If every interview ends with "Sounds great, let me know when it launches", you have not validated anything. You have learned that people are polite.
Market validation interviews are worth doing, but they answer one narrow question: is the problem real and painful enough to justify the next, more expensive test? They do not tell you whether people will buy your solution. They do not tell you how big the market is. They tell you whether a specific group of people has the problem now, what it costs them, and what they already do about it. That is a useful answer, and it is cheap to get.
This guide gives you three tools you can copy today: an evidence ladder that ranks what you hear, an interview guide that stays on the past, and a stop worksheet you fill in before the first conversation. It applies whether you are testing a product for customers, a service for partners, or a tool for your own staff.
Why compliments are the weakest evidence you have
A compliment costs the speaker nothing. "I'd definitely use that" costs nothing either. Neither one needs a memory, a number, or a decision.
An interview collects what people say, which is exactly why it needs discipline. Nielsen Norman Group is blunt about the problem: self-reported claims are unreliable, and so are people's guesses about their future behaviour.2
Psychology says the same from another direction. A review of the intention-behaviour gap begins from the plain observation that people do not always do the things they intend to do.3 So "I would use that" is a statement about a mood in the room. It is not a statement about next month.
Your job is not to stop collecting opinions. It is to rank them, and to spend most of your interview time on things that cost the speaker something to say.
The evidence ladder
Every sentence you hear in an interview sits on a rung. Write the rung next to your notes. The higher the rung, the more it cost the person to say or do, and the more weight it carries.
| Rung | What you hear or see | What it can support | What it cannot support |
|---|---|---|---|
| 1. Compliment | "Great idea." "I'd use that." | That the person was friendly | Anything about the problem or demand |
| 2. General claim | "Yes, that's a real problem for people like me." | The problem is recognisable | That it is frequent, costly, or urgent |
| 3. Specific recent story | "Last Thursday it happened again, and it took me two hours." | The problem happens, in what circumstances, at what cost | That it is typical, or that they would pay to remove it |
| 4. Current effort | A spreadsheet, a paid tool, a person hired, a monthly workaround they can show you | The problem already earns time or money from them | That your solution beats what they have |
| 5. Small commitment | An introduction to a colleague, access to real data, a booked follow-up, a place on a waiting list | Interest that cost them something | That they will pay |
| 6. Changed behaviour | A pre-order, a paid pilot, switching a real task to your manual service for a few weeks | The strongest signal on this ladder | A sales forecast. Nothing here is that |
If most of your answers sit on rungs 1 and 2, your questions are drifting toward the future. Rung 3 and rung 4 are where you earn your keep. Rung 5 can happen at the end of an interview, if you ask for a small step. Rung 6 needs an offer or a trial, which is the next test, not another round of questions.
The rule: only rungs 3 and above count toward your decision. Rungs 1 and 2 are the warm-up. Collect them, record them, and do not add them up.
If you want a wider view of what makes a finding trustworthy once you have notes, see what makes a customer insight evidence-backed.
An interview guide that stays on the past
Nielsen Norman Group names the weakness of interviews directly: you are asking people to remember past use or to speculate about future use.4 You cannot remove that weakness. You can narrow it by asking for one specific, recent occurrence instead of a general habit. The "last time" question does exactly that: the participant is asked to describe the most recent time.5 GOV.UK's guidance on in-depth interviews gives the same advice in plain words: focus on stories and real examples, and avoid generalities and talking about how things "should" happen.6
Here is a guide you can copy. Keep it to the order shown and do not describe your idea until the very end, if at all.
Market validation interview guide (30 minutes)
Before you start (2 minutes)
- Say who you are and why you are asking.
Say you are not selling anything.
- Ask if you can take notes. Follow your organisation's
consent and privacy rules.
1. Context (3 minutes)
- "Tell me about your role. What does a normal week involve?"
2. The last time (10 minutes)
- "When did [the problem situation] last happen to you?"
- "Walk me through it from the start. What happened first?"
- "What did you do next? And then?"
- "How long did it take? Who else got involved?"
- "What did it cost you, in time, money, or something else?"
3. What they do about it today (8 minutes)
- "What do you use to deal with this now?"
