TLDR
Useful customer discovery questions help a founder understand a real situation: what someone was trying to accomplish, what happened, how they handled it, and what would make a different approach worth considering. They should produce an account you can examine, rather than a collection of compliments about your idea.
For a bootstrapped or pre-seed team, start with one decision. Perhaps you need to choose a customer segment, understand a stalled workflow, or decide which assumption to test before building. Recruit people who have recently done the relevant work. Ask them to reconstruct an actual example before discussing your product.
This guide provides twelve core questions, follow-up prompts, a thirty-minute interview structure, an evidence record, and worked examples. Choose the prompts that answer your research question; do not turn the conversation into a checklist of everything you could ask.
Keep three things separate in your notes: what the participant said, what you directly observed, and what you think it means. A recurring problem deserves further investigation. It does not automatically establish a market, a workable price, or a commitment to buy.
Finish each round with a decision and a next experiment. The value of discovery is the change it makes to your work: a narrower customer definition, a revised onboarding flow, an abandoned feature, or a more focused test.
Quick Verdict
A good customer interview is a practical investigation. You are trying to understand the conditions under which a person makes a choice, including the reasons they might reasonably keep their current approach.
Prepare enough to know what you need to learn, then follow the participant's account. If they describe an unexpected handoff or exception, explore it before returning to your planned questions. That detail may explain why your original product assumption is incomplete.
Use a simple document and a small evidence log at first. More software will not improve an interview that starts with a pitch and ends with an unsupported conclusion. A shared workspace becomes helpful when several people need to review the same evidence and act on the resulting decisions.
The question to answer after a call is: “What do we understand better, and what should we do differently?” If the answer is only that the participant liked you, the conversation needs a sharper purpose.
What Are Customer Discovery Questions?
Customer discovery questions are prompts used to investigate a potential customer's work, circumstances, priorities, and choices. They help you test assumptions about a problem and the people affected by it before committing more time to a particular solution.
Imagine a founder considering an approval tool for small creative studios. The useful starting question is not whether studios want faster approvals. It is how a recent approval actually happened: who requested it, which version was reviewed, where comments arrived, and what happened when someone did not respond.
The answer could reveal a document problem, an ownership problem, a client-expectation problem, or no important problem at all. Each finding points toward different work. Asking only about the founder's preferred solution would make those alternatives harder to see.
Customer discovery is also broader than an interview. Harvard Business School's Customer Discovery Basics treats it as an iterative process and cautions against relying on interviews alone. Conversation can guide a later observation, prototype, or small real-world test.
Keep discovery distinct from other conversations. A sales meeting asks whether a particular offer fits a buying opportunity. A usability session examines how someone uses a design. A support call resolves an immediate issue. You can learn across all three, but record the context so you do not mistake a support complaint for a purchasing commitment.
Write the Decision Before Writing the Questions
A broad instruction such as “learn about customers” gives you little help choosing participants or deciding when a conversation was useful. Write the decision your next round of interviews is meant to inform.
For example: “We need to decide whether independent studio owners or account managers are the better starting audience for our approval workflow.” That decision creates specific questions about responsibility, recent incidents, and who can change the process.
Use a short research brief with five lines:
- Decision: The choice the founders need to make.
- Current assumption: What you believe happens and for whom.
- Missing evidence: What you need to understand before choosing.
- Contradicting evidence: What would make you reconsider the assumption.
- Next action: The kind of experiment or decision the findings could support.
The fourth line is easy to skip. Keep it. If you believe studio owners lose work because approvals are delayed, evidence that delays are rare or harmless should change your conclusion. Decide that before a friendly conversation makes the idea feel more certain.
Keep one brief for each round. When the question changes, date the new version instead of silently rewriting the old assumption. This gives both founders a shared account of what changed and why.
Recruit People Who Can Describe the Relevant Work
Define participants by experience, not only by company size or job title. “A founder at a small agency” is a category. “A founder who personally handled a client approval in the last month” identifies someone likely to have a concrete example.
A short recruitment screen can ask what role the person had, when they last did the task, and what type of team or customer was involved. Keep these questions neutral. Asking whether they suffer from your proposed problem can recruit only the people who already agree with your framing.
Include useful differences deliberately. A studio with recurring retainers may handle approvals differently from one doing one-off projects. Interview both if the distinction matters to the decision, then keep the groups separate during analysis. Mixing them into one average story can hide the actual opportunity.
