write-interview-guide

Start with open openers like How, What, Tell me about, or Walk me through instead of Did/Was/Do/Would. Strip embedded premises and value judgments, ask about past real behavior rather than predictions, and run each question through a lint check for leading, double-barreled, and hypothetical phrasing.

A 60-minute session holds roughly 8 to 12 main questions, since each question plus probes runs 3 to 5 minutes. Overstuffing eliminates probing time, so mark 2 to 3 questions as skip-if-short in advance.

A coverage matrix is a table mapping every interview question to the research questions it answers. Each research question needs at least two interview questions covering it, and any interview question mapping to nothing gets cut before the pilot.

Use a survey when the instrument is self-administered at scale rather than moderated live. Interviews suit exploratory questions needing probes and follow-ups; surveys suit quantifiable questions across many respondents, and each format has its own bias audit.

Prediction questions produce polite fiction because people cannot accurately forecast their own behavior. Ask about past real behavior instead, such as "Have you ever paid for something like this? Tell me about it," which yields recalled evidence rather than speculation.

Yes, no guide is done until piloted with a colleague or one real participant. The pilot checks total time, confusing questions, dead questions everyone answers identically, and awkward funnel ordering, and revisions are recorded in the guide header.