Problem discovery: a practical guide to talking to people about their problems

A note before you start. I wrote this guide for myself, and I use it. Most of it comes from The Mom Test by Rob Fitzpatrick and YC's Startup School, plus what I've picked up so far. Some parts, like the message templates, I haven't fully tested yet. I'll keep updating this page as I learn what works and what doesn't. Treat it as a starting point.

Last updated: 4 October 2026

The short version

• Ask about things that already happened to them. Never ask about your idea.
• Ask for 15 minutes, one person at a time. Say "I'd love to learn from you", never "interview".
• They talk 80% of the time. You take notes and ask "why?"
• Give something back after the questions: a summary of what others said, a quick tip, or an intro.
• You're looking for a pain they already spend money or hours on. "That would be nice" counts for nothing.

What you're looking for

You want a painkiller: a problem they need to solve right now. A vitamin is nice to have, and nobody rushes to pay for it.

By the end of a conversation, try to know these things:

Start broad. Don't zoom in on one problem until you hear a strong signal.

Who to talk to

• Pick 2 or 3 groups for your first 10 conversations (for example accountants, physios and small shop owners). Ten conversations with ten different professions give you ten unrelated pains, and nothing repeats.
• Start with people who know you. Former colleagues, classmates and friends of friends reply far more often than strangers.
• Include the person who pays. In a company, the people doing the work feel the pain, but a manager usually approves the purchase. Talk to both.

How to get them to talk to you

• Message one person at a time. Never send a group message.
• Ask for 15 to 20 minutes, not an hour.
• Use the words "advice" or "learn from you". Avoid "interview", "survey" and "feedback on my idea".
• Say why you picked them.
• Fit their schedule. A call, a voice note or a chat all work.
• Follow up once after 3 days. If there's still no reply, let it go.

Message for someone you don't know well:

Hi [name], I'm building a product for [type of person]. You've been running [company] for a while, and I'd love to learn how you handle [area]. Could I get 15 minutes of your time this week? I'm happy to work around your schedule. I'm talking to about 8 people like you, and I'll send you a short note on how the others handle it too.

Message for a former colleague or friend:

Hey [name], I'm building my own thing now and trying to understand where [teams like yours] lose time. You're one of the sharpest people I've worked with. Could I buy you a coffee, or get 15 minutes on a call? No pitch, I just want to learn.

The first approach

Be honest, ask permission, and make "no" easy. Keep the first ask tiny: two questions, not "an interview".

At events (the easiest place, since people came to talk):

"What are you working on these days?"
"What's the hardest part of that right now?"
"When did that last bite you?"

If it gets interesting, ask: "Could I get 15 minutes with you later this week? I'm trying to understand exactly this."

At cafés (people on laptops, or the owner during a slow hour):

"Sorry to interrupt. I'm building something, and I'm going around learning how different people actually work. Can I ask you two quick questions? If it's a bad time, no problem at all."

At shops and clinics (go off-peak, or while you're a paying customer):

"How long have you been running this? What part of the business eats most of your time that isn't the actual work?"

What helps:

• Being a customer first makes the ask much easier.
• Saying "I'm learning, not selling" lowers their guard.
• With strangers, don't take notes during the chat. Write them on your phone right after.

Approaching a stranger

Pick someone who looks free: standing alone at an event, waiting in the coffee queue, or taking a break from their laptop. Leave people alone if they're on a call, or have headphones in and are typing fast.

Say who you are and what you want in one go, and give them an easy way out:

"Hi, I'm [name]. I'm building something for [type of person] and I'm trying to learn how people handle [area]. Can I ask you a couple of questions? If you're busy, that's totally fine."

Keep that last line. People relax when they know they can say no.

If they're busy, move on

Short answers, eyes going back to the screen, or "maybe later" all mean no. Say "No worries, thanks anyway" and walk away. Don't ask twice. A forced yes gets you a bad conversation, and you'll probably run into them at the next event.

Expect a few no's, so plan to approach more people than you need. A no is usually about their day, not about you.

What they get in return

1. Being heard. Almost nobody asks a busy person "what's hard for you?" and then really listens.
2. A summary. Half a page on how others in their field handle the same problem. This is the strongest thing you can give. Send it within 2 days.
3. A quick tip on the spot. A shortcut, a tool, or a 2-minute AI trick that fixes a small annoyance. This works especially well with non-technical people.
4. An intro to someone useful to them.
5. Coffee or lunch on you, for longer chats.