Tom Eisenmann's Early Customer Research highlights the risks of relying mainly on convenient contacts and the need to understand the people involved in a purchasing decision. A willing friend can help you practice, but their enthusiasm does not establish that they represent your intended customer.
Make the invitation honest and specific. Explain the workflow you are studying, why their experience is relevant, and how much time you are requesting. If you are building a related product, say so. Do not disguise a sales call as independent research.
A possible invitation is: “I'm researching how small studios manage client approvals. You mentioned that you oversee delivery, so your experience would be useful. Would you be open to a twenty-five-minute conversation about a recent project? I'm building in this area, but the conversation is about how you work today.” Adapt it to the actual person and channel.
Open the Interview Without Setting the Answer
At the start, explain the purpose, the time available, who is present, and how you intend to use the notes. Make it comfortable for the participant to correct your assumptions or decline a question.
For example: “I'm trying to understand how this work happens today. There isn't a right answer, and it is useful to hear when the current process works well. I'd like to discuss one recent example, then ask a few questions about the choices around it.”
Ask before recording or bringing an automated note-taking service into the conversation. Give the participant an alternative if they would rather continue without it. Do not ask them to expose confidential customer records simply to make your notes more detailed.
The GOV.UK guide to in-depth interviews recommends preparing a discussion guide, explaining the research, obtaining informed consent, and allowing the conversation to develop naturally. For a small startup, that translates into a clear opening and a flexible set of topics rather than a rigid interrogation.
Twelve Customer Discovery Questions to Build Around
The following questions form an original working guide you can adapt. Each has a specific purpose, a possible follow-up, and a decision it might inform. You rarely need every question in one call. Choose a few that address the assumption you are testing and leave time to explore the answers.
1. What was happening when this task last became necessary?
Begin with the trigger. A request might have come from a customer, an internal deadline, a new employee, or a recurring operating routine. The trigger tells you when the work enters someone's day and what makes it relevant.
Follow up with: “What happened immediately before that?” If the participant starts describing the usual process, gently return to the specific occasion. The difference between a routine trigger and an unusual emergency may change the audience you should serve.
For the approval-tool example, a customer revision after a deadline is a different starting point from a routine weekly review. Record which situation you heard about instead of combining both into “approvals are difficult.”
2. What result were you responsible for delivering?
A task label does not always reveal the outcome. Someone may say they were “sending a report” when their real responsibility was helping a client decide whether to continue a project.
Ask: “Who needed that result, and what would have counted as finished?” This helps distinguish the person doing the work from the person judging it. It also shows whether your proposed improvement addresses something the participant actually owns.
Listen for competing expectations. The account manager may need a clear answer from the customer, while the designer needs comments tied to a particular version. A tool that helps one role but confuses the other may fail in everyday use.
3. Can you take me through that occasion from start to finish?
Let the participant describe the sequence in their own order. Capture the people, tools, handoffs, and decisions that appear. Ask for clarification when a step is unclear, without supplying the explanation you hoped to hear.
A useful follow-up is: “What did you do next?” Another is: “Where did you find the information you needed at that point?” These prompts reveal transitions that a high-level summary can skip.
You are looking for a workable account, not perfect recall. Mark gaps as gaps. If an important step is uncertain, a later walkthrough of a redacted example may be more useful than pressing the person to guess.
4. Which part required the most attention, and why?
This leaves room for the participant to identify the difficult part instead of accepting your chosen problem. They might name an unexpected coordination issue, or say that the task is straightforward.
Ask what they had to check, repeat, explain, or wait for. Then ask what happened when that step went smoothly. Comparing a difficult instance with a normal one can reveal the conditions that create the difficulty.
A founder building version control might discover that the files are already clear but no one knows who can approve them. That is useful contradictory evidence. Preserve it even if it points away from the feature you intended to build.
5. What did you use to handle it?
Ask about the whole current approach, including software, documents, messages, and manual checks. “We use a spreadsheet” is the start of the answer. The important detail may be that someone also checks three inboxes before updating it.
Follow up with: “Which part of that setup works well enough that you would want to keep it?” Existing solutions often contain advantages your product would need to preserve: familiarity, flexibility, a shared history, or a customer's established habit.
Record the workaround as a sequence, not a dismissive label. A manual process can be a rational choice when the task is infrequent or exceptions are common.
