1. Prepare
Choose the interview job, participant context, and practical session constraints. Write the decision, key unknown, and evidence needed before treating the guide as ready.
Prepare a customer interview
User Interview Script Generator (Free)
Turn a product-discovery goal into a structured B2B or B2C discussion guide with behavior-first questions, follow-up probes, leading-question guardrails, evidence notes, and an explicit post-interview handoff.
This is a deterministic script builder, not an AI research analyst. It creates the research instrument; your interviews create the evidence.
Before you generate: define the research frame
Decision
What product decision should this research inform?
Unknown
What uncertainty blocks that decision?
Participant
Who has relevant first-hand experience?
Evidence needed
What story or observation would reduce the uncertainty?
The generated research frame also leaves explicit placeholders for these fields when they are not supplied. Fill them before running the interview rather than treating a generic script as the research objective.
No login. Deterministic generation runs in your browser.
The research frame and editable guide are saved to this browser so you can continue later. Share URLs contain instrument context, so keep participant names, transcripts, recordings, and other sensitive raw research data out of them.
1. Research frame
Start with the decision and uncertainty. The tool uses these fields to judge readiness and keep the guide focused.
Script Readiness Check
Decision is missing or too broad
The guide cannot stay focused if the Product choice this research should inform is unclear.
Try: Write the choice the team expects to make differently if the evidence changes.
Key unknown is too broad
A broad learning goal produces a wandering script and makes it hard to know what evidence matters.
Try: Name one uncertainty that currently blocks the decision, without writing the conclusion into it.
Evidence need is vague
The instrument should make clear which story, observation, artifact, or decision sequence could reduce the uncertainty.
Try: Describe the real event or behavior you need to reconstruct, not a generic request for feedback.
Participant definition needs first-hand experience
A job title or generic segment does not prove that the participant experienced the workflow or decision you need to understand.
Try: Specify the role or segment plus the relevant recent experience or lifecycle context.
This is a deterministic checklist, not a research-quality score and not a claim that the interview is bias-free.
2. Adapt the guide
Interview context
Time budget
Session mode
Question Quality Lab
Paste one custom question. The deterministic checker flags common wording risks; it does not claim to detect every source of interview bias.
3. Runnable working guide
Edit the working copy after generation, or switch to Run Interview for a cleaner live view. Guide progress is not a mandatory checklist.
You can remove, rewrite, or reorder prompts. For problem discovery, keep concept or solution blocks after behavioral reconstruction unless the research job intentionally changes.
4. Question → evidence trace
Every planned prompt exists to surface specific evidence for the decision above, not to be completed as a checklist. Follow the participant's story, skip prompts the story already answered, and cut from the bottom first when time runs short.
| # | Prompt | Purpose | Evidence sought | Probe if useful | If time runs short |
|---|---|---|---|---|---|
| 1 | Tell me about the last time this situation happened. Start from the beginning.Core | Anchor on a recent real event | Trigger, sequence, and context | What triggered it? | Keep unless the story already answered it |
| 2 | Walk me through what you did from the trigger to the outcome.Core | Reconstruct the full sequence | Steps, actors, and handoffs | Where did the normal flow change? | Keep unless the story already answered it |
| 3 | Tell me about the last time the normal process did not work. What did you do instead?Core | Surface breakdowns and workarounds | Failure point and workaround | How did you start using that workaround? | Keep unless the story already answered it |
| 4 | What happened because the problem was not solved cleanly?Core | Expose consequences and cost | Consequence and cost of inaction | What changed for you or the team? | Keep unless the story already answered it |
| 5 | Walk me through the workaround or alternative you used and what led you to it.Core | Understand current alternatives | Alternative and selection reason | What did it solve well? | Keep unless the story already answered it |
| 6 | Who else was involved, and where did handoffs or constraints matter?Core | Map actors and constraints | Roles and handoff friction | What did each person need? | Keep unless the story already answered it |
| 7 | When does this matter a lot, and when is it easy to ignore?If time | Find severity conditions | High-stakes and low-stakes contexts | Can you give me one example of each? | Cut first |
| 8 | What have you tried before, and what made you keep, change, or abandon that approach?If time | Learn from past attempts | Tried approaches and change reason | What evidence changed your mind? | Cut first |
After the session, carry each row's evidence into the snapshot: what was observed, in what context, what you inferred, and what remains open, then update the decision.
Loading saved interview state...
