The Problem Interview Checklist for 2026: Script, Questions and How to Score Pain

A problem interview checks three things before you build anything: whether a problem exists, how often it happens and what people already do about it. You don’t show a product. You listen to stories about the last time something went wrong.

This is the fourth guide in our customer development series. It’s a working checklist: who to talk to, how many, where to find them, a 40-minute script, the questions to ask, the ones to avoid, and how to score what you hear.

Problem interview, solution interview or survey?

	Problem interview	Solution interview	Survey
When	Before you have anything to show	When you have a prototype or MVP	After interviews, to count how common a pattern is
What you show	Nothing	The prototype, a demo or a price	A fixed list of questions
What you learn	Which problems hurt, how often, what people do now	Whether your fix works and whether they’ll commit	How many people share what you heard
Main risk	Pitching by accident	Polite yeses	Questions that only reflect what you already think

Run them in that order. A survey before interviews only measures your own guesses, because you wrote all the answers. Once you have a prototype, move on to solution interviews.

Why this step is worth the time

In CB Insights’ March 2026 study of startups that shut down since 2023, 43% of those with a known reason cited poor product-market fit. Two-thirds of those were early-stage companies that never found a market at all.

There’s a 2026 shortcut that doesn’t work: asking an AI to play your customers. A 2026 study in the NIM Marketing Intelligence Review compared AI-made respondents with real people. The AI answers caught broad trends but were consistently too positive and less varied than real people. On a 7-point scale they were off by 1.2 points on average. A problem interview is meant to surprise you, and answers that are too positive and too alike rarely do.

Who to talk to

• If you don’t know your segment yet, talk to people who are as different as possible: different jobs, situations, incomes and places. You’ll see who feels the problem and who doesn’t.
• If you know your segment, talk to people who look like it. Match them on the things that change the problem, such as company size or how often they do the task.
• In B2B, talk to three roles: the person who does the work, the person who buys and the person who holds the budget. They often describe three different problems.
• Include non-users and unhappy users. People who tried a competitor and quit tell you more than fans.

How many interviews you need

These rules of thumb come from the earlier version of this guide. Treat them as a starting point, not a law:

Situation	Interviews	Time to plan
A clear segment, and you want to confirm what you know	10	1-2 weeks
A subject area you already know well	20	2-3 weeks
A new subject area	40	1-1.5 months if the area is complex

Research programs go further. Teams in the National Science Foundation’s seven-week I-Corps program are told they’ll complete a minimum of 100 interviews. Whatever the number, stop when new interviews stop telling you new things.

Where to find people

1. People you know and the people they know. End every interview with “Who else should I talk to?”
2. Communities and social networks where your segment already talks shop: groups, forums and channels.
3. People with an audience. The head of a trade association or a well-known practitioner can open many doors with one post.
4. Conferences and meetups, where people expect to talk.
5. Your existing users, if you have any, through an email or an in-product message.

Hand off the scheduling to someone else if you can. Never hand off the talking.

Before the interview: three steps

1. Map what you think you know in a quick mind map of the problem, the people and the current fixes.
2. Split people into groups, for example users and non-users, frequent and rare users, or small and large companies.
3. Turn each guess into a question about the past. “Business class travelers pay $1,000 and board first” becomes “What did you pay for your last ticket?” and “Walk me through boarding last time.”

Write the guesses down as hypotheses with a pass and fail line first. Our hypothesis testing guide has a template.

The 40-minute script

Minutes	Part	What you do
0-5	Opening	Thank them, say you’re learning and not selling, ask if you can take notes or record
5-10	Context	Their role, a normal day or week, the tools they use
10-30	Problems	The checklist below; repeat questions 4 to 7 for each problem they raise
30-35	Ranking	Ask them to rate each problem from 1 to 10
35-40	Close	“Who else has this problem?” and “Can I come back to you later?”

