How to prepare for a product sense (or product thinking) interview

Are you reading sample answers and hoping the prompt turns out to be one you've seen? Then the interviewer says "design a product for gardeners," and because silence feels dangerous, you start listing ideas within thirty seconds. That's the most common way to fail this interview, and it has nothing to do with how good your ideas are.

This article covers what interviewers actually score, and the three shapes the questions come in. It also covers a step-by-step structure for answering, a worked example you can copy the mechanics from, and a practice plan. Read the structure section twice, then spend your remaining prep time saying answers out loud, because reading quietly doesn't prepare you to talk for 35 minutes.

The advice comes from Ben Erez, who interviewed more than 50 PM candidates at Meta and now coaches product interviews full time; Julie Zhuo, who spent 14 years at Facebook and became its VP of product design; Shreyas Doshi, who led product teams at Stripe, Twitter, and Google; Scott Belsky, Adobe's chief product officer at the time; Jiaona Zhang, then head of product at Webflow; and hiring leaders from Netflix, Rippling, and EvenUp. Every quote plays the exact moment it was said.

What interviewers actually score

Product sense interviews, sometimes called product thinking or product design interviews, were popularized by Meta and Google and have spread across the industry, including Stripe, OpenAI, Block, and plenty of earlier-stage companies. The setup is consistent: a 45-minute slot, roughly 35 usable minutes after introductions, and one open-ended prompt.

Ben Erez, who sat on the interviewer side of this exact loop at Meta, describes a rubric with five dimensions. You need a passing score on each one, and a brilliant performance on one can't rescue a miss on another:

• Clear communication: an easy-to-follow answer with visible structure.
• Product motivation: you examine why the product exists and what it's trying to do before building anything.
• Segmentation: you define the audience, split it into meaningfully different groups, and pick one on purpose.
• Problem identification: you list real pain points for that group and prioritize one.
• Solution development: you brainstorm more than one approach, pick with reasoning, and scope a concrete V1.

One more thing worth knowing about the person across the table: they're usually a working PM squeezing your interview between meetings, filling out that rubric from memory afterward. Your job is to make their evidence easy to collect. Announce what you're about to do, do it, then check in before moving on. Erez calls this waypointing, and it's the easiest part of the rubric to score well on.

Note that this round tests your thinking on a made-up product. Questions about your past work belong to the behavioral round, and the behavioral interview questions playbook covers those separately.

The three question shapes

Almost every prompt is one of three shapes. The structure for answering barely changes, but each shape has its own trap.

"Design a product for X"

Examples: design a product for gardeners, build a podcast product for Netflix, design something for people learning a language. The trap is treating the prompt as an invitation to pitch the first product you can picture. The whole point of the shape is watching you narrow an enormous space: which gardeners, trying to do what, stuck where. Spend your first minutes shrinking the problem, and don't propose anything until you've picked a user and a pain point.

"How would you improve Y"

Examples: improve Google Maps, improve a product you love, improve our product. The trap here is feature listing. Candidates who open with "I'd add dark mode and better search" have skipped the question of what the product is failing to do for whom. Shreyas Doshi's model of product work explains the difference interviewers hear instantly.

A feature list is execution-level thinking. A strong improvement answer starts at the impact level: what outcome does this product exist to create, where is it falling short, and for which users. Only then do you propose changes, each one tied back to that outcome.

The ambiguous prompt

Some interviewers deliberately hand you something underspecified: "our engagement is flat, what would you do?" or "should we enter this market?" The vagueness is the test. Jiaona Zhang, who has hired PMs at Airbnb, WeWork, and Webflow, describes what she listens for when candidates hit ambiguity.

She also looks for candidates who ask for inputs along the way instead of declaring one true path, because the odds of charting the right course on the first try are low in real product work too. Stating assumptions, proposing a plan, and inviting correction is exactly that behavior in miniature.

A structure that covers every signal

Erez coaches candidates to walk the five dimensions in a linear flow, and to say the plan out loud before starting. Here's the arc with rough timings for a 35-minute exercise:

The step candidates most want to skip is problem identification, because ideas are more fun than pain points. Scott Belsky, who built Behance and ran product for Adobe's creative tools, watched that exact mistake sink real teams for years.

In interview terms: if you describe your user's day in enough detail, the solution section gets much easier, and the interviewer gets to watch you work your way to an idea rather than defend one you brought with you.

A worked example: "How would you improve Spotify?"

Here's the structure applied, compressed. In the real interview each step gets the full timing above.

Assumptions and game plan. "I'll assume I'm a PM on consumer Spotify, focused on the mobile music experience rather than podcasts, in a mature market like the US. I'd like to spend a couple of minutes on what Spotify is trying to do, then segment listeners and pick one group, walk their listening day to find problems, and bring a couple of solutions down to a V1. Does that work for you?" Thirty seconds, and you've already banked most of the communication score.

Product motivation. Spotify makes money when people keep listening and keep subscribing, and people stay when the app reliably gives them the right audio for the moment they're in. Mission sentence: "help people get the right audio for every moment of their day."

