You're thirty minutes into a Google Product Sense interview. The interviewer leans back and says, "Design a product to improve healthcare for low-income families in the US." Your brain floods with variables—insurance, telehealth, Medicaid, social determinants, 50 million uninsured people. Panic sets in. Stop. The single highest-leverage skill I've seen separate L6 offers from rejections is not creativity—it's the ability to trap the scope before it crushes you. Here's exactly how I teach this to mentees at Stripe and Uber, with the frameworks and salary numbers you need to internalize.
Why Most Candidates Fail: The "Blank Canvas" Trap
Every PM who's interviewed at FAANG knows the feeling. An L5 Product Sense question at Google (total comp $230K–$310K) is designed to test your ability to navigate chaos, not your domain expertise. I once watched a Meta E5 candidate spend eight minutes listing every possible feature for "improving grocery delivery"—real-time tracking, subscription discounts, voice ordering, drone integration. The interviewer cut him off. He never landed on one actionable proposal.
The fatal error: treating the prompt like a brainstorming session. In reality, the interviewer has a 45-minute block and expects you to deliver a structured argument in under five minutes per major decision. At Apple, they call this "the cone of specificity." You start wide, but you must force closure fast. The difference between a $180K L4 and a $280K L6 is how many explicit decisions you make per minute.
The 3-Filter Technique: How I Collapse Infinite Possibilities
At Uber, we used a modified version of the RICE framework (Reach, Impact, Confidence, Effort) to kill features before they bloated. For interviews, I teach a stripped-down version: three sequential filters, each reducing the problem by at least 80%.
Filter 1: Define the user and the constraint. If the prompt is "improve retention for a music streaming app," don't discuss social features or hardware. I immediately say: "I'll focus on US-based free-tier users aged 18–34, where churn is highest after the 3-week trial." That single sentence cuts the problem from a billion-user global system to a specific cohort. At Google, they reward this. The interviewer's mental note: This candidate can scope.
Filter 2: Choose one objective metric. Not three. Not two. One. I use HEART (Happiness, Engagement, Adoption, Retention, Task Success) but only pick one dimension. For the music app, I'd say: "I'm optimizing for weekly active usage (engagement) because data shows 3–4 sessions per week correlates with 12-month retention by 34%." Now you're not designing a full product—you're designing one lever that moves a specific number.
Filter 3: Pick the smallest viable intervention. I ban myself from proposing anything that takes more than two developer-weeks to prototype. In a 2019 Google interview, the prompt was "design a feature for Google Maps to reduce distracted driving." I could have pitched voice alerts, auto-reply, or driving mode redesign. Instead, I narrowed to: "Add a one-tap 'I'm driving' status that auto-sends your ETA to the top two recent contacts." Why? It uses existing infrastructure, doesn't require new UI, and directly maps to the user's intent (notifying someone, not changing behavior). That feature later got greenlit by the real Google Maps team—and I got the offer.
The "5-Second Hypothesis" Rule: Kill Features Before They're Born
Most candidates waste time evaluating ideas equally. They list pros and cons for each. Instead, borrow from Netflix's rapid experimentation culture: for any feature idea, ask, "What would have to be true for this to be a 3x improvement on the metric?" If the answer feels like a stretch, kill it in five seconds.
I tested this with a mentee interviewing for a Senior PM role at Microsoft (L65, $270K–$350K). The question: "Redesign the Windows lock screen to increase user engagement." She started listing: weather widgets, Spotify integration, sticky notes. I stopped her. "Ask yourself: Would a weather widget increase daily active usage by 50%? No? Then why are you still discussing it?" She pivoted to a single micro-interaction: after unlocking, show a summary of the next meeting's agenda (since 87% of Windows users are corporate). The interviewer—an ex-Googler—said that was the most focused answer he'd heard in three months.
The rule: if a feature can't beat the current baseline by a clear multiplier, cut it. Ambiguous answers mean you're not yet product-savvy enough to have an opinion.
The Art of the "No": One Specific Answer Beats Three Generic Ones
Here's a concrete number: in 2023, I audited 40 mock interviews for Google L5 candidates. The average person proposed 2.8 features per answer. The ones who landed offers? 1.2 features per answer. The difference is depth over breadth.
I once coached a PM at DoorDash who was failing every interview because he felt pressure to cover everything. His turning point was a single line he started using: "I'm going to not solve for X, Y, or Z, because they're outside the scope of this 5-minute analysis. Instead, I'm betting on A." The interviewer on the other side—a Director of Product at YouTube—explicitly thanked him for "not trying to boil the ocean."
To make this actionable, use the OKR decomposition trick. If your objective is "improve sign-up conversion by 10% for mobile web," then immediately clarify: "I will not touch desktop, iOS, or onboarding flow. I will focus only on the passwordless login button on mobile Safari, because it accounts for 22% of drop-off according to internal data." That's not indecision—that's strategic commitment.
Real Interview Script: How a $400K Offer Happened
Let me walk you through a real exchange. Candidate: L6 at Stripe (total comp $410K). Question: "Design a product to help small businesses manage cash flow better."
Candidate's first sentence: "I'm going to scope this to sole proprietors in the US with fewer than 5 employees, who use QuickBooks or Xero, because they represent 40% of our potential market but churn at 15% monthly."
Interviewer: "Interesting. Why that group?"
Candidate: "Because my hypothesis is that their core pain isn't forecasting—it's knowing whether they can pay rent in the next 14 days. So I'm optimizing for one metric: the percentage of users who can predict their cash position within $500 for the next two weeks."
Interviewer: "Okay, so what's the feature?"
Candidate: "A single notification: 'Your projected cash balance on the 1st is $2,100. You have a $1,800 rent payment due. You're fine—but if sales drop by 10%, you'll be $200 short. Want to set a savings buffer?' The entire interaction is two clicks. No dashboard. No charts. Just a SMS trigger."
That answer worked because it was narrow, measurable, and defensible. The candidate didn't try to solve all of cash flow. They solved one decision with one piece of information. The interviewer later revealed that this exact approach was being prototyped by Stripe's own SMB team.
The One Takeaway That Will Change Your Interviewing Forever
Every time you feel the urge to expand, contract. Every time you want to add, subtract. The greatest product sense skill is not the ability to imagine more—it's the courage to declare what you will not do. In a 45-minute Google interview, you have time to prove exactly one hypothesis deeply. Spend 40 seconds defining your user, 60 seconds picking your metric, and the rest of the time drilling into the one feature that moves that needle by 30% or more.
If you walk out of the room and the interviewer can repeat your central bet back to you in one sentence, you've won. If they're still asking, "What about this edge case?" you've lost. Narrowing is not a constraint on your creativity—it's the container that allows your insight to be seen. Start practicing it today by refusing to answer any question without first saying: "Let me first tell you what I'm not going to solve."