Discussion Guide Design

How many questions should a discussion guide have?

Far fewer than most guides contain. A primary question explored with real probing takes 4 to 8 minutes and a reconstructed episode takes 8 to 12, so a 60-minute depth interview supports four to six lines of questioning, which is eight to twelve questions the moderator will actually ask. A guide with forty questions has usually written every probe and every prompt as though it were a question, and in the room it produces forty shallow answers instead of five deep ones.

What is the difference between a discussion guide and a questionnaire?

A questionnaire is an instrument for the respondent: fixed wording, fixed order, asked identically every time so that answers can be counted. A guide is an instrument for the moderator: it fixes the purpose, the sequence and the boundaries, and deliberately leaves the route between the opening question and the exit condition free, because the valuable material is usually in the direction nobody predicted. Writing a guide with questionnaire discipline removes exactly the thing qualitative research is for.

What order should topics go in?

Broad to narrow, and four rules in priority order. Unaided before aided, without exception, because once you have named a brand or a reason, spontaneous salience of it cannot be recovered. Spontaneous before prompted, so the prompt list checks coverage rather than structuring the answer. General before specific, because a specific failure raised early re-frames everything after it. Behaviour before attitude on the same object, because attitudes reported after an episode get rationalised into consistency with it.

How do I write probes that do not lead?

Write probe classes rather than scripted follow-ups, and test each one four ways. It must not introduce a noun the participant has not used. It must not offer a candidate answer ("was it because it was expensive?"). It must not carry an evaluation, because repeated enthusiasm teaches the participant which answers are rewarded. And it must not be answerable "yes" by someone with nothing to say. Balance matters too: probing positives twice and negatives once leads through attention, even when every individual probe is neutral.

Where should stimulus be shown in an interview or group?

Late, and behind a boundary you draw explicitly in the guide. Everything above that line is pre-exposure data and can never be collected again: spontaneous vocabulary, current behaviour, unaided awareness, and the problem as the participant frames it. Below the line, everything is a reaction to a document. Where more than one stimulus is shown, rotate the order across sessions and record which order each session used, or a concept effect and an order effect become indistinguishable.

Are projective techniques worth using?

Only when three things are true: a direct question would predictably produce flat, normative or socially managed answers; the technique reaches material the direct route cannot; and there is an interpretation rule agreed before fieldwork saying what the output means. Otherwise it is decoration, and decoration in a 60-minute session costs a whole line of questioning. Two disciplines apply: the collage or the completed sentence is a prompt for talk and the talk is the data, so budget the debrief as the larger half; and a projective output never establishes an unconscious motivation, it generates material an analyst interprets, and the interpretation must be visible as interpretation.

How do you stop one person dominating a focus group?

By design, not by moderation heroics. Private written responses before open discussion, timed go-rounds, direct nomination by name, splitting into pairs for five minutes, and physical artefacts that occupy hands and equalise turn-taking. These are structural remedies that can be specified in advance and repeated. Asking a moderator to manage a dominant participant in the moment is a hope, not a design.

When is a focus group the wrong instrument?

When the material requires privacy: debt, health, shame, family conflict, workplace grievance, anything a participant manages the presentation of in front of peers. Also when the objective needs an individual decision journey reconstructed, when prevalence will be inferred from what was said, or when status or power differences within the group would silence the lower-power participants. A group produces a socially negotiated account, which is genuinely valuable for range, vocabulary and norms, and genuinely unsuitable for the private and the counter-normative.

Should I ask people whether they would buy or use something?

No. Stated future behaviour in a moderated setting measures politeness and imagination, and the answer is systematically generous. Ask what they did last time, what they nearly did, what they considered and rejected, what stopped them, and what would have had to be different. The same applies to asking people to explain their own unconscious motivations directly: they will give you the most culturally available explanation, fluently.

Can I change a discussion guide once fieldwork has started?

Yes, and often you must. A guide is a hypothesis about the conversation, and the first sessions are the test. Treating it as fixed once fieldwork has shown it is wrong wastes every remaining session. The condition is that the change is logged: what changed, when, why, and which sessions used which version, so the analyst can tell whether a difference between session three and session eight is a participant difference or an instrument difference. Where a tracked or multi-market study depends on comparability, changing the fixed spine is a researcher's decision, not an editing decision.

How do you adapt a discussion guide for other markets and languages?

Comparability lives at the level of objective, section purpose and sequence, not at the level of the sentence. Hold five things fixed across markets: objectives, section order, the unaided-before-aided boundary, the stimulus and its exposure point, and the probe classes. Let five things vary: exact wording, warm-up, examples, time distribution, and the choice of enabling technique, since some do not travel. Translate the intent rather than the words, log every adaptation so market differences can be told apart from instrument differences, and get a local reading before findings are fixed.

Can AI write a discussion guide?

It can draft one quickly, and the risk is specific: generated guides tend to be fluent, comprehensive, evenly balanced across every objective, and roughly three times too long for the session booked. Used properly, AI should be required to produce the section brief, the timing arithmetic and the cut log alongside the guide, should never invent stimulus content or concept copy that was not supplied, and should never generate example participant answers, which have a habit of migrating into analysis documents. A guide with nothing cut has not been designed.