By the Wavefield Research team · Published Sep 24, 2026
Qualitative survey questions collect descriptions, stories, and reasons in the respondent's own words — the material that becomes themes rather than percentages. The craft is a specific request plus one follow-up probe: ask for the last time something happened, not a general opinion about it. Below: 24 examples across six research goals, each with its probe.
Also on this page, the two things the example lists ranking for this search skip: a complete sample questionnaire for qualitative research — six questions in working order, not a grab bag — and the honest answer to how many responses themes need, which is a different number than percentages need.
The commonest misunderstanding in this territory — documented in the methods literature as exactly that: researchers writing on qualitative questionnaires in Perspectives on Medical Education note that newcomers commonly assume open-ended questions alone constitute qualitative research design. They don't. A text box is a format; qualitative research is a set of commitments — the goal is meaning and experience rather than measurement, the sample is chosen for relevance rather than statistical representation, and the analysis codes answers into themes rather than computing shares.
The distinction matters because it runs both directions. A survey full of open-ended questions analyzed by skimming is not qualitative research — it's uncoded text. And a mostly-quantitative survey with three well-placed qualitative research survey questions, coded properly, is doing real qualitative work in its reserved slots. That mixed design is how most commercial studies should carry qualitative: the closed questions measure, and each qualitative slot earns its place by answering something no closed question can. The mechanics of the open-end format itself — required versus optional, box sizing, placement — are covered in open-ended survey questions; where the format sits among the others is in survey question types.
Six research goals, four questions each — and every example carries its follow-up probe, because the pair is the craft. A qualitative question without a planned probe collects the rehearsed first answer; the probe is where the specific, codable material lives. Adapt the bracketed placeholders to your product or category.
Before you know what the closed questions should even ask
Exploration answers become the option lists of your next quantitative survey — the categories respondents volunteer are the ones a closed question can later count.
A metric moved and the closed data can't say why
Pipe the respondent's own rating into the question stem — a probe anchored to their answer reads as attention, not interrogation, and the verbatims code cleanly against the score.
You need the sequence of events, not a verdict
Ask for a story, not an opinion — 'describe the last time' produces specifics that can be coded; 'how do you feel about' produces adjectives that can't.
Roadmap and positioning decisions need the gaps named
Needs hide in behavior, not requests — the workaround question outperforms 'what features do you want' because respondents are unreliable feature designers but reliable witnesses to their own effort.
Messaging, positioning, and survey wording need the customer's words
Report these verbatims unparaphrased — the analytical output is the exact vocabulary, so cleaning it up destroys the finding. Marketing copy written in captured language pre-tests better than invented language.
The respondent is leaving and this is the last honest conversation
Exit respondents have no incentive to be polite, which makes them the highest-signal qualitative source a company has — but only if the questions arrive within days of the decision, while the reasons are still specific.
Example lists show questions; instruments have an order and a logic. Here is a complete six-question qualitative exit study — the kind run when customers cancel — with the reasoning behind each slot. It's a working template for a questionnaire for qualitative research: short, sequenced from event to counterfactual, one closed anchor for crosstabbing, everything else open.
1. Which best describes how long you used [product]? (Under 3 months / 3–12 months / Over a year)
One closed question up front — it takes ten seconds, and tenure is the variable every theme will be read against.
2. What happened that made you decide to cancel?
The central question, asked plainly. No 'we're sorry to see you go' preamble — softening language pulls answers toward politeness.
3. When did you start considering leaving, and what was happening then?
Separates the trigger from the decision. The gap between 'started considering' and 'cancelled' is where retention work would have landed.
4. What did you switch to, or what will you do instead?
Competitive intelligence in the respondent's own frame — including the answer 'nothing,' which codes as a different loss than a switch.
5. What would have needed to be true for you to stay?
A conditional, not a satisfaction item — it forces a concrete requirement instead of a rating.
6. Anything else you want us to know? (Optional)
The safety valve. Always optional, always last, and reliably the source of the one theme the codeframe didn't anticipate.