How it works: Prepare → Run → Synthesize → Decide
2. Run
Use the guide as scaffolding. Stay with relevant real stories, leave room for silence, and use probes when they clarify the research question.
3. Synthesize
Capture observation and context separately from interpretation. Across interviews, inspect recurring mechanisms, contradictions, and meaningful segment differences.
4. Decide
Turn supported patterns into problem framing, prioritization input, another research method, or no action. Do not jump from one quote directly to a feature requirement.
B2B vs B2C: separate workflow from buying context
B2B interviews often require you to distinguish the person doing the work from champions, approvers, procurement, security, or budget owners. The B2B module covers those roles only when they matter. B2C research may have a simpler buying process, but the same evidence rule applies: reconstruct what happened instead of assuming that stated preference predicts future behavior.
Avoid leading questions
Leading questions insert the answer, sell the solution, or ask participants to predict behavior without context. The generated script includes weak-to-strong rewrites. For example, replace “Would you use a feature that does X?” with “Tell me about the last time you had to do X. What happened?” There is no universal percentage of questions that must use a particular syntax; use the wording that produces relevant evidence for the research job.
Behavior-first interviewing
Recent-event prompts make it easier to inspect a trigger, sequence, workaround, constraint, consequence, and decision in context. Hypothetical questions can still be useful for some research jobs, especially concept exploration, but do not confuse a participant saying what they might do with evidence that they actually behaved that way.
After the interview: evidence before themes
The output now uses an evidence-first note model: Observation → Context → Interpretation → Open question. The post-interview snapshot then separates Evidence → Inference → Assumptions → Decision implication. This keeps a memorable quote from silently becoming a universal fact or a roadmap requirement.
Pro tips
- Write the product decision and key unknown before you write questions; the interview exists to reduce uncertainty around that decision.
- Recruit people with first-hand experience relevant to the unknown, not merely people with the right job title.
- Start from a recent real event before asking for general opinions or future predictions.
- Use plain language you would actually say aloud and avoid internal framework vocabulary unless the participant uses it first.
- Leave space for silence and unexpected evidence instead of rushing through every question in the guide.
- Keep the user workflow separate from the buying process when B2B roles and decision rights differ.
- Pilot the guide with a teammate listening for leading wording, hidden assumptions, and unclear probes.
- Capture observation and context separately from your interpretation so synthesis does not turn inference into fact.
- Preserve contradictions and segment differences instead of forcing every interview into the same theme.
- After the session, state what changed in the decision, what remains assumed, and which evidence would reduce the next uncertainty.
Common mistakes
Participants give generic opinions but few useful examples.
Why: The guide asks what people think or would do instead of reconstructing a real event.
Fix: Start with a recent story, then probe the trigger, steps, workaround, consequence, and decision context.
People praise the idea, but the team still does not understand current behavior.
Why: The interviewer is pitching a feature or value proposition during problem discovery.
Fix: Return to current behavior and real examples. If concept reaction is the research job, make that objective explicit instead of mixing methods.
B2B evidence mixes end-user pain with buyer, procurement, and budget concerns.
Why: The research does not distinguish workflow roles from buying roles.
Fix: Use the B2B module only when relevant and label whose behavior or decision each observation describes.
The team cannot explain what decision the interview was supposed to inform.
Why: The script begins with questions instead of a research decision and key unknown.
Fix: Write the decision, unknown, participant, and evidence needed before running the session.
Notes blur participant statements with the researcher's conclusions.
Why: Free-form notes do not distinguish observation, context, interpretation, and open questions.
Fix: Use the evidence-first note table and keep interpretations explicitly labeled until cross-interview synthesis.
A memorable quote becomes a roadmap requirement.
Why: One interview is being treated as a market conclusion rather than evidence from one context.
Fix: Preserve the quote and context, compare it with other evidence, and route credible patterns into problem framing before solution prioritization.
Different interviewers ask the same topic in leading or inconsistent ways.
Why: The shared guide lacks probes and examples of wording to avoid.
Fix: Use the common probes and leading-question rewrites as guardrails while still following relevant participant stories.
Recordings and notes accumulate without affecting product choices.
Why: Research is not connected to a decision, synthesis step, or explicit remaining uncertainty.
Fix: Close each snapshot with evidence, inference, assumptions, contradictions, and the decision implication or next research method.
FAQ
Is this user interview script generator free?
