Open-ended questions let participants answer in their own words. Closed questions limit the answer to a short fact, category, rating, or choice.
In a user interview, use open-ended questions to uncover stories and surprises. Use closed questions later to confirm details. But question type is only part of the job: a useful question usually asks about a specific event that already happened.
Compare these two questions:
- Open but speculative: “What do you think about our reporting feature?”
- Open and specific: “Walk me through the last time you prepared a report for your leadership team.”
Both are open-ended. Only the second is likely to reveal the participant’s goal, sequence of actions, constraints, workarounds, and decision points.
That gives you a practical rule for reviewing an interview guide:
- Anchor the question to a specific past event.
- Make it open enough for the participant to tell the story.
- Use short closed questions only when you need to clarify what happened.
Open-ended vs closed questions: a quick comparison
Nielsen Norman Group defines open-ended questions as questions that allow a free-form answer. Closed questions restrict the participant to a limited answer.
| Open-ended question | Closed question | |
|---|---|---|
| Typical answer | A story, explanation, or description | A fact, choice, rating, or yes/no |
| Best use in interviews | Discovery, context, motivations, and unexpected details | Screening, confirmation, timing, and clarification |
| Example | “Tell me about the last time an export failed.” | “Was that before or after the migration?” |
| Main risk | A broad question can invite vague summaries or speculation | A narrow question can stop the story or reveal what you expect |
A question does not have to be yes/no to function as closed. “How often do you export data?” usually produces a short frequency estimate. It may begin with how, but the answer space is narrow.
Closed is also not the same as leading. “When did that happen?” is closed and neutral. “Was onboarding confusing?” is both closed and leading because it suggests the interpretation you want the participant to consider.
If you need help separating the two, use this guide to avoiding leading questions.
Why an open-ended question can still produce weak evidence
The usual advice—ask open questions, avoid leading questions, and do not ask people to predict the future—is necessary. It is not sufficient.
Teresa Torres demonstrates the gap in her story-based customer interview example. “Tell me about your experience with Netflix” is open, non-leading, and about the past. Yet it encourages the participant to summarise Netflix in general.
Change the prompt to “Tell me about the last time you watched Netflix,” and the answer becomes a story. In Torres’s example, that story reveals the participant’s setting, device, preparation, connectivity constraint, and what happened during the viewing session. Those details give a product team something it can investigate. A general preference does not.
The stronger distinction is therefore:
- Specific: one real event, decision, attempt, purchase, cancellation, or workaround;
- Speculative: what someone usually does, thinks in general, or imagines doing later.
This principle also sits at the heart of The Mom Test for founders: past specifics are more useful than generic opinions and future promises.
Use this two-axis test
Open versus closed and specific versus speculative are separate axes. Put them together and every draft question falls into one of four boxes.
| Specific past event | General or speculative | |
|---|---|---|
| Open | “Walk me through the last time you pulled a report for the board.” Start here. | “What do you think about our reporting?” Likely to produce a summary. |
| Closed | “Was that the same report you mentioned earlier?” Useful for clarification. | “Would you use scheduled exports?” Weak evidence for a future decision. |
The top-left box is the default for discovery interviews. It gives the participant room to speak while keeping the answer grounded in an event you can examine.
The bottom-right box is where false validation accumulates. A participant can sincerely say they would use a feature and still behave differently when the trade-off becomes real. Use prototypes, behavioural data, or a genuine commitment when you need to validate a solution. The article on discovery versus validation interviews explains where interviews fit in that decision.
There are exceptions. A general attitudinal question may be appropriate when attitude itself is the research subject. The point is not to ban opinions. It is to match the question to the evidence your decision requires.
Rewrite common interview questions around real events
The fastest way to improve a guide is to replace summaries and predictions with recent incidents.
| Weak question | Stronger question | What changes |
|---|---|---|
| “What do you think about reporting?” | “Tell me about the last report you prepared.” | Opinion becomes an event |
| “How do you normally prioritise work?” | “Walk me through the last time two urgent tasks competed.” | Typical behaviour becomes a real trade-off |
| “Would you use an approval workflow?” | “Tell me about the last approval that slowed the team down.” | Feature prediction becomes problem evidence |
| “Why did you cancel?” | “Take me back to the moment you decided to cancel. What was happening?” | A demand for a reason becomes a sequence |
| “Did onboarding make sense?” | “What happened when you first tried to set the account up?” | Suggested judgement becomes an observable experience |
Listen for answers that begin with usually, I would, I think, or probably. Do not argue with them. Bring the conversation back to evidence:
- “Can you tell me about the last time?”
- “What happened next?”
- “What did you do then?”
- “What made that difficult?”
- “How did you work around it?”
That is also how to get beyond a rehearsed explanation. This guide to asking why without repeatedly asking “why” gives you more neutral probes.
When to use open-ended questions
Open-ended questions should do most of the work in a qualitative interview. Use them for three jobs.
Start a relevant story
Good opening prompts are broad enough to avoid planting an answer, but bounded enough that the participant knows which experience to recall.
- “Tell me about the last time you prepared a customer renewal.”
- “Walk me through the most recent support issue you escalated.”
- “Take me back to the last time you compared tools for this job.”
Avoid prompts so broad that the participant has to guess what you mean. “Tell me about your workflow” could refer to a day, a tool, a team, or a career. Name the event without naming the answer.
Follow the participant’s language
Open probes let you explore what the participant introduced:
- “You called that step fragile. What makes it fragile?”
- “Tell me more about the workaround.”
