Marcus spent a month asking people about his idea for a mobile dog grooming van. He asked neighbors, coworkers, the regulars at his gym, and everyone in his family group chat. The question was always some version of the same thing: "If there was a grooming van that came to your driveway, would you use it?" Nearly everyone said yes. A few said they would tell their friends. One cousin said he should have done this years ago.
He bought a used van, outfitted it, and opened for bookings. In the first six weeks he groomed eleven dogs, and four of them belonged to the people who had cheered him on. The rest of the yeses never called. When he finally asked one neighbor why, the answer stung in its honesty: "Oh, I already have a groomer. She's fine. I just thought it was a cool idea."
Nobody lied to Marcus. They answered the question he asked, and the question he asked was about an imagined future, put to people who liked him and did not want to hurt his feelings. That combination produces cheerful, sincere, useless answers. Customer interviews work, and they are the cheapest research you will ever do, but only when the questions are built to get past politeness and into what people actually do.
Why "would you buy this" fails
People are bad at predicting their own future behavior and very good at being kind. Ask whether someone would use a product, and they picture an idealized version of themselves with more time, more money, and fewer existing habits. Then they look at your hopeful face and say yes. There is no cost to them for being wrong, so the answer carries no weight.
Past behavior is a different kind of evidence. If someone tells you about the last time they dealt with the problem, what they tried, what it cost, and what annoyed them about it, they are reporting something that happened. Memory is imperfect, but it beats prediction, and it reveals what you most need to know: whether this problem is painful enough that people already spend time or money trying to fix it.
That is the core principle of a good discovery interview. Talk about their life, in the past tense, and keep your idea out of the room for as long as you can. The moment people know what you are building, the conversation shifts from describing their world to evaluating your pitch, and evaluation is where the politeness creeps back in.
Decide what you need to learn first
Before you write a single question, write down two or three things you need to learn. Not things you hope to confirm. Things you genuinely do not know and cannot guess.
Priya was thinking about a bookkeeping service for food truck owners. Her learning goals were plain: how do these owners handle their books today, what part of it eats the most time or causes the most stress, and what do they currently pay, in money or hours, to deal with it. Notice that none of those goals mention her service. They are all about the customer's current reality.
Also decide who you are talking to. "Small business owners" is too broad to produce patterns, since a florist and a software consultant share almost nothing. Priya picked food truck owners who had been operating for at least a year, because new owners were still figuring out everything at once and would muddy the picture. A narrow segment makes it possible to hear the same problem described five different ways, which is exactly the signal you are looking for.
How to write an interview guide
An interview guide is a short list of open questions, usually somewhere around eight to twelve, that you use as a map rather than a script. Good discovery questions share a few traits. They start with what, how, when, or "tell me about," so the person has to describe something real. They ask about specific past events. They never mention your product. And they ask one thing at a time.
Here is what weak questions look like next to stronger versions of the same intent:
| Weak question | Why it fails | Stronger version |
|---|---|---|
| Would you use an app that tracks your receipts? | Hypothetical, and it pitches your idea | Tell me about the last time you had to find a receipt for your taxes. |
| How much would you pay for this? | Imagined prices mean nothing | What do you spend on this today, in money or in hours? |
| Isn't it frustrating to do your books by hand? | Leading, it supplies the feeling | How do you feel about the way your books get done now? |
| Do you have trouble with scheduling? | Yes or no, so you get one word back | Walk me through how you scheduled last week's jobs. |
| What tools do you use and how do you like them? | Two questions, they will answer the easier one | What tools do you use for this? (Then ask about each one.) |
A useful guide often moves in a rough arc. Open with context about their work or routine. Narrow to the last time the problem showed up. Ask what they did about it, what they tried before that, and what each attempt cost. Ask what is still annoying even after the workaround. Close with whether there is anyone else you should talk to, which is the fastest way to find your next interview.
Leave room in the guide for the question that matters most: "Why?" When someone says the spreadsheet is a pain, ask why. When they say they gave up on an app, ask what happened. The first answer is usually the surface. The second or third is where the real reason lives.
Running the conversation
Keep it short. Asking for "twenty minutes about how you handle your books" gets far more yeses than asking for feedback on your idea, and most people enjoy telling a story about their own work.
Talk less than you think you should. If you hear yourself explaining, defending, or describing your product, stop. Your job is to listen and to follow the energy. When someone's voice speeds up, or they sigh, or they say "oh, don't get me started," that is the moment to ask for more detail, not the moment to move to question four.
Watch for three kinds of answers that sound like evidence but are not. Compliments ("that sounds great") tell you about their manners. Generalities ("I usually," "I always," "I would") are predictions in disguise, so pull them back to a specific instance by asking about the last time. And feature requests ("it should have a calendar") are solutions, when what you need is the problem underneath. Ask what the calendar would let them do that they cannot do today.
Money is the hardest subject to raise and the most important. Do not ask what they would pay. Ask what they pay now, whether that is a subscription, a part-time helper, a spouse's Saturday, or the late fees they keep eating. Someone who spends real money or real hours on a problem has proven it matters to them. Someone who shrugs and says "I just deal with it" has told you something too.