Yes. The generator is free to use in the browser and does not require a login. You can change the research context, interview goal, participant type, duration, and other inputs, then copy or export the generated guide. The tool creates a deterministic research instrument; it does not conduct the interview or manufacture research findings for you.
Is this for customer research or Product Manager hiring interviews?
This tool is for customer and user research in Product Management, not for hiring candidates. Its prompts reconstruct customer behavior, workflows, buying decisions, churn, pricing context, concepts, and usability. If you are preparing for a Product Manager job interview, use CraftUp’s separate PM Interview Guide, question bank, and interview simulator instead.
Can I use the generator for both B2B and B2C interviews?
Yes. Both flows share the same evidence-first foundation, but B2B now separates workflow roles from buying and approval roles. Buying questions appear when the research job or context actually makes them relevant. B2C interviews usually need less organizational decision structure. Do not assume every B2B session requires procurement, security, budget, or approver questions.
How does the tool help avoid leading questions?
The script uses recent-event and workflow prompts, plus deterministic checks for common wording risks such as assumed pain, future-intent predictions, feature ranking, solution leakage, double questions, and internal jargon. These guardrails reduce avoidable leading language, but they cannot remove recruiting, interviewer, interpretation, confirmation, social-desirability, or recall bias from the research process.
Why use this instead of copying a generic list of customer interview questions?
A generic list can teach useful wording patterns, but the final guide should match a specific research job. This generator starts with Decision, Key Unknown, Participant, and Evidence Needed, then adapts core questions, optional depth, probes, B2B context, notes, and the post-session snapshot. Use the Question Guide for examples; use this tool when you need a runnable instrument.
Can I share an interview script with my team?
Yes. The Share action creates a URL snapshot that reconstructs the research frame and current edited guide. Local autosave also lets you continue in the same browser. Treat the link as an instrument, not a research repository: keep participant names, transcripts, recordings, raw notes, and other sensitive research data out of share URLs and in systems approved by your organization.
How should I use this for churn interviews?
Choose the churn goal and recruit people who can reconstruct a real cancellation journey. The guide asks about first doubt, attempted recovery, expectation gaps, the final trigger, and what happened after cancellation. During synthesis, compare mechanisms and contexts rather than simply counting complaints. A churn interview can inform a decision, but it does not prove one universal reason explains every cancellation.
Does this support pricing and willingness-to-pay research?
Yes, but it treats interviews as qualitative buying evidence rather than a precise price estimator. The pricing flow reconstructs comparable purchases, value anchors, budget ownership, alternatives, approvals, and purchase constraints. If you need to estimate a market-wide willingness-to-pay distribution or causal pricing effect, combine interviews with an appropriate quantitative or experimental method rather than relying on interview count alone.
What should I do with the notes and post-interview snapshot?
Use the note structure to separate observation, context, interpretation, and open questions. Then complete the single-interview snapshot with strongest evidence, contradiction, current inference, remaining assumption, decision update, and next evidence needed. Across interviews, compare recurring mechanisms and meaningful differences with the Insight Clustering Helper. Only after synthesis should credible patterns feed problem framing or prioritization.
How does this connect to CraftUp’s discovery workflow?
The generator is the preparation and run-instrument step, not the whole discovery process. Learn the Customer Interview method first, prepare and pilot the guide here, run sessions, cluster observations with the Insight Clustering Helper, then turn supported patterns into a Problem Statement or another research step. If several credible opportunities remain, use Prioritization instead of ranking problems by interview mention count.
Learn the method around the tool
The generator prepares the conversation. Use the customer-interview guide for conduct and synthesis, the specialist guides for question wording and sample-size decisions, and the real CraftUp discovery courses when you want a structured learning sequence.
Do not stop at the script
Run the interview, preserve evidence, synthesize patterns and contradictions, then choose the next product decision explicitly.
Freshness
Last updated: 2026-09-23
- Exposed Decision, Key Unknown, Participant, and Evidence Needed as first-class builder inputs instead of hidden engine fields.
- Added deterministic Script Readiness and Question Quality checks without pretending to score or eliminate research bias.
- Made duration and B2B adaptation operational: shorter guides prioritize core prompts, while B2B workflow and buying roles are separated by research job.
- Added editable and live run views, ephemeral evidence-first notes, edited-guide share/export, and clearer privacy boundaries for share URLs.
- Added a question → evidence decision trace: every planned prompt now carries its purpose and the evidence it is designed to surface, in the UI and exports.