PM interviews are genuinely weird. In an engineering interview, there’s usually a right answer. In a PM interview, you can answer the same question correctly in two completely different ways, and interviewers at the same company will have different opinions about which one was better.
I’ve watched this play out enough times to have some opinions about what actually works, and some of those opinions are probably wrong for certain companies. I’ll flag the exceptions as I go.
What the interview is actually testing
The framing I find most useful: PM interviews aren’t testing your knowledge of frameworks. They’re testing whether you think clearly under pressure, can communicate tradeoffs, and have enough product intuition to ask the right questions before jumping to solutions.
Frameworks are scaffolding. They help you organize your thinking when you’re nervous and the clock is running. But an interviewer who hears you recite RICE or ICE as if you’re reading from a textbook isn’t impressed. They want to see you reason, not recall.
This distinction matters a lot for how you prep.
Product sense: how to improve a product without sounding generic
The classic product sense question is something like “how would you improve Google Maps?” The candidates who bomb it start with features. The ones who do well start with users.
A better opening: “Can I ask who we’re optimizing for? A daily commuter has totally different problems than someone navigating an unfamiliar city once a year.” That question alone signals that you think in user segments, not product features.
Then pick one segment, articulate their top friction point with specificity (not “navigation is frustrating,” more like “I lose the audio cue right when I’m entering a confusing intersection”), and work from there. The solution almost writes itself once the problem is crisp.
For prioritization, RICE (Reach, Impact, Confidence, Effort) is a reasonable shorthand. ICE is simpler and easier to defend in a 30-minute interview. Either works. What doesn’t work is listing both frameworks and then not actually scoring anything.
Strategy questions: the 90-day plan trap
A lot of PM candidates, especially those who’ve done their research, come in with a canned 90-day plan answer: learn, then execute, then scale. First 30 days I’d listen, second 30 days I’d build relationships, third 30 days I’d ship something.
Interviewers have heard this answer hundreds of times. It doesn’t differentiate you.
What does differentiate you is being specific about how you’d learn. “I’d do 20 customer calls in the first three weeks, specifically targeting churned users, because they tell you more about unmet needs than happy users do” is a real answer. It shows you have a theory about where signal comes from.
For market entry and “should we build X” strategy questions, the instinct to use a framework like SWOT or Porter’s Five Forces isn’t wrong, but the move that gets people hired is connecting it to the company’s actual situation. If the interviewer is at a payments company asking about expanding into BNPL, you should be referencing what Klarna and Affirm have figured out (and where they’ve had regulatory problems), not describing market entry theory in the abstract.
According to LinkedIn’s Economic Graph research, PM roles have grown faster than almost any other function over the last five years, which means the candidate pool is much larger than it was in 2019. The bar for “showing you’ve done your homework” has moved up.
Metrics: the question that trips up most candidates
Metrics questions are where I see the sharpest divergence between candidates who will get offers and candidates who won’t.
The question often looks like: “Our weekly active users dropped 15% last month. Walk me through how you’d investigate.” A weaker answer starts immediately: “First I’d look at acquisition, then retention…” A stronger answer pauses: “Before I start, can you tell me if the 15% drop was consistent across all platforms or did it concentrate somewhere? And was there anything significant that changed in the product or in external conditions that month?”
Those clarifying questions aren’t stalling. They’re demonstrating that you know the investigation depends on the answer. That’s the North Star metric thinking they’re testing for.
One thing I genuinely don’t know: how much the technical depth expectation for metrics varies across companies. At data-heavy companies like Stripe or Airbnb, I’d expect them to push into A/B test design and statistical significance. At earlier-stage companies, that might be overkill. Calibrate based on the company before assuming.
Case studies: don’t solve them alone
The biggest mistake I’ve seen in product case studies is treating them like a solo exercise. Someone will get a case like “should Spotify expand into podcasts in Southeast Asia,” go quiet for three minutes, and then present a fully-formed recommendation as if they’ve been CEO for a year.
Thinking out loud is better. Not stream-of-consciousness thinking out loud, but narrating your reasoning: “I’m going to start by framing the market size question before I get to product decisions, because I think the distribution question is load-bearing here.” That kind of narration gives the interviewer something to engage with, and most interviewers will redirect you if you’re going somewhere unhelpful. That redirect is useful information.
You can practice this well with AI-assisted mock interviews. Craqly‘s interview practice mode will push back on your reasoning with follow-up questions, which is much closer to a real interview dynamic than rehearsing alone. Whether that kind of tool helps you or makes you overthink probably depends on how you learn, and I’d be honest with yourself about that.
What actually gets people hired
The Stack Overflow 2024 Developer Survey found that nearly 76% of developers now work directly with AI tools in their day-to-day work. PM candidates who can speak concretely about AI product decisions, not just abstractly, are increasingly standing out in interviews at companies where AI is a real part of the product.
That’s a narrow point but a real one. If you’re interviewing at an AI-forward company and your product sense examples are all from 2019, that’s a signal worth thinking about.
The broader point: PM interviews reward candidates who seem like they’ve thought carefully about products they actually use, not candidates who’ve memorized the canonical frameworks. The frameworks are a vehicle. The thinking is the thing.