6. What happened because the task took that path?
Move from frustration to consequence. Did another person wait? Was work repeated? Did a decision change? Did the participant simply finish later than expected? The consequence helps you judge the importance of the problem without inventing its cost.
If a person offers a time or money estimate, ask how they arrived at it. Record whether it comes from a tracked measure, a specific example, or a rough impression. Those are different kinds of evidence.
Do not turn “it felt like half my day” into a precise savings claim. A product experiment might later measure the difference, but the interview alone has not done that work.
7. When does this go differently?
Exceptions can be more informative than another description of the same frustration. Ask about a recent occasion when the task was easy, did not happen, or was handled by someone else.
Follow up with: “What was different about the customer, project, or team that time?” You may discover that the problem appears only with a particular contract type, a missing role, or a handoff between companies.
That distinction can improve your positioning and product scope. “Studios need approval software” may become “studios handling several external approvers need a clearer final decision.” The narrower description gives you something more specific to test.
8. Have you tried to change this process before?
Ask what prompted the attempt, what the person tried, and what happened afterward. An abandoned change can reveal a requirement that is invisible in a feature wish list.
Follow up with: “What made you return to the previous approach?” Perhaps customers did not adopt the new portal, historical records were difficult to move, or the person responsible left. A replacement must account for those conditions.
Also allow “no” to be meaningful. The participant may tolerate the problem because other work matters more. Ask how they decide what is worth changing, rather than assuming that a lack of action proves they need more education about your product.
9. How was the current tool or approach chosen?
This question investigates an actual decision. Ask who proposed the change, who evaluated it, who approved the spending, and who had to use the result. In a tiny team, one person may hold several roles.
Follow up with: “What had to be true before you could start using it?” The answer might include a trial, a customer request, an integration, a colleague's recommendation, or a particular budget window.
Separate the participant's own experience from their guess about another person's reasoning. If they were not involved in approval, the next useful step may be a conversation with someone who was.
10. What else competed for attention when you considered changing it?
A problem can be real and still lose to more pressing work. Ask what else was happening and how the team chose among those priorities. This helps you understand urgency in context.
A small company might postpone a workflow improvement while delivering a major customer project. Another might prioritize it because the same project exposes repeated failures. The surrounding situation matters more than a generic urgency score.
Listen for a concrete event that changes the decision: another client, a new teammate, a renewal, or a repeated exception. Do not interpret every busy participant as uninterested, or every frustrated participant as ready to buy.
11. What would make replacing the current approach difficult?
Invite the participant to explain switching costs in their own terms. Existing data, customer habits, setup effort, permissions, training, and uncertainty about a new vendor can all matter.
Ask: “Which part could your team change itself, and which part depends on someone else?” A founder may be able to adopt a tool internally but unable to require every customer to create another account.
This question can change your first product experience. The appropriate experiment may be a smaller workflow that fits around existing tools, rather than a complete replacement. Treat that as a hypothesis to test, not an automatic roadmap decision.
12. What should I understand next to avoid misreading this?
End by giving the participant a chance to correct the picture. Ask whether you missed a role, an unusual condition, or a reason the current process makes sense.
Where appropriate, ask if they would be comfortable with a specific follow-up: clarifying one point, reviewing a redacted example, or introducing someone involved in the decision. Explain why that step would help.
A refusal is not a failed interview. They may have limited time or be unable to share more. Record the boundary and the learning you already have. Do not turn an agreed research follow-up into permission for unrelated marketing.
Rewrite Questions That Invite Easy Agreement
The wording of a question can quietly supply the answer. Before a call, read your prompts and look for assumptions hidden inside them. Replace the assumption with something the participant can describe or correct.
Instead of: “Would an automated approval tool save you time?” Try: “During the last approval, which steps did you handle yourself?” The first asks the person to imagine your desired benefit. The second helps you locate the work before deciding what should change.
Instead of: “How frustrating are scattered client comments?” Try: “Where did the comments arrive, and what did you do with them?” Maybe the comments were scattered but easy to manage. Maybe the difficult part came later.
Instead of: “Would you pay thirty dollars for this?” Try: “How did you decide what to spend on the approach you use now?” This informs your understanding of a past decision. Testing a particular offer still requires a separate, concrete commercial conversation.