Segmentation. The ecosystem has listeners, artists, and advertisers; you're focusing on listeners. Behavioral segments might be: curators who build playlists, lean-back listeners who press play on whatever's offered, and social listeners who treat music as something to share. Discovery for lean-back listeners is already well served by the algorithmic playlists, so pick social listeners: large reach, and the app treats sharing as an afterthought.

Problem identification. Walk a social listener's day: they hear a song that reminds them of a friend, send it, and then nothing happens, because the conversation is now in a chat app instead. They can't see what friends are into now, and reacting to a shared song means leaving Spotify entirely. Prioritize the reaction gap: people share often, and they rarely hear anything back.

Solution development. Three different approaches: reactions and short notes on shared songs inside the app, live listening rooms with friends, or a weekly blend recap you and a friend both see. Listening rooms are high effort and demand synchronous friends; the recap is low effort but slow feedback. Pick in-app reactions: high frequency, modest build. V1 is a react-and-reply layer on shared tracks, shown to the recipient when they play the song. Risk: it needs a critical mass of friends on Spotify, so seed it inside the existing Blend feature first. And that closes the loop: more moments where the right audio comes from someone who knows you.

None of these choices is the one right answer. A different segment and a different problem pass just as well, as long as every choice comes with a spoken reason.

What hiring managers screen for now

The rubric above hasn't changed much in years, but the bar around it has moved. In a panel conversation on The Skip, hiring leaders from Netflix, Rippling, and EvenUp agreed that most candidates are running a playbook that's about two years out of date. AI has made polished, well-structured documents trivial to produce, so the screens have shifted toward what a tool can't fake: depth of thought, the ability to make a complicated problem simple, and the judgment to question the prompt itself. The Netflix leader on the panel calls out one red flag in case exercises specifically: candidates who jump straight into answering without ever asking whether it's the right problem to solve.

That instinct, pausing to ask "is this the right problem?", is precisely what the assumptions step and the problem prioritization step give you room to show. The structure does more than help you remember the steps, because working through it forces you into the behavior senior interviewers are looking for.

How to practice in the weeks before

Erez is blunt about this: no one wings these interviews successfully. His own prep for the Meta loop took four intense weeks. The good news is that the practice compounds, because you're building the same muscle the job uses. Julie Zhuo's advice on developing product sense in the first place doubles as interview prep.

Zhuo's full method: observe your own reactions to every product you touch, then discuss products with other people and dissect the builders' decisions, then check your hypotheses against data wherever you can. Every one of those conversations is a low-stakes rehearsal of the interview.

For the interview-specific work, over two to three weeks:

• Build an opening template. Write three or four placeholder assumptions you can adapt to any prompt, so the first two minutes never start from a blank page.
• Rehearse 20 to 30 openings. Erez has candidates practice just the assumptions-and-game-plan opening across dozens of prompts, because a strong start settles your nerves for everything after it.
• Drill problems, separately from solutions. Take five products and phrase pain points as "users struggle with Y because Z." If your sentence names a feature, you've written a solution and called it a problem, so rewrite it.
• Map one persona's day per week. Sketch their day hour by hour and mark every touchpoint where a product could help. This builds the journey-mapping step until it's fast.
• Do at least three full mocks out loud, timed. With a friend, a coach, or a tool, and record yourself. The clock is a big part of what makes this hard, and you only find that out by running the whole thing under time.

If your product sense round sits inside an onsite with several other interviews, the final-round playbook covers how to prepare for the rest of the panel around it.

Frequently asked questions

Should I use a named framework like CIRCLES?

Use a structure, but own it. Interviewers at big tech companies have heard hundreds of candidates recite the same acronym, and naming it earns you nothing on its own. The five-part flow above hits every rubric dimension, so practice it until you can bend it to the prompt instead of marching through letters.

What if I know nothing about the domain in the prompt?

That's expected, and it's partly why the domain is chosen. You're being scored on structure, reasoning, and user empathy, never on gardening expertise. State an assumption ("I'll assume home gardeners in apartments, since that's who I can reason about"), and reason carefully from what you do know. Asking the interviewer a clarifying question or two is fine, but interrogating them for domain facts doesn't go over well.

What if the interviewer interrupts or challenges my choice?

Treat it as steering, and usually as help. "Are there other problems you'd consider?" usually means they want more breadth before you narrow. Pushback on your prioritization is an invitation to show your reasoning, and sometimes a test of whether you get defensive under pressure. Engage with curiosity, adjust if they've shown you something real, and keep your structure.

How is this different from a product strategy or analytical interview?

Product sense tests user-centered design thinking on an invented problem. Analytical or execution rounds test metrics, goal-setting, and diagnosis, and strategy rounds test market-level reasoning. Companies mix them differently, so ask your recruiter which rounds you're getting. Recruiters answer that question happily, and it costs you nothing to ask.

How long should I prepare?

If the interview is days away, focus on the opening template and two timed mocks. If you have a few weeks, run the full practice plan above. Erez spent four weeks preparing for his Meta loop while already working as a PM, and that's a reasonable benchmark for a competitive company.

The product sense brief on one page