Take notes in every interview, even if you also record. Notes make people feel heard. If you record, ask first and say why. Some people talk less freely on tape, so watch for it.

The question checklist

1. How do you deal with this problem now? How much do you pay for it, how much time does it take, and how happy are you with it?
2. What happens if you can’t solve it? Not for every interview, but it helps you rank segments when you talk to many people.
3. What’s the most annoying part of dealing with it? Expect a polite answer the first time. Ask again later.
4. When did it last happen? This tells you how often the problem comes up.
5. Why was it hard to solve?
6. How did you feel when it happened?
7. Rate that feeling from 1 to 10. What would a 10 be?

Repeat questions 4 to 7 for each problem. After 40 minutes you’ll usually have several problems, each with a frequency and a score. That list becomes the first draft of your backlog.

Keep every question open: who, what, when, where, why and how. Ask about facts and past events, not about what people usually do or might do.

Questions that ruin a problem interview

• “Would you buy this?” or “Would you use this?” People say yes to be nice, so a question about the future gives you hope instead of data.
• Leading questions: “Isn’t it frustrating when…?” hands them the answer.
• Anchors: show someone a scary terms-of-service screen and ask how scary it is, and you’ve set the answer. Ask how they feel about the whole screen instead.
• Anything about your idea. In the best interviews, people never learn what you plan to build.

How to score what you heard

Put every problem from every interview into one sheet, one row per problem. Then mark the signals.

Signal	Strong	Weak
Frequency	It happened in the last week	They can’t remember the last time
Current spend	They pay for a tool, a person or overtime	It costs them nothing
Workaround	They built their own fix, like a spreadsheet	They’ve never tried to fix it
Pain score	Above 7 out of 10	5 or below
Who has it	Many people in one group	One person in each group

Focus on problems rated above 7. That’s where people are most likely to pay. Lower-scored problems can still work for mass consumer products, where many people pay a little.

How we would estimate this

Our project estimate can give you your first interview list. It suggests customer groups and gives each one a TAM, SAM and SOM with a note on what the number covers. Treat them as hypotheses to test, not verified demand. Pick the group you’d sell to first and book ten problem interviews with people in it.

The estimate also turns your brief into user stories. You confirm what’s in scope, and every hour in the budget traces back to a story you kept. We check the hours the model suggests against what our team actually books on finished projects, then multiply by our published rate of $39-55 an hour. When your interviews kill a problem, uncheck its stories and see what the build would cost without them.

What we would do

1. Write your top five guesses about the problem as hypotheses.
2. Pick one group and book ten interviews over the next two weeks.
3. Run the 40-minute script. Don’t mention your idea.
4. Score every problem in one sheet the same day, while you remember the details.
5. Keep the problems that score above 7 and come up often. Drop the rest, or test them in a new group.
6. Turn the top problems into the scope of a first version. If you need a fixed price and a technical plan for that scope, a discovery sprint takes a week. Our software development for startups page shows the price of each stage after that.

If you’re short on time or money, the small-budget guide shows which tests to run before and after interviews.

FAQ

What is a problem interview in customer development?

A conversation with a potential customer before you have a product. You ask about the last time a problem happened, what they did about it and what it cost them, to learn whether the problem is real and worth solving.

How many problem interviews should you do?

Ten to confirm a clear segment, about 20 in an area you know and about 40 in a new one. Stop when new interviews stop telling you anything new.

What questions should you ask in a problem interview?

How they deal with the problem now, what it costs, when it last happened, why it was hard, how they felt and how they’d rate it from 1 to 10. Repeat the last four for every problem they raise.

How long should a problem interview take?

About 40 minutes: 5 to open, 5 on context, 20 on problems, 5 on ranking and 5 to close. Get the most important questions into the first 15 minutes in case they have to leave.

Can AI replace problem interviews?

No. A 2026 study found AI-made respondents were too positive and less varied than real people. Use AI to prepare questions and sum up notes, not as the customer.