Instead of: “Which features should we build?” Try: “What were you trying to accomplish when you wanted that feature?” A requested dashboard could mean several different jobs. Understand the job before deciding whether a dashboard is the answer.
Neutral wording does not remove every source of bias. Your role, recruiting choices, and reactions also shape the conversation. The aim is to make your assumptions easier to detect and your conclusions easier to inspect.
A Thirty-Minute Interview Structure for a Small Team
Use this as a starting agenda rather than a required research duration. A complicated workflow may need a longer session or a separate observation. Agree on the time with the participant and keep the scope realistic.
- Minutes 0–3: Explain the purpose, confirm how notes will be used, and understand the participant's role.
- Minutes 3–15: Reconstruct one recent occasion, from the trigger to the result.
- Minutes 15–22: Explore the current approach, consequences, exceptions, and earlier attempts to change it.
- Minutes 22–27: Discuss the actual decision process and any barriers relevant to your research question.
- Minutes 27–30: Check your understanding, invite corrections, and agree on any useful follow-up.
If one founder interviews while another takes notes, agree on how the observer will contribute. A quiet handoff near the end gives them room to ask a missing question without turning the session into two people interrupting the participant.
A solo founder can use the same structure. Take brief notes during the call and leave time immediately afterward to expand them. If something was not captured accurately, label it as a paraphrase rather than rebuilding a polished quotation from memory.
When the conversation moves into product evaluation, name the transition. “We have covered the current process. Would it be useful to discuss the approach we're exploring?” Keep that later feedback separate from the earlier account of how the person works today.
Worked Example: An Approval Problem That Changes Shape
The following conversation is fictional. It illustrates how a founder can move from a broad claim to a more specific learning decision.
Participant: “Client approvals are always a mess.”
Founder: “Which project are you thinking about most recently?”
Participant: “A website launch last Thursday. We had comments in email and in the design file.”
Founder: “What did you do when those comments arrived?”
Participant: “I copied them into our task list. That part was fine. The problem was that two people at the client disagreed, and neither was sure who could make the final decision.”
The original statement could have encouraged the founder to build another comment aggregator. The sequence reveals a different possibility: collecting feedback worked, but the final decision had no clear owner.
The founder continues by asking whether that situation happens on every project. The participant explains that repeat customers usually name an approver at kickoff, while new customers sometimes do not. The problem now has a condition: an early expectation was missing.
A reasonable next experiment is small. The founders could test an explicit approval-owner step with a few relevant projects before building more software. They would still need to learn whether the step is adopted and whether it changes the outcome.
The interview has not proven that every studio needs a new product. It has helped the team avoid treating an ambiguous complaint as a complete specification.
Worked Example: A Strong Complaint with a Weak Buying Case
Consider a second fictional participant who dislikes preparing monthly client summaries. They spend an afternoon collecting information and call the process tedious.
Further questions reveal that the founder only prepares four summaries a month, has a working template, and uses the exercise to notice issues in the accounts. They tried a reporting tool but returned to the manual process because checking its output took similar effort.
That does not mean the problem is imaginary. It means the proposed improvement has to preserve something valuable: the founder's understanding of each account. Removing the work entirely might remove a useful review habit.
The startup could explore assistance with gathering inputs while leaving judgment with the founder. Alternatively, it might learn that a team preparing forty summaries has a different need and deserves a separate research round.
Record both possibilities without declaring either one correct. The useful output is a clearer distinction between customer situations and a focused next question. Strong language about a task is one part of the evidence, not a purchasing decision.
Keep an Evidence Record That Another Founder Can Read
After each interview, complete a short record while the details are fresh. The record should let someone who missed the call understand what happened without inheriting your confidence automatically.
- Interview reference and date: A simple identifier and the date of the conversation.
- Relevant context: The participant's role, team situation, and experience with the task.
- Specific occasion: The event discussed and how recent it was.
- Sequence: The main steps, handoffs, and tools.
- Consequence: What changed because of the problem, including how any estimate was obtained.
- Current strengths: What the existing approach does well.
- Decision context: Who can change the process and what else matters.
- Evidence: Exact quotations where captured accurately, observations, or clearly labeled paraphrases.
- Interpretation: Your explanation, kept separate from the evidence.
- Open question and next action: What remains unknown, who owns the follow-up, and when it is useful.