- “What happened after the hand-off?”
These probes are where an interview becomes adaptive rather than a spoken questionnaire.
Discover what you did not know to ask
Closed questions constrain the answer to your existing model of the problem. Open questions leave space for a motivation, constraint, or alternative you did not anticipate.
That does not mean writing twenty main questions. NN/g’s funnel guidance says a user-interview guide will usually have five to eight open-ended questions, followed by open or closed follow-ups. Treat that as a planning range, not a quota. A focused interview may need fewer.
When to use closed questions
Closed questions are useful when they have a defined supporting role.
Screen before the interview
Job title, product usage, company size, recent experience, and eligibility often belong in a screener. Asking them before the interview protects time for stories.
NN/g specifically recommends moving factual questions into the screener when they determine who should take part. Do not spend the first ten minutes making a participant prove that they qualified.
Clarify a story
Once a participant has introduced the topic, a closed question can pin down a detail:
- “Was this before or after the pricing change?”
- “Did anyone else need to approve it?”
- “Was that the first time the error appeared?”
Keep the question neutral and return to an open probe if the detail matters.
Quantify with the right method
If you need to know what percentage of customers use Postgres, which plan they pay for, or how often a button is clicked, use product data or a structured survey. Closed survey responses can be aggregated. Interview stories cannot establish prevalence.
An interview can help you understand why a pattern occurs or what it means in context. It should not be used to turn a small qualitative sample into a market statistic.
Use the funnel: open first, closed last
The funnel technique moves from broad to narrow. Repeat the funnel for each main topic:
- Open with a specific story: “Tell me about the last time you prepared the monthly report.”
- Probe what the participant introduced: “What happened when you combined the data?”
- Explore the consequence: “How did that affect the deadline?”
- Clarify missing details: “Was the finance team involved?”
- Start a new funnel for the next research question.
This order matters. Ask about a feature too early and the participant knows what you hope to hear. Start with their recent experience and you learn how the problem appears before your product language enters the room.
Follow-up probes can matter more than the opening question
A strong first question does not guarantee a strong answer. The next question often creates the depth.
A 2026 paper in the Hawaii International Conference on System Sciences proceedings tested three follow-up strategies in a chatbot survey. The final sample included 151 German-speaking participants with a German Abitur, and the topic was German retirement policy.
Every group answered the same open-ended questions:
| Follow-up condition | Participants | Mean words | Mean distinct meaning units |
|---|---|---|---|
| No probe | 51 | 41.88 | 8.51 |
| Generic scripted probe | 50 | 75.02 | 12.04 |
| Contextual GPT-4o probe | 50 | 106.02 | 14.52 |
The differences in response length and thematic richness were statistically significant. The contextual probe performed best, but even the generic follow-up produced longer, richer answers than no follow-up.
Do not carry those exact effect sizes into every user interview. This was a chatbot survey about one policy topic, with a deliberately constrained educational and language sample. Word count is also not the same as useful evidence. The safer conclusion is narrower: plan to probe, and make the probe respond to what the participant actually said.
A checklist for reviewing your interview guide
Before fielding a study, check every main question:
- Does it serve the research objective?
- Does it ask about a specific event that already happened?
- Is it open enough for a story?
- Does it ask one thing at a time?
- Does it avoid suggesting an answer or interpretation?
- Have you written two or three neutral probes?
- Can any factual closed question move to the screener?
- Are you using product data or a survey for questions that require counts?
Then pilot the guide with one person. If they ask what you mean, give only short answers, or drift into hypotheticals, revise the question before recruiting the rest of the sample.
Keep your notes tied to the event, not just the conclusion. The user-interview note-taking guide includes a lightweight format for preserving quotes, observations, and context.
How Maren handles the question sequence
Maren turns a research objective into a guide using a purpose-built interview style. During the asynchronous conversation, it asks adaptive follow-up questions based on what each participant says. It then creates individual summaries and synthesises themes across interviews with supporting evidence.
That workflow is useful when the hard part is following several participants’ stories consistently without scheduling and moderating every call yourself. It does not remove the researcher’s responsibilities. You still need the right objective, the right participants, and judgement about what the evidence can support.
Maren’s conversation remains text-based, although participants can use optional voice dictation. That keeps the record easy to review and quote, but it does not capture vocal tone, facial expression, or observed product use. Choose the medium that matches the decision.
Frequently asked questions
What is the difference between open-ended and closed questions?
An open-ended question allows a free-form answer, such as a story or explanation. A closed question limits the response to a short fact, choice, rating, or yes/no answer. In interviews, open questions support discovery; closed questions support screening and clarification.
Are closed questions bad in user interviews?
No. They are useful for confirming details after a participant has told a story. Problems arise when closed questions dominate the guide, suggest an answer, or replace evidence you could collect through a screener, survey, or product analytics.
What is the best open-ended question for a user interview?
A reliable starting pattern is: “Tell me about the last time you…” Complete it with a real event connected to your research objective. Follow with “What happened next?”, “What did you do?”, and short clarifying questions.
How many open-ended questions should an interview have?
NN/g suggests that a typical user-interview guide contains five to eight main open-ended questions. The right number depends on the study goal, depth, and time available. Fewer well-probed questions are usually better than a long checklist.
Ask for the story, then follow it
Open-ended questions create room. Specific past events give that room a useful boundary. Follow-up probes turn the first answer into evidence.
Review your guide in that order: past event, open prompt, adaptive probe, closed clarification. That sequence will teach you more than replacing every did with how and calling the guide finished.