Six questions is deliberate. Qualitative answers cost respondents real effort — a thoughtful open-end takes a minute or more against five seconds for a closed question — and quality collapses before quantity does. A ten-open-end survey doesn't collect ten answers; it collects three good ones and seven increasingly terse ones.
Two different claims hide inside every qualitative report, and they need different sample sizes. “These themes exist” needs saturation: somewhere around 15–40 substantive answers, new responses stop producing new themes and mostly repeat existing ones. Past that point, more collection adds confidence in coverage, not new findings — which is why a qualitative survey with 30 rich verbatims can be finished while a poll with 30 respondents hasn't started.
“32% of customers mentioned price” is a different claim entirely — the moment a theme gets a percentage, it inherits ordinary sampling math. At 40 coded responses that 32% carries a margin of error around ±15 points; defending it takes a couple hundred responses, sized with the same sample size calculator as any closed question, and the count must be respondents mentioning the theme, never mentions — one person raising price three times is one respondent. Themes with real bases can then be crosstabbed by segment like any other variable, letters and all, per the conventions in cross tabulation. The coding step that turns verbatims into those countable themes — codeframes, AI-assisted assignment, human review — is worked through in open-ended survey questions.
Ask for an instance, not a position. “Describe the last time you exported a report” retrieves a memory; “How do you feel about our reporting?” retrieves a rehearsed verdict. Specific episodes contain the actors, sequence, and friction that coding needs — verdicts contain adjectives.
One probe, planned in advance. The follow-up is where the material lives, but stacking probes turns a survey into an interrogation. Write the single probe that matters into the question flow and stop there — a survey cannot improvise the way an interview or focus group can, and pretending otherwise burns goodwill.
Keep qualitative questions optional. A required open-end fills with “asdf” and single words that poison the codeframe. Optional slots collect fewer answers of far higher quality, and the motivated respondents who fill them are exactly the ones with something to say.
Neutral stems, no leading language. “What did you think of the new dashboard?” outperforms “What do you love about the new dashboard?” for the same reason any leading question fails: it pre-codes the answer. In qualitative work the damage is worse, because the bias contaminates the respondent's vocabulary — the very thing the question was fielded to capture.
Questions that collect descriptions rather than selections: 'Walk me through the last time you used the product,' 'What happened that made you decide to cancel?', 'How would you describe this to a colleague?', 'What do you do right after using it that you wish it did for you?' The pattern is a request for a specific story or explanation, usually paired with one follow-up probe, answered in the respondent's own words.
It can be either, or both — the instrument doesn't decide, the design does. A survey is qualitative when the research goal is meaning and experience, the sample is chosen for relevance rather than statistical representation, and the analysis codes answers into themes. Most commercial surveys are quantitative instruments carrying two or three qualitative questions in reserved slots, which is a legitimate mixed design rather than a compromise.
For themes: roughly 15–40 substantive answers, because themes saturate — after a few dozen rich responses, new answers mostly repeat existing themes. For percentages: far more, because '32% mentioned price' inherits ordinary sampling math, and at 40 responses that figure carries a margin of error around ±15 points. Decide which claim you need before fielding, and size accordingly.
Only with the base size to support them and only counting respondents, not mentions — one person who raises price three times is one respondent mentioning price. With 30 coded verbatims, report themes in frequency order without percentages; with a few hundred, percentages by respondent are defensible and can be crosstabbed like any closed question.
Quantitative questions offer predefined options and produce counts — shares, means, significance tests. Qualitative questions collect free text and produce themes — reasons, stories, vocabulary. The deeper difference is when each works: quantitative questions require you to already know the possible answers; qualitative questions exist for when you don't. That's why exploration runs qualitative first and measurement runs quantitative second.
By coding: reading the verbatims, building a codeframe of themes, and assigning each response to the themes it expresses. Counts of respondents per theme then behave like closed-question data. Traditionally this took an analyst days per few hundred responses; AI-assisted coding now proposes the codeframe and assigns themes in minutes, with a human reviewing the assignments before anything is reported.
Related: open-ended survey questions · survey question types · market research methods · how to analyze survey results
Brief the agent and it places the open-ends properly; when responses arrive, AI-assisted coding proposes the codeframe and you review the themes. From $99 per study.
Start a study