Do not turn this into a permanent collection of every personal detail mentioned. Keep what serves the research purpose, limit access appropriately, and avoid copying customer documents into a shared workspace without permission.
A practical test is to ask your cofounder to read one record and state the conclusion. If they cannot distinguish what the customer said from what you inferred, revise the record before using it to make a product decision.
Review Patterns Without Inventing Precision
Review interviews by the customer situation you set out to study. Put accounts of the same task beside one another and compare triggers, sequences, consequences, current alternatives, and reasons for inaction.
Use plain statements with counts when they help: “Three of five studio owners described uncertainty about who approves final work; two described a clear owner and no recent delay.” This is a description of those interviews. It is not an estimate that sixty percent of the market has the problem.
Look for disconfirming cases. If two people avoid the problem, investigate what they do differently. Their process may reveal a useful solution, an exclusion criterion, or a reason your proposed product would not be necessary.
Choose one next decision from the review. You might narrow the audience, revise the problem statement, inspect a workflow directly, test a prototype, or stop pursuing an assumption. Write what evidence led to that choice and what result would make you revisit it.
For commercial follow-through, keep the learning record connected to the relationship without treating every research participant as an active deal. The founder sales pipeline guide explains how to distinguish a contact from an opportunity with an agreed next step.
How Many Interviews Are Enough?
There is no interview count that automatically validates a startup. The useful number depends on how specific the audience is, how varied the task is, and what decision you are trying to make.
For a small initial round, you can choose a manageable batch and a review date. The number is an operating constraint that helps you begin; it is not a statistical guarantee. Review after the batch rather than collecting conversations indefinitely because more interviews feel productive.
If every interview produces a completely different workflow, check your participant definition. You may have combined several customer situations. If the same account repeats, ask whether you understand the conditions well enough to design a test, including examples that do not fit.
Move to a different method when the uncertainty changes. An interview can explain why someone hesitates. Observing setup can show where they struggle. A real evaluation can show whether the proposed workflow is adopted. A commercial offer can test a concrete buying decision. Keep the evidence from each method in its proper scope.
How Wave Supports the Discovery Routine
A useful discovery process needs a place for evidence and a way to turn the review into work. Wave can support the surrounding team routine without replacing the founder's judgment.
Wave's Meetings area includes Agendas. Use an internal review agenda to discuss the interviews, compare interpretations, and choose the next learning decision. The Meetings introduction explains the broader meeting workflow.
Wave CRM provides Contacts, Companies, and Deals. Keep the participant's relationship context in the appropriate record, and distinguish research contact from buying opportunity. An interview alone should not move someone into a sales stage that implies more intent than they expressed. See the Wave CRM guide for the product's current feature overview.
When the review produces work, Wave Projects provides tasks with ownership and due dates. An action might be “test the approval-owner step” or “clarify who evaluates the current tool.” The evidence record in this article is a process template, not a claim that Wave automatically creates those research fields or validates the conclusions.
Keep the first setup small: an evidence document, a recurring internal review, and clear next actions. Add structure when it solves a coordination problem the team actually has.
Use AI to Organize Notes, Then Check the Source
AI can help organize notes you are permitted to use, but it should not supply missing customer evidence. Give it a bounded task and preserve the original notes so a teammate can check the result.
A useful prompt is: “From these notes, separate the participant's reported events, direct quotations, interviewer observations, and possible interpretations. List unanswered questions. Do not infer budget, authority, urgency, or commitments that are not stated.”
Review the output against the source. A fluent summary can make an uncertain observation sound settled or combine two participants into one story. Do not publish an AI paraphrase as a quotation, and do not let a generated priority score replace the founders' explanation of the evidence.
Common Mistakes That Make Discovery Less Useful
Explaining the solution too early. A detailed pitch can change the rest of the conversation into feedback on your framing. Establish the current situation first and label any later product discussion.
Interviewing only people who are easy to reach. Warm contacts are useful, but check whether they recently did the relevant work and whether your sample excludes important customer situations.
Treating a feature request as the problem. Ask what the person wanted to accomplish and what happened without the feature. A requested interface may be only one possible answer.
Ignoring what already works. A replacement that removes familiar advantages can create more work than it saves. Understand why the current approach survives.
Converting small samples into market claims. Keep counts and context visible. Your interviews can guide the next experiment without becoming a reliable market percentage.
