Reference

Interview guide design and question bank

Part of the Primary Research Synthesis skill · loaded on demand from SKILL.md

Read this file when designing or reviewing an interview guide, briefing an interviewer, or diagnosing why a set of transcripts contains opinions rather than evidence. The principles come from the two frameworks in references/frameworks.md: the guide must let categories emerge (Grounded Theory) and must collect theory-in-use as well as espoused theory (Argyris and Schön).

1. Design principles

Funnel, do not grid. Open with the interviewee's own account of their work and the problem, then narrow to the topics you need, then close with the specific probes. A guide that starts with the specific probes tells the interviewee what you think the answer is, and the rest of the interview confirms it.

Ask for incidents, not habits. "How do you usually handle X?" yields the espoused theory. "Tell me about the last time X happened" yields the theory-in-use. The second is slower and always more useful. Every topic in the guide should have at least one incident question.

Pair every policy question with an incident question. The pairing is what makes the espoused/in-use table possible. Code the answers separately.

Probe for conditions and consequences. After an incident: what made that happen? what would have made it go differently? what happened next? who knew? These are the Strauss and Corbin paradigm elements phrased as questions, and they fill the axial table later.

Reserve a quarter to a third of the time. Fifteen to twenty minutes of a 60-minute interview should be unscripted, spent following whatever the interviewee raised that the guide did not anticipate. If you finish the guide with time to spare, you have not been listening.

Iterate the guide. After the first two or three interviews, revise. Add probes for what emerged, drop questions that produced nothing, and move the most productive question earlier. Keep a version log; the write-up must say which interviews used which version.

Ask the interviewee to disconfirm. Near the end: "Who would disagree with what you have told me, and what would they say?" This gives the counter-evidence line and the next sampling move.

Never ask a question the interviewee cannot know the answer to. A sales rep does not know the bonus mechanics of a manager. Route questions to the segment positioned to observe.

2. Structure of a 60-minute guide

BlockMinutesPurpose
0. Framing and consent3Purpose, anonymity, recording, right to skip
1. Role and context7Interviewee's own description of their job and what a good week looks like; establishes vocabulary
2. Core topic, open8"Tell me about how X works from where you sit"; espoused theory, in their words
3. Core topic, incidents15Two or three recent incidents in detail; theory-in-use
4. Conditions and variation7When does it go differently? Where? Who does it differently?
5. Unscripted follow-up15Whatever emerged
6. Close5Who would disagree; who else should we talk to; anything we should have asked

3. Question bank by purpose

Opening (espoused theory, vocabulary)

  • Walk me through what you are responsible for and what a good week looks like.
  • If you had to explain to a new joiner how <topic> works here, what would you tell them?
  • What does the company say it does well? Do you agree?
  • What is the official process for <topic>? Who owns it?

Incident elicitation (theory-in-use)

  • Tell me about the last time <topic> came up. Start from the beginning.
  • What exactly did you do first? Then what?
  • Who did you speak to? What did they say, as close to their words as you can remember?
  • How long did that take?
  • What did you expect to happen? What actually happened?
  • Can you think of a time it went the other way?
  • When did you last see <policy> enforced? What happened to the person or the order?

Conditions and variation (axial material)

  • What made that happen the way it did?
  • What would have had to be different for you to do something else?
  • Is there anyone here who handles this differently? What do they do?
  • Where in the business does this work best? Worst? Why?
  • What happened as a result? Who found out? What did they do?
  • What does your manager see of this? What do they not see?

Incentives and governing values (Model I probes)

  • When something goes wrong on <topic>, who gets asked about it?
  • What is the worst outcome for you personally in this situation?
  • What are you measured on? How does that affect what you do on <topic>?
  • What would you have to say out loud in a meeting that you currently would not?

Value and negotiation (for commercial topics)

  • In the last negotiation you ran, what did the customer ask for, and what did you say?
  • At what point did you mention <the company's advantage>, if at all?
  • What do customers say when you hold price? What do you do then?

Customer or counterparty view

  • Compared with the other suppliers on your panel, how does <client> behave in a negotiation?
  • When you push back on price, what happens?
  • What do you value about <client> that you would pay for? Do they ever ask you to?

Closing

  • Who would disagree with what you have told me? What would they say?
  • Who else should we talk to, and what should we ask them?
  • What did we not ask that we should have?
  • Is there anything you said that you would rather we did not attribute, even by role?

4. Probes to keep in reserve

Short, neutral, and usable after any answer:

  • "Say more about that."
  • "What did that look like?"
  • "For example?"
  • "And then?"
  • "How often?"
  • "Who else was involved?"
  • "What did you make of that?"
  • (silence for four seconds)

5. Questions that produce bad data, and their replacements

AvoidWhyAsk instead
"Do you follow the discount matrix?"Yes/no; invites the espoused answer"Tell me about the last quote you escalated."
"Is pricing a problem here?"Leads; imports the consultant's frame"What gets in the way of a good margin on a big order?"
"Would you use a tool that did X?"Hypothetical; people are poor predictors of their own behaviour"What do you use today to do X? Walk me through it."
"Why don't reps use the value story?"Assumes the premise; invites blame"In that negotiation, when did service come up?"
"On a scale of one to five..."Converts qualitative data to a number without a definitionAsk for the incident; rate intensity yourself at coding
"What should the company do?" (early)Turns the interview into a wish listSave for the close, and code it separately as recommendation, not evidence

6. Briefing the interviewer

  • Two people per interview: one asks, one takes near-verbatim notes and watches for topics to return to.
  • The asker holds the guide but does not read from it; the order flexes to follow the interviewee.
  • Note timestamps against topics so the transcript can be navigated quickly.
  • Immediately after, spend ten minutes writing the "surprise note": what was unexpected, what contradicted earlier interviews, what to add to the guide. This note is the seed of the analytic memo.
  • Do not discuss conclusions with the client between interviews. The client will push you toward their hypothesis, and the sampling should be driven by the data.

7. Observation protocol (field visits and ride-alongs)

Observation captures theory-in-use directly and is the strongest source for the in-use column of the gap table.

  • Agree with the host what will be observed and that individuals will not be named.
  • Record what happens in sequence, with times, without interpretation. "Rep texted manager at 10:14; reply 10:18; discount confirmed 12 percent" rather than "manager rubber-stamped".
  • Note what is absent as carefully as what is present. The absence of a service mention in a price negotiation is a finding.
  • After the session, ask the person observed to narrate what they did and why. The narration is espoused theory; the record is in-use. Code both.
  • One observation is an anecdote; three that agree are a pattern; treat them as such in the evidence counts.