Replace leading customer interview questions
Rewrite leading customer interview questions into neutral prompts about real behavior, then code observations separately from preferences and interpretations.
What makes a customer question leading?
A leading question embeds the answer or the researcher’s preferred explanation. “Would an easier dashboard save you time?” invites agreement without establishing the current problem. A neutral question asks the participant to describe the last time they performed the task, including what happened and what they used.
GOV.UK’s interview guidance recommends open, neutral questions, useful follow-ups and real stories rather than generalities about how work should happen. The script below applies that method to a fictional reporting task. It does not turn a small interview set into a representative demand estimate. [GOV.UK Service Manual: using in-depth interviews]
A before-and-after interview script
| Leading prompt | Neutral replacement | Evidence sought |
|---|---|---|
| Do you hate copying numbers manually? | Tell me about the last report you prepared. | Actual sequence before naming a problem |
| Would automation save you hours? | Which steps took the most time, and how do you know? | Observed or estimated effort with context |
| Our tool would make this easier, right? | What did you try when that step became difficult? | Existing alternatives and workarounds |
| Would you pay for a better dashboard? | How was the current solution selected and funded? | Past decision process, not hypothetical purchase intent |
| So the main problem is accuracy? | What happened after the report was shared? | Consequences without imposing a theme |
Do not force every participant through the same exact wording if their account has already answered a topic. Keep the research questions consistent while allowing follow-up. The guide should support listening rather than become a checklist that interrupts the useful story.
Follow one real episode far enough
- Ask when the task last happened and who was involved.
- Trace inputs, tools, handoffs and the completed output.
- Ask what went wrong, if anything, before suggesting a problem.
- Request an example or artifact only with appropriate permission.
- Clarify whether a time or cost figure was measured or estimated.
- Ask what happened next and whether the outcome changed a decision.
A participant might say the report “takes forever,” then describe ten minutes of copying and two days waiting for approval. Those are different mechanisms. One suggests a production task; the other suggests an ownership or review process. Preserve the timeline instead of translating both into demand for automation.
Separate behavior, preference and interpretation
| Statement | Code | Allowed interpretation |
|---|---|---|
| I copied three tables yesterday | Reported recent behavior | A concrete task to investigate |
| I would love one button | Hypothetical preference | A design idea, not confirmed adoption |
| My manager checks it before release | Reported process | Review is part of the workflow |
| They need an AI dashboard | Researcher interpretation | Hypothesis requiring further evidence |
| Another participant uses a template successfully | Contradictory observation | Explore conditions rather than discard it |
Keep a source location for each note and avoid turning a paraphrase into a direct quotation. If a participant estimates effort, retain that qualifier. An interview can explain context and mechanisms while leaving prevalence, causation and purchasing behavior unresolved.
What should the interview change?
Choose a next question or prototype test that follows from the evidence. If approval waiting dominates, test a clearer handoff rather than immediately building a data-import tool. If copying errors appear repeatedly, inspect the actual source and output with permission. The next step should reduce the uncertainty that affects the decision.
I would keep the awkward counterexample in the research summary. It often reveals the condition under which the proposed solution is unnecessary. That is useful product direction, even when it makes the original idea less exciting.
Sources
AI-assisted writing and an AI-generated editorial cover. Source links and checked dates are listed below; illustrative examples are labeled in the text.
- GOV.UK Service Manual: using in-depth interviews — checked 2026-10-18