Short answer. Ask about the last real event, what happened next, what it cost and who made the decision. Keep the proposed solution out of the opening questions, request records where appropriate and treat future enthusiasm as a hypothesis rather than demand.
Customer interviews can create the answer a founder hopes to hear. Explain the idea enthusiastically, ask whether it sounds useful and most polite people will help by imagining a favourable future. The conversation feels validating, but the evidence is mostly about courtesy and the wording of the question.
The commercial cost of a poor interview is not the hour spent talking. It is the confidence to build, hire or borrow on the strength of answers the interviewer helped manufacture.
The job is to recover an event, not win approval
The GOV.UK guide to in-depth interviews recommends open, neutral questions and a focus on stories and real examples rather than generalities about what should happen. The UK Statistics Authority's qualitative-research guidance explains why: leading questions can align responses with the researcher's goals, while even the researcher's reactions can signal which topics appear important.
A discovery interview should therefore reconstruct a recent event:
- What triggered it?
- What happened first, second and third?
- Which person, tool or supplier was involved?
- What workaround was used?
- What time, money, risk or delay followed?
- Who noticed the consequence?
- Who could approve a change?
This does not make the interview objective in the statistical sense. Memory is incomplete, selection can be biased and the founder still interprets the account. It does make the evidence more inspectable than a prediction about an imagined product.
If the starting point is still vague, first separate a costly problem from an ordinary annoyance.
Six rules for questions that produce usable evidence
1. Anchor the time
"Tell me about the last time" is usually stronger than "Do you ever". A date or recognisable event reduces the room for a respondent to combine several experiences into one idealised account.
2. Ask for sequence before explanation
Begin with what happened. Ask why only after the steps are clear. Explanations can be useful, but people often rationalise actions after the event.
3. Remove your proposed answer
"Would an automated dashboard help?" contains both the problem frame and the solution. "How did you notice the delay?" leaves room for an answer the founder did not anticipate.
4. Look for sacrifice and workaround
A complaint is evidence of dissatisfaction. A spreadsheet maintained every Friday, an emergency contractor, a refund or a missed deadline shows what the person already does because the problem matters. Theflght's comparison of complaints and workarounds explains why the distinction matters.
5. Separate user, buyer and approver
The person experiencing the problem may not control the budget. Ask who requested the last purchase, who approved it, which budget was used and why it was accepted or rejected.
6. Probe neutrally
Useful follow-ups include "What happened next?", "Can you give me a recent example?", "How do you know?" and "What made that difficult?" Avoid praise, surprise or finishing the participant's sentence. The aim is not a cold interrogation; it is a conversation in which warmth does not reveal the desired answer.
Practical asset: rewrite 25 weak questions
Use the right-hand column as a starting prompt, then follow the participant's real event.
| Weak or leading question | Evidence-seeking rewrite |
|---|---|
| 1. Do you like this idea? | What happened the last time you faced this task? |
| 2. Would you use an app for this? | What tools did you use last time, in order? |
| 3. Is this a big problem? | What consequence followed the most recent occurrence? |
| 4. How often is this frustrating? | On which dates did it happen in the last four weeks? |
| 5. Would this save you time? | How long did the last instance take, and how was that measured? |
| 6. Would you pay £50? | What did you last pay to solve or avoid this problem? |
| 7. What features do you want? | Which step in the current process creates the greatest cost or delay? |
| 8. Don't you find spreadsheets difficult? | Show me how the spreadsheet is used from start to finish. |
| 9. Would your team love this? | Who used the current process last time, and what did each person do? |
| Weak or leading question | Evidence-seeking rewrite |
|---|---|
| 10. Would your boss approve it? | Tell me about the last related purchase and who approved it. |
| 11. Is speed important? | What happened when the task took longer than planned? |
| 12. Would you switch providers? | When did you last consider switching, and what stopped or enabled it? |
| 13. How much would you pay? | Which budget paid for the current alternative, and what else competes for it? |
| 14. Would reminders be helpful? | How do you remember now, and when did that system last fail? |
| 15. Do customers complain about this? | What was the last complaint, and what action followed? |
| 16. Would automation reduce errors? | Describe the last error and how it was detected and corrected. |
| 17. Would you buy this next month? | What decision must occur before any purchase can be made? |
| 18. Is compliance a concern? | Which recent requirement changed what you did or bought? |
| 19. Would a dashboard improve visibility? | What decision was delayed because information was unavailable? |
| 20. Do you need better reporting? | Which report was last requested, by whom and for what decision? |
| 21. Would an all-in-one service be easier? | Which hand-off between current services failed most recently? |
| 22. Would a free trial convince you? | What evidence did you require before the last comparable purchase? |
| 23. Do you prefer monthly or annual billing? | How was the last comparable purchase funded and approved? |
| 24. Can I send you a proposal? | What, if anything, is the agreed next decision after this conversation? |
| 25. Who else would want this? | Who else experienced or approved the last real event? |
Do not use the table as a rigid script. GOV.UK planning guidance says each research round should have clear, actionable objectives and a method chosen around them. A question belongs in the guide only if its answer could affect the decision.
