Customer Interviews

What an interview can—and cannot—tell you

A customer interview is a structured conversation used to learn from someone with relevant experience. For an entrepreneurial team, its value is usually depth and context: what the person was trying to do, what happened, what alternatives they considered, what language they use, and which conditions changed the experience. The interview is not a sales call disguised as research, and it is not a referendum on whether a team likes its own idea. A participant may be a current user of an existing process, a likely user of a proposed service, or another person who directly experiences the workflow the team needs to understand. The selection must follow the learning question. If the question concerns how independent tutors manage schedule changes, interviewing people who merely know a tutor is weaker evidence than interviewing tutors who coordinate recurring sessions.

A useful interview round begins with one or two decisions the team needs to inform and a short discussion guide. The guide gives every participant a comparable core set of topics, while leaving room to pursue a relevant detail. GOV.UK's guidance on in-depth interviews recommends open, neutral questions and attention to stories and real examples. That is why a prompt such as “Tell me about the last time a schedule changed” is often more informative than “Would you use our schedule-change app?” The first asks for a remembered event; the second introduces a proposed answer and invites courtesy, imagination, or a quick yes.

Interview evidence is qualitative and exploratory. A small set of deliberately recruited participants can surface recurring patterns, contrast situations, and generate a better next question. It does not calculate a population percentage, establish a market size, or prove that a particular product will sell. GOV.UK contrasts small rounds commonly used for interviews with survey-style methods that need much larger samples for clear quantitative findings. The precise number appropriate for an interview round depends on the question, heterogeneity of the people involved, access constraints, and what has already been learned. Rather than borrow a magic number, state who was interviewed, why they were relevant, what differences were included or missed, and what the sample cannot represent. Market sizing and statistically representative research require their own design.

Plan recruitment and questions without steering the answer

Recruitment is part of the method, not an administrative afterthought. Write inclusion criteria that connect to the behavior or situation under study: for example, tutors who independently schedule at least several recurring sessions, or families who recently changed a session. Deliberately seek meaningful variation when it could change the experience—such as a new versus experienced tutor, a different scheduling arrangement, a disability or access need, or a different tool setup. This is not a claim of statistical representativeness. It is a way to avoid letting the most convenient or enthusiastic contacts stand in for everyone. Record the recruiting route and likely sources of bias. Friends, early supporters, and people recruited from a founder's network may be useful to learn from, but their relationship to the project can shape what they say.

Use a guide with an opening, a small number of topic areas, and a closing. Begin by explaining the purpose in plain language and confirming consent; then ask broad questions before narrower ones. Open questions commonly start with “how,” “what,” “tell me about,” or “walk me through.” Neutral wording does not contain the preferred conclusion. Compare “What happened when you needed to reschedule?” with “How frustrating was it when the confusing system made you reschedule?” The second question assumes both that the system was confusing and that frustration occurred. It pressures the participant to agree with the interviewer’s framing. Avoid stacking two questions together as well: “How do you schedule and pay and communicate?” gives no clear answer target.

Questions about an actual episode are usually more useful than questions about a hypothetical future purchase. “What did you do next?” “What else did you try?” and “What made that option workable or not?” can expose constraints, workarounds, and alternatives. A participant's claim that they would buy something is still a statement about a possible future, not evidence that they will do so. The point is not to catch people being wrong. It is to understand their present practice before treating a solution as the center of the conversation. This lesson therefore does not build a customer-discovery framework or teach a final value proposition; it teaches the interview craft that can supply bounded evidence to those later activities.

Listen, record, and analyze with care

Good moderation makes the participant, not the interviewer, do most of the explaining. Listen for the sequence of a real event: trigger, goal, actions, tools, obstacles, choices, and outcome. Pause rather than filling silence with an answer. A brief reflection can check understanding—“It sounds as if the text message was easy to miss because you were teaching”—but frame it as a check, not a correction. Then ask a simple, non-leading follow-up: “Can you tell me more about what happened after that?” or “What made that difficult?” Follow the participant's relevant detail while returning gently to the guide when needed. Do not debate, promise a solution, or treat a participant’s politeness as approval.

Before a session, make voluntary participation and consent practical, not ceremonial. Explain who is conducting the interview, its purpose, what will happen, whether anyone will observe, whether you will take notes or record, what information will be collected, who may receive it, how long it will be kept, and how a participant can decline or stop. Seek separate, clear permission for a recording instead of assuming that joining the call authorizes it. Government guidance is written for its own setting and legal regime, so it should not be read as universal legal advice. Still, these disclosures express a sound ethical baseline. Applicable privacy, consumer-protection, employment, accessibility, and research rules vary by location, industry, participant, and data type; obtain appropriate local advice when the stakes or data warrant it.

Keep the data footprint small. Do not collect a full name, address, sensitive history, or recording merely because it might be convenient later. Separate contact details from research notes when possible, restrict access, use secure storage appropriate to the setting, and remove identifiers from findings when identity is not necessary. Notes and recordings can contain personal information even when the discussion seems routine. A person must understand the intended use before agreeing to it.

Analysis begins soon after the round. Preserve an observation separately from the interpretation it may suggest. “Participant 04 checked three messaging apps before finding the update” is an observation. “Tutors need a new messaging platform” is an interpretation and a proposed solution; it does not follow automatically. Extract relevant observations from notes or an authorized recording, label the participant context, and sort similar observations into provisional themes. Then write a bounded finding: for example, “Several interviewed tutors described missing schedule changes when updates were spread across more than one channel.” Report contrary cases too. Finally, identify what the evidence supports, what it does not establish, and the next question. That discipline turns interview notes into an auditable learning record rather than a collection of memorable anecdotes.