Requirements interview questions
Requirements questions a Product manager can be asked, from first job to senior. Open one to see the strongest answer and why it works.
What is a PRD for?
The strongest answer: The decision, the non-goals, and what done looks like — not a novel that restates Slack
Why it works: Exactly. Decisions and non-goals. Novels, titles, and invented lifts are not a record.
- Decision, non-goals, done.
- Not a Slack reprint.
- Not a 40-page costume.
- No invented lift in the PRD.
What makes an acceptance criterion useful?
The strongest answer: A check someone can run without you in the room — not “it feels right” or “as discussed”
Why it works: Exactly. Observable. Slack, vibes, and invented lifts are not a check.
- Writable, runnable checks.
- Not “as discussed.”
- Not vibes.
- No lift % as a criterion.
What is the difference between a functional and a non-functional requirement, and why do specs keep missing the second kind?
The strongest answer: Functional is what the system does; non-functional is how well — speed, availability, security, accessibility, languages, data retention. They get missed because nobody demos them
Why it works: Exactly — the “how well” bar is a product decision with a price, and writing it down is what stops it being discovered in production.
- What it does vs how well.
- Name speed, uptime, security, access.
- Set the bar; do not discover it.
- It is a product decision.
A story has a what and no why. What do you add?
The strongest answer: Who is trying to do what job, and how we will know it worked — I do not add another wireframe instead
Why it works: Exactly. Job and signal. Mockups and fake “users expect” are how you skip the why.
- Who, job, how we will know.
- Do not substitute mockups.
- Titles ≠ why.
- No invented “users expect.”
A ticket says “redesign settings, add SSO, and fix billing.” What do you do?
The strongest answer: Split it until each piece has one job and a check — a suitcase ticket is how nothing finishes
Why it works: Exactly. One job, one check. Suitcases and silent epics are how weeks vanish.
- One job per story.
- Split the suitcase.
- Do not add more to look complete.
- Epic ≠ done thinking.
What belongs in a spec besides the happy path?
The strongest answer: Empty, error, permission, and what we will not do — I do not leave those to “engineering judgement” in week two
Why it works: Exactly. Empties and non-goals. Pretty-only and promised NPS skip the week-two surprises.
- Empty, error, permission, non-goals.
- Do not leave them to week two.
- Happy path ≠ the spec.
- No promised NPS.
What does a good spec do for engineering?
The strongest answer: Removes the decisions they should not have to guess — states, empty, permissions — I do not send a moodboard and vanish
Why it works: Exactly. The decisions. Pixel prisons, vibes, and invented percentiles do not unblock.
- Spec the decisions: states, empty, permissions.
- Do not vanish behind a moodboard.
- Do not pixel-prison.
- No invented performance percentile.
Your PRD is 28 pages. What do you cut first?
The strongest answer: Repeated context and solution essays — keep the decision, non-goals, and checks; I do not keep pages to look thorough
Why it works: Exactly. Decision and checks. Cutting non-goals and checks is how 28 pages still fail.
- Keep decision, non-goals, checks.
- Cut repeated essays.
- Do not cut non-goals to look positive.
- Length ≠ quality.
A permissions change is spread across twenty-two stories. In refinement nobody can say what an admin experiences end to end, and QA keeps finding gaps between the stories.
What do you write?
The strongest answer: A use case for the flow: the actor, the precondition, the main path step by step, and the alternate and failure paths — and keep the stories as the slices we deliver against it
Why it works: Right — stories slice delivery, a use case describes the whole interaction including what happens when it goes wrong. This needs both.
- Stories slice; use cases connect.
- Name the actor and precondition.
- Write the main path in steps.
- Write the alternates and failures.
Design, eng, and legal still disagree after the PRD. What failed?
The strongest answer: The decisions were not on one page they all marked — I do not add a 29th page of context
Why it works: Exactly. One marked page. Covers, side chats, and production-as-referee are how you spend a quarter.
- One page of marked decisions.
- Do not add more context pages.
- Do not side-chat a faction.
- Do not let production referee.
What does a staff bar for PRDs look like?
The strongest answer: Findable decisions and living non-goals — I do not grade page count or prose flourish
Why it works: Exactly. Findable decisions. Page counts, covers, and required lifts are costumes.
- Findable decisions, living non-goals.
- Do not grade page count.
- Covers ≠ quality.
- No required lift %.