How to answer the product discovery interview question

Written by Reed, an AI career analyst at PM Career Desk.Reed, an AI, drafted this guide from the Interview Coach question bank and its answer check. The question is shown word for word as Interview Coach asks it. It is general guidance and promises no outcome. The example answers and figures are invented.

The short answer

Interview Coach asks this question in these words:

How did you find out what problem to solve on your last product? What surprised you?

Pick one product decision that came from research you did yourself. Say who you talked to or which data you used, and how much of it. Then name the one finding that surprised you and changed the plan. End with how that finding reached the roadmap and what happened next. The surprise is the heart of the answer: research that only confirmed the plan is weak evidence that you listen.

Check your own answer in Claude with Interview Coach, free

What the interviewer is checking

The question tests whether your roadmap comes from customers and data or from the loudest voice in the room. The second half, "what surprised you", is where most answers go flat. The bank lists three things a strong answer does:

Names the research: who you talked to or which data you used. Twelve interviews with clinic managers, six months of support tickets, a funnel report. Say the size and the source.
Describes the finding that changed your plan. One finding, said plainly. "Users wanted a better experience" is not a finding; "managers re-entered every booking by hand on Mondays" is.
Shows how the finding reached the roadmap. A roadmap item moved, a feature dropped, a new problem statement. Say who agreed and when.

Name the people you learned from by role, never by name or company if the work was confidential.

A structure that works

Interview Coach describes a story answer in five parts:

Situation. Two or three sentences on the context and what was at stake. Here: the product, the open question and why it mattered.
Action. What you did yourself, said with I, in the order you did it. Here: the research you ran and how you read it.
Result. What changed, and how you know. Here: the finding and the change it caused.
Evidence. A figure before and after, or a record a reader could ask for. Here: the research notes, the survey, the roadmap or the analytics dashboard.
Length. About one to two minutes when spoken. Spoken at a normal pace, one to two minutes; the check marks an answer under 60 words as short.

The answer check reads the answer for a situation set before the action, at least two actions said with I, a result near the end, and a before and after pair of figures or a named record. It also reads the balance between "we" and "I", the length of a spoken answer, and words that weaken an answer: filler, hedges, weak ownership and vague claims.

Answers as written, and what the check says

Illustrative, not a real client. Each answer below was sent on its own through the answer check of Interview Coach. All of them are also marked short for a spoken answer, because each is one or two sentences; the table shows the other finding, word for word, and one way to fix it.

Answer as written	What the check says	What to change
When weekly use of the booking product was flat, clinic managers were interviewed and the support tickets were read for a quarter, and bulk editing was approved ahead of the reports page, per the research report.	No sentence says what you did yourself, with I.	Say which of these steps were yours, with I.
When weekly use of the booking product was flat, I interviewed clinic managers, and bulk editing was approved ahead of the reports page, per the research report.	Only one first person action found. Say what you did yourself.	Add the second step that was yours, in the order you took it.
When weekly use of the booking product was flat, we interviewed managers, we read tickets, we mapped journeys, we ran a survey and we rewrote the brief. I interviewed clinic managers and I read the support tickets for a quarter, and bulk editing was approved ahead of the reports page, per the research report.	Mostly we and the team. The interviewer cannot tell which part was yours.	Give the team's work one sentence, then say which calls were yours.
I interviewed clinic managers and I read the support tickets for a quarter, and bulk editing was approved ahead of the reports page, per the research report.	No context is set before the action. Add what was at stake.	Open with what was at stake before you acted.
When weekly use of the booking product was flat, I interviewed clinic managers and I read the support tickets for a quarter, per the research report.	No result is named. Say what changed.	Say what changed in the end, and where it shows.
When weekly use of the booking product was flat, I interviewed clinic managers and I read the support tickets for a quarter, and bulk editing was approved ahead of the reports page.	No figure and no named record. Name where the result shows, if you have it.	Name where the outcome is recorded, or give the figure before and after.
When weekly use of the booking product was flat, I interviewed clinic managers and I read the support tickets for a quarter, and weekly use rose by 18 percent after the bulk edit release.	A figure is given, with no starting point.	Give the starting figure too, or name the record that holds both.
When weekly use of the booking product was flat, I kind of interviewed clinic managers and I read the support tickets for a quarter, and bulk editing was approved ahead of the reports page, per the research report.	Filler words found.	Drop the filler word and say the step plainly.

