Ask a question about their past behavior. The moment you are explaining your product, you have stopped learning.

Active Listening Techniques

Silence

The most underused interviewing technique is saying nothing ^just-enough. After a participant finishes speaking, wait 3-5 seconds before responding. The silence feels uncomfortable to you; to the participant, it is an invitation to continue. The additional information they share during these pauses is often the most valuable.

"Tell Me More"

Three words that unlock more data than any clever question. When a participant says something interesting, resist the urge to ask a specific follow-up. Try "Tell me more about that" first. It gives the participant control over what they elaborate on, which reveals what they think is important — not what you think is important.

Minimal Encouragers

Small verbal cues — "mm-hmm," "I see," a nod — signal that you are listening without steering the conversation. They keep the participant talking without introducing bias.

Reflecting Back

"So what I'm hearing is..." followed by a paraphrase of what the participant said. This technique serves two purposes: it confirms your understanding, and it gives the participant a chance to correct you. Often the correction is more insightful than the original statement.

Note-Taking Systems

Interview Snapshots

Torres recommends creating an interview snapshot immediately after each session ^cdh — a one-page summary that captures:

• Key quotes (verbatim when possible)
• Opportunities identified (unmet needs, pain points, desires)
• Insights (what you learned that you did not know before)
• Surprises (anything that contradicted your assumptions)

The snapshot is completed within 15 minutes of the interview ending, while memory is fresh. It becomes the canonical record that the team references during synthesis.

The Symbol System

Fitzpatrick suggests using a simple notation system during the interview itself ^mom-test:

• :) — Positive emotional signal (excitement, enthusiasm about a problem)
• :( — Negative emotional signal (frustration, anger, disappointment)
• $ — Mention of money, budget, or willingness to pay
• ! — Something surprising or contradicting your assumptions
• ☆ — A specific feature request or product idea (to excavate later)

The goal is to capture the type of signal in real-time without disrupting the flow of conversation. Detailed notes come later, in the snapshot.

The Researcher-Notetaker Pair

Lombardo and Bilgen advocate for a two-person interview team ^prr: one person leads the conversation while the other takes detailed notes. This division of labor is important because it is nearly impossible to listen deeply, formulate follow-up questions, and write comprehensive notes simultaneously.

The notetaker captures near-verbatim quotes, timestamps for key moments, and observations about body language and tone. The interviewer maintains eye contact and presence. After the session, both debrief together.

The Five-Act Interview for Prototype Testing

Jake Knapp's Sprint methodology introduces a specialized interview format for evaluating prototypes ^sprint. It is structured in five acts:

Act 1 — Friendly Welcome (5 minutes): Build rapport. Explain the format. Make it clear that you are testing the product, not the person, and that honest reactions help.

Act 2 — Context Questions (10 minutes): Open-ended questions about the participant's relevant background and current behavior. This is standard discovery — learn about their world before showing them anything.

Act 3 — Introduce the Prototype (2 minutes): Brief setup. Provide only enough context for the participant to understand the scenario. Avoid explaining how the prototype works.

Act 4 — Tasks and Reactions (15 minutes): The participant interacts with the prototype while thinking aloud. The interviewer asks: "What are you thinking? What would you do next? What do you expect to happen?" Resist the urge to help or explain.

Act 5 — Debrief (5 minutes): Summarize what you observed and ask the participant to reflect. "Of everything you saw today, what stood out?" "How does this compare to what you currently use?"

The Five-Act structure works because it separates discovery (Acts 1-2) from evaluation (Acts 3-5). You learn about the person's context before you contaminate their thinking with your solution.

See Sprint Interview Protocol for a ready-to-use version.

Remote Interviewing Considerations

Remote interviews — video calls, phone calls — are now the default for most teams ^prr. They work well, with a few adjustments:

• Video on: Facial expressions and body language carry important data. Ask participants to enable video when possible, but do not require it.
• Recording with consent: Always ask permission to record. Recordings are invaluable for revisiting key moments, but the request can make some participants guarded. Frame it as "so I can focus on our conversation instead of taking notes."
• Screen sharing for context: If you are discussing a workflow, ask the participant to share their screen and walk you through it. Observation beats description.
• Technical friction: Budget an extra 5 minutes for connection issues. Test your setup before the session. Nothing kills rapport faster than "Can you hear me now?"
• Shorter sessions: Remote attention spans are shorter. Aim for 30 minutes instead of 45-60 for remote discovery interviews.
• More structured transitions: Without physical cues, use explicit verbal transitions between topics: "I'd like to shift gears and ask about..."

Debriefing After Sessions

The 15 minutes immediately after an interview are as valuable as the interview itself ^{prr cdh}. Without a structured debrief, individual impressions harden into narratives that resist correction.

Debrief within 15 minutes of the session ending. Every team member who observed the interview participates.

Structured debrief process:

1. Individual notes first (3 minutes): Each person writes down their top observations, surprises, and quotes before discussing. This prevents the loudest voice from anchoring everyone else's memory.
2. Share around (5 minutes): Each person shares their notes. The facilitator captures them on a shared surface.
3. Identify patterns and contradictions (5 minutes): What did multiple people notice? Where do interpretations diverge?
4. Update the interview snapshot (2 minutes): Finalize the canonical record.

Torres emphasizes that the debrief is where insights emerge ^cdh. A single person's interpretation of an interview is incomplete. Three people who watched the same interview will notice different things. The overlap is signal; the divergence is a prompt for further investigation.

Common Mistakes and How to Fix Them

The Learning Bottleneck: Why the Whole Team Must Participate

The single most impactful change a product team can make to their research practice is this: stop delegating customer contact to a single person ^{mom-test sprint cdh}.

When only the product manager talks to customers, they become a bottleneck. Every insight must pass through their interpretation, their memory, their communication skills, and their biases. The engineer who could have noticed a performance complaint, the designer who could have spotted a workflow gap — both are cut off from the primary source of truth.

Fitzpatrick puts it bluntly: if only one person on the team talks to customers, the team does not have customer understanding — that person has customer understanding, and the team has a second-hand summary ^mom-test.

The Sprint model solves this by having the entire team watch interviews on the final day ^sprint. Everyone sees the same evidence. The post-session discussion is richer because each person brings a different lens — engineering feasibility, design coherence, business viability.

Torres extends this to continuous discovery ^cdh: the product trio (product manager, designer, engineer) jointly conducts weekly interviews. They share the preparation, the facilitation (rotating who leads), the note-taking, and the synthesis. No one person owns the customer relationship. The team collectively owns the understanding.

This is not about efficiency. It is about the quality of decisions. Teams that share direct customer contact make better products because their decisions are informed by firsthand evidence rather than translated summaries. The learning bottleneck is the most expensive bottleneck in product development — and the easiest to remove.