Two rules:

• Give after the questions, not before. If you help first, they feel they owe you and get polite about everything.
• Keep help under about 15 minutes. Anything longer turns into free consulting. If they want more, note that as a signal and ask: "Would you pay someone to set this up?"

During the conversation

• They talk 80%, you talk 20%.
• Use their exact words back: "You said X is the worst part. Tell me more."
• Take notes and let them see it (except with strangers, see above).
• Don't pitch. If they ask what you're building: "Still figuring out the problem. That's why I'm talking to people."
• End on time. They'll be happy to talk again.
• Say thanks the same day, and send the summary you promised.

Keep them talking

People talk most freely about themselves, so make the conversation about them.

• Start with their story. "How did you get into this?" or "What are you working on right now?" are easy to answer and warm them up.
• Follow their energy. When they speed up, complain, or laugh about something that went wrong, stay on it. That's usually where the pain is.
• Keep asking small follow-ups: "Why?", "What happened then?", "How did you sort it out?"
• Keep your own stories to one line. If you catch yourself explaining, stop and ask a question.
• Don't correct them or argue, even when you know a better way. You're there to learn how they see it.
• If they're stuck on something you can fix, write it down and help at the end, after your questions (see What they get in return). A quick fix then builds trust and often gets you a second conversation.
• Watch for signs they want to leave: checking their phone, looking around the room, shorter answers. Wrap up by asking "Is there anyone else I should talk to?", thank them, and go.

Questions to ask

The one rule: ask about the past, never the future. "Would you use X?" gets polite lies. "When did this last happen?" gets facts.

Core questions (use with anyone)

Open (2 minutes)

• "What does a normal week look like for you right now?"
• "What are your big goals or focus areas right now?"

Find the pain (10 minutes)

• "What's the most annoying part of [your work] right now?"
• "When did that last happen? Walk me through it."
• "What did you do about it?"
• "What did it cost you? Money, hours, or a missed chance?"
• "Have you tried to fix it? What did you try?"
• "Do you pay for anything to help with it now?"

Dig deeper (use any time)

• "Why was that hard?"
• "Can you show me?" (a spreadsheet, a script, a chat, a tool)
• "What happened next?"

Close (3 minutes)

• "Is there anyone else I should talk to about this?" This one question gets you the next 10 conversations.
• "Can I come back to you if I build something?"

Business owners and non-technical people

• "What do you still do by hand, on paper, or in WhatsApp that feels like it should be automatic?"
• "What do you do every week that you hate doing?"
• "Who have you hired, or wanted to hire, just to handle admin?"
• "What's the most expensive mistake you made last year? How did it happen?"
• "Show me your phone. Which apps do you use for the business?"

Engineers and technical people

• "Walk me through the last nasty bug or slow project. Where did the days go?"
• "Have you tried AI coding tools on this kind of work? Where did they fail?"
• "What does your team build internally because nothing good exists to buy?"
• "What scripts or hacks have you written because no tool did the job?"

Managers and people who approve purchases

• "What tool did your team last pay for? Who approved it, and how long did it take?"
• "If a tool fixed this for $50 a month, could you pay with your own card, or would it need approval?"
• "What would make you say no to a tool like that, even if your team wanted it?"

Ask these early. If engineers describe real pain but managers say "we don't buy tools for that, we build it ourselves", you need to know before you build.

Testing an idea (only with people who fit it)

Keep this separate from discovery. Once you mention your idea, people turn polite and stop describing their week. Describe the idea only at the end.

1. Ask how they handle the problem today. "Which tools did you use for this last week?"
2. Ask for the last time it went wrong. "When did that last happen? What happened?"
3. Ask what they've tried. A homemade workaround is a strong signal.
4. Ask if they pay for anything that helps.
5. Only now, one sentence: "I built something for myself that does X." Then stay quiet and watch.
6. Ask for a commitment, not an opinion: "I'm setting up 10 people by hand this month. Want to be one? It takes 20 minutes with me." Time, intros or money count. "Sounds cool" doesn't.

Good and bad signals

After each conversation

Write 4 lines within the hour:

1. The pain
2. What it costs them
3. What they use now
4. Did they commit to anything? (a follow-up, an intro, a trial)

After 10 conversations, look for a pain that shows up 3 or more times within one group.

If 10 conversations give you nothing that people already spend money or hours on, the fix is usually how you ask. Re-read The Mom Test and run 10 more. Going back to desk research won't find it for you.

Resources

• The Mom Test by Rob Fitzpatrick. Short, and the source of most of this guide.
• YC Startup School videos on talking to users.