The check never adds a figure or a record for you. If the research changed nothing, choose a different story.

An illustrative answer

Illustrative, not a real client. Every figure is invented. This answer gets all nine marks of the story check: length, situation, own action, team versus me, result, evidence, and no vague words, filler or hedges.

I was the product manager for the booking tool of a chain of physiotherapy clinics. When I joined, weekly use by clinic managers had been flat for two quarters, and the roadmap had a new reports page as the next big item. Before we started the build, I interviewed twelve clinic managers and I read three months of support tickets. The surprise was that managers did not need reports: every Monday they re-entered the week's bookings by hand, because the tool could not edit a series of appointments at once. I wrote the finding up in a two-page research report with quotes and ticket counts, and I took it to the head of product with a proposal to replace the reports page with bulk editing. She agreed, and the reports page moved to the next half. The bulk edit release went out in six weeks. Weekly use by managers rose from 41 to 63 percent of active clinics in the quarter after release, per the product analytics dashboard.

Why it works. The research has a size and a source. The finding is concrete and contradicts the plan. The person who decided is named, the roadmap change is visible, and the effect has a before and after figure from a named dashboard.

What the check cannot see

The answer check reads the words of your answer, not the research behind them. This answer, sent on its own, gets every mark except length:

When use was flat, I interviewed managers and I read tickets, and bulk editing was approved, per the research report.

It has a situation, two actions, a result and a record, so the check passes it on structure. An interviewer would ask what the surprise was, which roadmap item moved, and what happened after. Without the finding itself, it is a description of a research process, not of a discovery.

Choosing your story

Pick research you ran yourself. Reading another team's study is fine as a source, but the interviews, the data pull or the test should be yours.
Pick a finding that changed a plan. The question asks what surprised you; research that confirmed the plan shows less.
Pick one that reached the roadmap. A finding that stayed in a slide deck tells the interviewer you could not land it.
Size the research honestly. Five interviews is a real number. Do not round it up; interviewers ask follow-ups.

Your resume is where an interviewer looks for these stories, and a line such as "led customer discovery" invites this question. Get a free AI review of your resume and see which lines Reed flags first.

Prepare the follow-up questions

How did you decide how many customers to talk to?
Who disagreed with the finding, and how did you handle it?
What would you research differently next time?

Run the free review on your own resume lines.

Check your answer with Interview Coach in Claude

This question is in the product manager track of the Interview Coach question bank. Write your answer the way you would say it and ask for a check.

It reads the answer for a situation, your own action said with I, a result, and a before and after figure or a named record.
It flags filler, hedging, weak ownership and vague claims, and says when the answer is short or long for a spoken answer.
It does not judge whether the story is true or good, and it never adds a figure you did not give.

Your answers start from your resume. The free review quotes your own lines and flags the 5 to fix first, so you know which lines an interviewer may question.

Check my answer in Interview Coach, free Review my resume, free

Free in Claude, no account. We do not store the text you send through it.

Questions

What if I had no budget for research?

Most discovery is cheap: support tickets, sales call notes, five calls with customers, a look at usage data. Say what you used and why it was enough. The interviewer is listening for the habit of checking, not for the size of the study.

Is this question only for product managers?

It is in the product manager track of the Interview Coach question bank, and product owner titles meet it too. Business analysts often face a close cousin about requirements; the same structure works.

Does Interview Coach know whether my story is true?

No. Interview Coach runs in Claude and checks the structure of an answer by fixed rules: a situation, your own action, a result, a figure or a record. It does not judge whether the story is true or good. You decide what to say.

Keep going

Interview questions: Product manager interview questions
Interview questions: Product owner interview questions
Guide: How to answer the stakeholder conflict question
Guide: How to answer the influence without authority question
Guide: How to answer the hardest deadline question
Guide: How to answer the product metric question
Guide: How to answer the feature you did not build question
Guide: How to answer the failed project question
Guide: How to answer the scope change question
Guide: How to answer the project risk question

Last updated: October 14, 2026