Collecting notes without making decisions. Set a review date and choose an owner for the next action. Otherwise the research becomes a growing archive that has little effect on the product.
A Seven-Day Plan to Start
Day one: define the decision. Write the research brief, choose the participant criteria, and select the few prompts most likely to reveal the missing evidence. Rehearse the opening with a teammate to catch unclear wording.
Days two and three: recruit and schedule. Contact a small set of relevant people through appropriate channels. Explain the purpose and time request. Keep track of which customer situations your participants represent so convenience does not quietly define the whole sample.
Days three through five: conduct the first conversations. Leave time after each call to complete the evidence record. Improve unclear wording as you go, while noting material changes to the guide so later comparisons remain understandable.
Day six: review together. Compare the actual accounts, identify contradictions, and separate repeated evidence from strong individual opinions. Discuss what the current approach does well as carefully as what fails.
Day seven: choose the next experiment. Turn the review into one concrete action with an owner, a question to answer, and a review point. If recruitment takes longer, extend the schedule honestly; a calendar does not create evidence.
Before Your Next Interview
- Can you state the decision this conversation will inform?
- Does the participant have relevant, recent experience?
- Have you identified evidence that could contradict your assumption?
- Does the opening explain the purpose and handling of notes?
- Can your questions be answered without agreeing with your idea?
- Have you left time to explore one specific occasion?
- Will you capture current strengths as well as difficulties?
- Can you separate evidence, interpretation, and uncertainty?
- Is there a review date for deciding what happens next?
If several answers are unclear, simplify the research brief before adding more questions. A focused conversation that changes one decision is more useful than a long script that leaves the team with the same assumptions.
Final Takeaway
Customer discovery works best when the founder is willing to learn that the situation differs from the pitch. Ask about a real event, understand the current approach, investigate the decision around it, and preserve what remains uncertain.
Use the twelve questions as a flexible guide. The useful result is an evidence record that supports a specific next decision, not a score for how much the participant liked the idea.
When the work needs a shared home, explore Wave to organize the team review, customer context, and follow-through. Start with the smallest routine that helps the next conversation improve the next decision.
FAQ
What should I ask in a first customer discovery interview?
Ask about the participant's role and one recent occasion when they did the relevant task. Explore the trigger, desired result, sequence, current tools, consequence, and decision context. Choose follow-ups based on what they describe rather than trying to complete every question in a preset list.
How is customer discovery different from a sales call?
Discovery investigates a situation and the assumptions behind a product opportunity. A sales call evaluates a particular offer and a possible purchase. They can happen with the same person, but make the transition clear and record which evidence came before and after discussing the product.
Should I show my product during the interview?
You can, if product feedback serves the research goal and the participant agrees. First establish enough context to understand their current work. Label the product discussion separately so enthusiasm about a demonstration is not mistaken for evidence about past behavior or a commitment to adopt.
Can I interview friends or existing customers?
Yes, when they have relevant experience. Record that relationship and the context it creates. Friends can help you practice, while existing customers can explain actual use. Include people outside those groups when the decision concerns a new audience or the reasons someone has not adopted your product.
How should I ask about budget?
Explore a past purchasing decision: how spending was approved, who participated, and what alternatives were considered. Mark missing information as unknown. A response to a hypothetical price is different from acceptance of a defined offer, and someone who uses the product may not control the budget.
What if someone says they love the idea?
Thank them, then return to the situation you are studying. Ask what experience makes the idea relevant and whether they can describe a recent example. If a concrete evaluation is appropriate, discuss it explicitly. Keep a compliment, a research follow-up, and a buying commitment as separate signals.
Do I need to record every interview?
No. Choose a method that fits the conversation and the participant's preferences. Clear notes can be enough, particularly when you review them immediately afterward. If you record or use an automated note-taker, explain it and obtain permission first. Preserve uncertainty when wording was not captured exactly.
What should I do when interviews contradict each other?
Compare the participants' situations, roles, and recent examples before averaging their opinions. They may be describing different workflows or conditions. Use the disagreement to refine your customer definition or choose a follow-up investigation. Contradiction can improve the decision instead of being something to remove from the summary.
When should discovery lead to building something?
Build a focused test when you understand enough to state what it should teach you and how you will review the result. Keep its scope tied to the remaining uncertainty. Interviews can justify learning through a prototype or real workflow, but they do not guarantee adoption or commercial success.