Grade claims by the evidence behind them
After the interview, do not code every positive sentence as equal. Use this claim-evidence ladder:
| Level | Evidence | Editorial treatment |
|---|---|---|
| 1 | General opinion or compliment | Context only; do not call it demand |
| 2 | Future intention | Hypothesis to test |
| 3 | Specific recent event | Evidence the situation occurred |
| 4 | Record, artefact or observed process | Stronger evidence of sequence, frequency or cost |
| 5 | Existing workaround or sacrifice | Evidence that the problem has consequences |
| 6 | Authorised commercial action | Evidence of commitment within stated terms |
The ladder is a house framework, not a validated scientific scale. Its purpose is to prevent an attractive quote from outranking a less exciting invoice, calendar entry or failed workflow. It also helps explain why stated enthusiasm may not turn into payment.
Illustrative example
Assume a founder is considering a scheduling service for independent clinics. A practice manager says, "Yes, missed appointments are terrible and an AI reminder would be brilliant." That is a favourable opinion.
A useful probe is: "Tell me about the last missed appointment that caused a problem." The manager describes the date, the staff member who called, the unfilled slot and the invoice adjustment. They open a weekly call list and show who maintains it. This does not prove a market or a willingness to buy, but it identifies an observable workflow, consequence and possible user. The next research question is who controls the relevant budget and what evidence they require before changing the process.
A 35-minute discussion guide
The timings below are illustrative.
1 Purpose and permission - 3 minutes. Explain the research, how notes or recordings will be used, the voluntary nature of participation and any limits to confidentiality.
2 Role and context - 5 minutes. Understand the participant's responsibilities without pitching.
3 Last event - 12 minutes. Reconstruct the trigger, sequence, people, tools and outcome.
4 Cost and workaround - 7 minutes. Ask for time, money, delay, risk and existing response. Request an artefact only when appropriate and permitted.
5 Decision path - 5 minutes. Identify users, influencers, approvers, budget and procurement steps.
6 Close - 3 minutes. Ask what was missed, explain what happens to the data and agree any genuine next step.
Test the guide with a colleague before using it. The government interview guide recommends checking clarity and flow, while the Government Analysis Function's questionnaire guidance recommends qualitative testing as part of responsible question design.
Treat recordings and notes as personal data
An interview is not automatically low-risk because it is informal. Names, voices, job details, opinions and screen recordings can identify a person. The GOV.UK recording guide says research materials should be used only for what was agreed, stored securely and stripped of unnecessary identifying detail where possible.
Do not say "we need consent" as if that settles every UK GDPR question. The organisation must identify an appropriate lawful basis, provide privacy information, minimise collection, control access and set retention. If it relies on consent, the ICO says it must be able to demonstrate who consented, when and what they were told. Special-category data, children, vulnerable participants, employee research and international research tools require more careful review.
Collect the minimum evidence needed. Do not copy a participant's customer list, medical information or confidential screen merely because it appears during a demonstration.
End with a decision, not a quote deck
After each session, separate direct observation, participant statement and researcher interpretation. After the round, record the strongest disconfirming evidence as well as the repeating patterns. If the result is unclear, use the diagnosis in why a business-idea test gives no clear answer before recruiting another convenient sample.
The immediate next step is to rewrite the five questions most capable of producing praise. Add a recent-event anchor, one neutral probe, one evidence request and the decision that the answer informs. Reconsider the guide when participants repeatedly misunderstand a term, a new buyer role emerges or the interview stops producing new mechanisms and the next uncertainty is prevalence or paid behaviour.
Sources
- GOV.UK Service Manual: Using in-depth interviews - accessed 9 November 2026.
- GOV.UK Service Manual: Plan a round of user research - accessed 9 November 2026.
- GOV.UK Service Manual: Taking notes and recording user research sessions - accessed 9 November 2026.
- UK Statistics Authority: Ethical considerations associated with qualitative research methods - accessed 9 November 2026.
- Government Analysis Function: Questionnaire design guidance - accessed 9 November 2026.
- ICO: How should we obtain, record and manage consent? - accessed 9 November 2026.
- ICO: Data minimisation - accessed 9 November 2026.
Cover photograph: Focused café conversation by Vitaly Gariev. Unsplash License. Adapted with an editorial headline and The FLGHT branding.
Related guides
Comments (0)