Choose one decision and recruit for recent experience

Start with the decision the research must inform, such as which buyer obstacle deserves one evidence-led article. Recruit actual or likely users who encountered the relevant situation recently, including people who did not complete the task. Define inclusion criteria before reading responses so enthusiasm for the idea does not determine the sample. Set a bounded round and stopping rule: the team proceeds when it can name recurring episodes, meaningful contradictions, and the next test—not when it has collected a desired number of compliments.

Evidence: Government Digital Service

Draft a guide from topics, not preferred conclusions

Open with context, then ask for the last time the person tried to accomplish the task. Follow the timeline: trigger, first action, alternatives, obstacle, workaround, consequence, and outcome. Use 'what happened next?' and 'can you show or describe that?' rather than 'would you agree?' or 'how useful would our guide be?' Pew documents how wording and order can change responses. Put concept reactions at the end and mark them separately so they do not contaminate the problem narrative.

Evidence: Pew Research Center; Oxford Academic

Interview by following evidence-rich turns

Listen longer than you speak. When a participant uses a broad word such as easy, expensive, or trustworthy, ask what made it so in the remembered episode. Request artifacts only when consent and privacy allow—perhaps a blank comparison sheet or a redacted checklist, not personal purchase data. Note moments when your prompt supplies vocabulary that the participant later repeats. Do not reward positive answers with extra enthusiasm. A neutral pause often produces more useful detail than a sequence of increasingly specific suggestions.

Evidence: Government Digital Service

Enter episodes into an anonymized ledger

Create one row per recent event with participant code, recruitment source, situation, trigger, exact phrase, action, alternative considered, workaround, consequence, evidence type, uncertainty, and contradiction. Strip unnecessary names and sensitive details. Record whether an action was observed, self-reported, or inferred. The GOV.UK analysis method emphasizes extracting observations before interpreting them; preserve that separation. A colorful quote is not automatically important if it lacks a costly or recurring task behind it.

Evidence: Government Digital Service

Synthesize patterns and actively search for disconfirmation

Group rows by situation and barrier, not by the participant demographics the team finds memorable. Count episodes cautiously, then read the contradictory rows. Ask whether recruitment, wording, or the moderator could have produced the apparent pattern. Compare with another evidence source such as recurring support questions or search-result tasks. If only people already loyal to a product described the issue, state that boundary. Revise the problem statement when a deviant case reveals a different trigger or a successful workaround.

Evidence: Government Digital Service; Oxford Academic

Close the round with one next test and one non-claim

Write the proposed audience, situation, desired progress, obstacle, current alternative, and content decision. Attach supporting and conflicting episode IDs. State what the interviews cannot tell you—population frequency, transaction likelihood, or whether the proposed article will earn attention. Choose a cheap next test, such as analyzing actual query patterns or publishing a narrowly scoped explainer with an explicit learning metric. Stop research that the team cannot act on; repeated interviewing without a decision creates data burden rather than insight.

Evidence: Government Digital Service; Pew Research Center

Debrief before memory becomes a new source of bias

Reserve time immediately after each session for the interviewer and note-taker to separate observations, tentative interpretations, and questions for the next round. Record any leading prompt, missed probe, technical problem, or participant discomfort. Do not rewrite the guide silently; version changes and note which sessions used each version. A methodological mistake does not automatically invalidate every observation, but it changes the confidence and comparison that the table should support.

Evidence: Government Digital Service; Government Digital Service

Sources and further reading

These references informed this article. A source supports a claim; it does not imply endorsement of TenMultigure or any future product reference.

  1. Using in-depth interviewsGovernment Digital Service · Accessed August 10, 2026

    Guides recruitment, consent, discussion-guide structure, and neutral follow-up behavior for the bounded interview round.

  2. Analyse a research sessionGovernment Digital Service · Accessed August 10, 2026

    Supports the ledger workflow that extracts observations before the team groups them into editorial findings.

  3. Writing Survey QuestionsPew Research Center · Accessed August 10, 2026

    Provides concrete evidence that wording and ordering choices can influence participant responses and apparent priorities.

  4. The Effect of the Question on Survey Responses: A ReviewOxford Academic · Accessed August 10, 2026

    Adds a research review of question effects used to challenge patterns that may have been elicited by the moderator.

Reviewed for clarity and evidence

Reviewed by TenMultigure Editorial Review. See an error or a source that has changed? Tell the editorial team.

Review method: AI-assisted desk research with editorial checks. Reviewed ; next scheduled review . Turned the interview brief into a consent-aware seven-step procedure, added an episode ledger and disconfirmation pass, and separated problem evidence from later concept reactions.