- "Can you show me? Share your screen or the file, with
anything private hidden."
- "What do you dislike about it?"
- "What have you tried and dropped?"
- "How much time or money does that setup take each month?"
4. Priority (4 minutes)
- "Where does this sit among the things you are trying to
fix right now?"
- "What would have to happen for you to change how you do it?"
5. A small step (3 minutes)
- "Is there someone else I should talk to about this?"
- "Could I follow up in a few weeks to hear how it went?"
Never ask
- "Would you use this?"
- "Would you pay for this?"
- "Do you think this is a good idea?"
- "How much would you pay?"
Three habits make this guide work.
Ask for the artefact, not the adjective. "It's a nightmare" is a rung 2 answer. "Here is the spreadsheet, it has six tabs, and Dana updates it every Friday" is a rung 4 answer. Follow every adjective with "Can you show me?"
Stay quiet about your idea. Once you describe your solution, treat everything that follows as politeness, not evidence. If you must show something, show it last, and record the reaction as a rung 1 comment.
Write down the rung as you go. Mark each quote 1 to 6. At the end, count how much of the conversation sat above rung 2. That count is your result, not the mood of the call.
For more on getting past a first, polite answer, see how to ask customer interview questions that move beyond surface answers. That article's follow-up ladder is about probing a single answer. The ladder above ranks what you hear.
Your real competition is what they do today
Do not stop at asking who the competitors are. Ask a better question: what is this person doing about the problem right now? It might be a paid tool, a spreadsheet, a favour from a colleague, or nothing at all.
GOV.UK's guidance for the early discovery stage tells researchers to find out how people do the task currently, including what services or channels they use, and to observe them to see what problems or barriers they face.7 That is your competitor analysis. It is also your best evidence.
Read the workaround as a price tag. Someone who pays for a clumsy tool, or loses an afternoon every month to a manual routine, has already shown you what the problem is worth to them. Treat "yes, it's annoying" from someone who has never tried to fix it as a warning, not as proof.
Two things follow. First, a problem with no workaround is a warning, not an opening. Before you count it against the problem, ask whether they tried and failed: a tool they tried and dropped shows the problem once cost them effort, but it is not a current workaround, so it does not count as rung 4. Second, your idea does not compete with other products. It competes with a habit, and you have to give people a reason to drop it.
Decide what would make you stop, before you start
Do not let a mixed result be called "promising" after the fact, or a cold one "the wrong segment". Write your decision rule before the first interview, while you are still neutral.
The idea is borrowed from research practice. The Center for Open Science describes preregistration as specifying your research plan in advance of your study, and says doing so helps to distinguish planned from unplanned work.8 You are not running a formal scientific study. You are borrowing the habit: write down what would change your mind while you are still neutral.
This worksheet sets the stop line for a single validation round. Choosing which study to run is a separate decision, and a link near the end of this guide covers it.
Copy this worksheet. Fill it in, date it, and show it to someone else before interview one.
What would make us stop? (fill in before the first interview)
Decision this serves
We will decide whether to [run a landing page test /
offer a manual pilot / drop this idea].
Who counts as the right person
[Role, situation, and how recently the problem must
have happened.]
How many conversations
We will hold [number] interviews with the right people,
then decide. We will not extend the number because we
like the result, except as set out under "If the result
lands in between".
Continue if (both must be true)
At least [number] of them describe a specific occurrence
within the last [period], with a cost they can name
(rung 3 or above).
At least [number] of them show a current workaround that
takes real time or money (rung 4).
Stop, or change the problem, if (either is enough)
Fewer than [number] describe a recent specific occurrence.
Fewer than [number] show a current workaround (rung 4).
If the result lands in between
Neither the continue rule nor the stop rule is met. We will
run [number] more interviews with only the group that had
recent occurrences. Then we apply one rule: continue only
if at least [number] of all participants show a current
workaround (rung 4). Otherwise we stop. Nothing else.
What we will not count
Compliments, "I would use that", and anything said after
we described our idea.
Signed and dated:
The middle outcome is the easiest one to leave blank, and it is where bias lives. Decide beforehand what an in-between result means. Do not leave it to be settled afterwards, when you already know which answer you want.
