An interview is a chance to understand how somebody handles a real problem. It is much less useful when it becomes a sales presentation followed by “Would you use this?” People can be polite, interested or optimistic without changing their behaviour. Build the conversation around what happened recently and the circumstances behind it.
GOV.UK's service research guidance recommends making research objectives explicit and turning unsupported assumptions into questions. Its public-service context differs from a small business, but that discipline is useful here. The interview plan below is a practical starting point, not a representative market survey.
Choose a decision before choosing questions
Suppose a repair business is considering evening appointments. The decision is whether to pilot those appointments, not whether customers “like convenience.” Useful research questions include: How do people arrange repairs now? What causes them to postpone a booking? What happened the last time a daytime appointment was difficult?
Choose participants who have recently faced the problem. Include people who used your service and people who chose another option, where you can recruit them appropriately. A conversation with a colleague who imagines being a customer is useful rehearsal, but it is different evidence from an actual customer's account.
Ask for a recent example
Open with a neutral prompt: “Tell me about the last time you needed an appliance repaired.” Follow the sequence. How did they notice the problem? Where did they look for help? Who else was involved? What information did they need before booking? Which part took more effort than expected?
Ask “What happened next?” when the answer becomes general. Instead of “Would evening appointments solve this?” ask “How did the appointment time affect your day?” The second wording allows the person to say the time was unimportant or that a different issue mattered more.
Avoid filling every silence. Give people time to remember. Do not interpret hesitation as agreement. When an answer is unclear, reflect it back tentatively: “Have I understood that you delayed because you could not arrange access?” Let the person correct your interpretation.
Keep observations distinct from interpretations
Use notes with three columns: what the person described, your possible interpretation, and a follow-up question. “Asked a neighbour to wait in” is an observation. “Needs evening appointments” is an interpretation. “Would a shorter arrival window have worked?” is a question to investigate, not a finding.
Ask permission before recording and explain how notes will be used. Gather only the information you need, avoid unnecessary identifying details, and manage the material according to your organisation's privacy procedures. Do not put customer stories into public marketing simply because they were shared in a research conversation.
Turn the round into a small test
Imagine several interviewees describe uncertainty about the technician's arrival, while only one mentions evening availability. You might first test a clearer arrival window and proactive updates. That is a reasoned next action; it is not proof that evenings lack demand. The recruitment method and small sample limit what you can conclude.
- Record the decision the interviews should inform.
- Recruit people with relevant recent experience.
- Ask about events before showing your proposed solution.
- Separate described behaviour from your explanation of it.
- Capture contradictory accounts alongside recurring themes.
- Choose one test and define what you will observe next.
Report what you do not know
A useful readout might say: “Interviewees described difficulty planning around arrival times. We will pilot clearer windows with a small set of bookings. We have not measured overall demand for evening appointments.” This keeps the next decision connected to the evidence.
Interviews can uncover language, obstacles and workarounds that a spreadsheet misses. They cannot estimate market size or predict conversion by themselves. Use them to sharpen an offer and its next experiment, then compare the learning with actual enquiries, bookings and delivery experience.
Sources and further reading
This original guide was prepared with AI assistance. Examples are illustrative. The edition date is an assigned date in this 2026 catalogue. See our editorial policy and report a correction.