Capture what they said, word for word
Take notes during the call, or record it with permission, and write up the interview the same day while it is fresh. The single most valuable thing to capture is verbatim quotes: the exact words the person used, in quotation marks, not your summary of what they meant.
Your summary is already filtered through what you hope is true. If Priya writes "owners find bookkeeping stressful," she has recorded her own interpretation. If she writes down what one owner actually said, "I do the sales tax at midnight in the truck because it's the only time it's quiet, and I'm always scared I got it wrong," she has a piece of evidence that will still be true in six months and that might end up word for word in her marketing.
For each interview, also note who they are (a role and situation, not necessarily a name), what they do today to handle the problem, what it costs them, how intense the pain seemed, and any signal about paying. Be honest about that last one. A clear no, a polite maybe, a verbal yes with no commitment, and a deposit or signed order are four very different levels of evidence, and lumping them together is how founders talk themselves into a market that is not there.
How many interviews is enough
There is no magic number, but there is a useful signal: you have done enough when the interviews stop surprising you. Early on, every conversation teaches you something new. Somewhere along the way, you start predicting what people will say before they say it, and the same three or four pains keep coming up in different words. That repetition is the pattern you were looking for.
For most small businesses, ten honest interviews within one narrow segment is a reasonable first bar, and plenty of founders find the picture is clear well before then. If ten conversations produce ten completely different stories, that is also a finding. It usually means your segment is too broad, and the fix is to narrow it and start again, not to keep interviewing the whole world.
Priya's turned out to be clarifying in a way she did not expect. Food truck owners did not much mind bookkeeping. What they dreaded, almost to a person, was sales tax across the different cities and events they worked, each with its own rules. She had planned a general bookkeeping service. The interviews pointed her toward something narrower and far easier to sell.
Turn interviews into evidence-backed personas
A persona is a short profile of a type of customer, and it is one of the most misused tools in business. The usual version gets written at a whiteboard: "Busy Brenda, 34, loves yoga and lattes, wants to save time." It feels useful and is almost entirely fiction. Made-up personas mislead because they are built from the founder's assumptions, so they confirm whatever the founder already believed, and they give that guess the authority of a document.
A real persona is built from the interviews, and every claim in it should trace back to a conversation. Start by laying out your notes and grouping people who described the same situation and the same main pain. Each group is a candidate persona. Then, for each one, write down:
- ›A descriptive handle based on their situation, such as "the multi-city food truck owner," rather than a cute name.
- ›The situation they are in when the problem shows up.
- ›The job they are trying to get done.
- ›Their main pains, each linked to the interviews where you heard it.
- ›Their current workaround and what it costs them.
- ›Two or three verbatim quotes that capture the problem in their words.
- ›What you learned about their willingness to pay, and what triggers them to look for help.
- ›The open questions your interviews have not answered yet.
That last item is the one founders skip, and it is the most honest part of the document. A persona that admits "we do not yet know who in the household makes this decision" is more useful than one that confidently invents an answer. It also tells you what to ask in your next round.
A good rule: if a persona is supported by only one interview, it is an anecdote, not a pattern. Hold it loosely until you hear the same thing again.
The mistakes that make interviews worthless
The most common mistake is the one Marcus made: interviewing friends and family about the future. They love you and they are guessing, which is the worst possible combination. Talk to strangers in your target segment, and ask about the past.
Pitching mid-interview is a close second. Many founders cannot resist. But the instant you describe your solution, you have turned a research call into a sales call, and every answer after that is shaped by politeness. A founder named Dev ran fifteen interviews for a tutoring marketplace and described his platform in the first five minutes of each one. He came away with fifteen opinions about his website and almost nothing about how parents found tutors.
Another trap is hearing what you want to hear. When an interview goes well, it is tempting to remember the enthusiastic parts and forget the "I already have something that works." Verbatim notes are the cure, because they keep the inconvenient parts on the page.
Finally, there is the founder who interviews endlessly and never acts. Once the patterns are clear, state your value proposition and test it.
This is where the Founder Workbook in PushStartGo was built to help. Its customer discovery step checks each interview question as you type and flags the usual problems, like yes or no questions, hypotheticals, leading phrasing, double questions, and any mention of your own idea. It gives you an interview log with a place for verbatim quotes, the current workaround, money spent today, pain intensity, and an honest payment signal. When you build personas, the workbook asks you to link each one to at least two interviews and checks that the quotes in it actually appear in your notes, so a persona cannot quietly drift into fiction.
Back to the van
Marcus did eventually get the grooming business working, and the turnaround started with better conversations. He stopped asking whether people would use a van and started asking dog owners to tell him about their last grooming appointment. He learned that most were content with their groomer, but a specific group was not: owners of large, anxious, or elderly dogs, for whom the car ride and the noisy salon were the real problem. Those owners were already paying extra for house calls when they could find them, which was rarely. He narrowed his marketing to them, and the van filled up.
The interviews he needed were cheap and available the whole time. He just had to ask about what people had done instead of what they might do.
This is the first step in validating an idea. Once your interviews show a clear pattern, the next moves are writing a value proposition that is not generic and testing it with a low fidelity prototype before you build. For the bigger picture of what to prove before spending money, see the seven questions every founder must answer. And when you want a place to write your guide, log each conversation, and build personas that hold up, start in the Founder Workbook. The customer discovery step is free.