A worked example
Here is an invented example to show the method.
A two-person team wants to build a service that chases late invoices for freelance designers. Before any interview, they fill in the worksheet. The decision it serves is whether to offer a manual pilot: continue means offering one, and stop means dropping the idea. The right person is a freelance designer who invoiced at least one client in the last month. They plan twelve conversations.
They will continue if at least six describe a late payment in the last two months with a named cost, and at least four show a current workaround that takes real time or money. They will stop if fewer than four describe a recent late payment, or if fewer than two show a workaround. If the result lands in between, they will run five more interviews with designers who had a late invoice in the last two months. After those, they will continue only if at least six of all seventeen people show a workaround, and otherwise stop. They will not count compliments, "I would use that", or anything said after they describe the idea.
They run twelve interviews, and they show the idea only at the end of each call. Nine people say it sounds useful. That is nine rung 1 comments, so the team writes them down and sets them aside. Seven say late payment is a real problem, which is rung 2. Five describe a specific late invoice in the last two months, with an amount and the days lost chasing it, which is rung 3. Two of those five show a spreadsheet they update by hand and one pays for an accounting add-on, which is rung 4: three workarounds, one short of the four needed. One person offers to introduce a designer friend, a rung 5 act.
The result falls in the middle. Five recent stories clear the stop line of four but not the continue line of six. Three workarounds clear the stop line of two but not the continue line of four. The rule they wrote first says what happens next. They run five new interviews with designers who had a late invoice in the last two months, using the same guide and the same rungs. Those five meet the story count by how they were chosen, so only workarounds matter now: they need three more, which makes six of seventeen. The compliments did not influence the decision. The worksheet did, because it was written first.
When to leave interviews and run a behavioural test
Stop interviewing when the worksheet gives you an answer: the number you set, or the number you set plus the extra round, if the in-between rule applies. If new interviews start repeating the same stories before then, check that your participants are spread across the group you care about. Do not treat repetition as permission to stop early. Once the worksheet gives its answer, change the kind of evidence you collect. Nielsen Norman Group's advice on method choice is that nearly all projects would benefit from multiple research methods and from combining insights.1 Interviews tell you the problem is worth testing for. A behavioural test tells you whether anyone acts.
Three simple forms, in order of effort:
- A landing page with a real offer. Say plainly what you would deliver, for whom, and by when. Give visitors a real next step, such as joining a short list or booking a call. Count what they do, not what they say.
- A manual trial. Deliver the result by hand for a small number of people, with you playing the part of the product. People see the outcome, and you see what the job really involves. Tell them it is manual.
- A pre-order or paid pilot. Ask for payment or a firm booking against a clear promise. Only do this if you will deliver or refund. Follow your organisation's and your payment provider's rules for taking money.
Write a stop worksheet for the test too, with the same sections. Decide what result means "continue" before you launch.
Be honest in every version. Say what exists and what does not. Do not run a test that leaves people believing a product is ready when it is not.
To decide which study to run before this one, or after it, see which customer research study should you run.
What this will not tell you
This is the one place where the limits go, so read them as a set.
Interviews will not tell you how many people have the problem. Twelve conversations describe twelve people. A story is not a rate.
Even the best rung 6 result does not predict sales. Do not treat the people who pre-order as a sample of the market. They chose to act first, and the distance between a first signal and a lasting business is still ahead of you.
Even a "last time" story depends on memory, the weakness named earlier. That is why the ladder asks you to see the workaround and not just hear about it.
You can also be wrong about whether you asked the right people. If the problem is real but your participants are the wrong kind of person, you will get a clean "no" that means nothing. Write the participant description before you recruit, and check it against who you actually spoke to.
Finally, the stop worksheet reduces self-deception. It does not remove it. Someone who wants a yes can still set a low bar. Show your bar to a person who does not share your hopes.
Your next step
Today, fill in the stop worksheet and nothing else. Then book the first conversation with someone who had the problem recently, open the guide at "the last time", and mark each thing you hear with its rung. For an example layout, see the market validation use case page.
