Meta Product Sense Interview (2026 Guide)

Meta's product sense round asks you to design a product from scratch, defend your decisions under follow-up pressure, and (increasingly) prototype your solution live using AI tools.

Read more about the product sense interview.

Meta's Product Sense Interview

The product sense round sits inside a broader Meta PM interview loop, but the loop structure changes depending on the level you're interviewing for.

Read more firsthand interview experiences from Meta candidates.

For IC roles (L5/P4–P5), the typical sequence is two phone screens (one product sense, one product execution/analytical), followed by an on-site with another product sense, another analytical, and a behavioral/leadership round.

One candidate described it as "two rounds, each with product sense and analytical thinking, then behavioral in the second round."

Some candidates report a gate between phases, where Meta reviews your phone screen performance before advancing you to the on-site.

For manager-level roles (L7 equivalent), the loop expands significantly. A Google director who interviewed for Meta's PM leadership track described six on-site rounds: people management, cross-functional partnership and influence, execution, operating/project retro, product sense, and metrics.

He went through multiple recruiter screens and a PM aptitude screen before even entering the formal loop, a process that took months.

Each round is typically 45 to 60 minutes. The interviewer is a Meta PM, though their team and background may have nothing to do with the question they're asking you.

Meta introduced an AI product sense round in late 2025.

In this round, you spend the first 30 minutes on a traditional product sense case (clarifying questions, user segmentation, solution design), then switch to a Llama-based prototyping tool for the remaining 30 minutes to vibe code your solution. More on this below.

Meta Product Sense Interview Questions

Read more Meta product manager interview questions.

Phone screen questions:

• "Design a parking solution." No context, no constraints. The candidate had to scope it entirely through clarifying questions.
• "You're a product manager at Meta tasked with making a tool to connect handymen with consumers. How would you design the product?" The interviewer specified domestic only, web and mobile, and that the product needed an ROI.

On-site product sense questions:

• "Design a product for volunteering." This prompt appeared across multiple candidates, confirming it's in active rotation. One version was: "You're a PM at Meta. You've been put in charge of a brand new product for volunteering. What would you do and why?"
• "Design a parking solution for Google Maps." Similar to the phone screen prompt but anchored to a specific platform.
• "Design a solution for contractors." Intentionally broad. When the candidate asked whether it should live on an existing Meta platform, the interviewer said yes. The candidate built the solution on Facebook.
• "Build me a Facebook events product." At the L7 level. When the candidate asked for clarification, the interviewer narrowed it to "going to the movies together."

AI product sense round:

• The same "volunteering" prompt, but with the second half spent prototyping in Meta's Llama-based tool.
• A second candidate's friend received: "If Meta wanted to create a map app, how would you go about creating it?" Same 30/30 format.

Analytical questions (often bundled with product sense):

• "Determine a north star metric for Instagram Reels" with a trade-off: prioritize Reels over Stories and defend your reasoning. The interviewer also asked about counter/guardrail metrics.
• "Define north star metrics for live events on Facebook."
• "You're the PM for notifications (the bell icon in Facebook). Why would Meta build this, how would you set goals, and how would you measure success?"

Follow-Ups and Trade-Offs

The initial prompt is the easy part. Meta's evaluation happens in the follow-ups. The interviewer introduces a constraint that forces two metrics into conflict, then watches how you reason through it.

One candidate got this after proposing a handyman marketplace: "The monthly active users of this new handyman tool are up from both the consumer and business side. However, monthly active users on Facebook are down. How would you approach that?" Same candidate, different follow-up: "The handyman product is operating well in New York but not in California. What steps would you take to look into the regional difference?"

In the notifications round, the constraint was: "The percentage of users engaging with notifications is going up weekly for the last six weeks, all users, all geographies, all mobile apps. But time on site is stable or declining. What do you do?"

The candidate asked a clarifying question about what "engaging" meant. Was it reading the notification or clicking through? The interviewer revealed it was specifically comment notifications. People were clicking, leaving a reply, and immediately exiting the app.

The standard follow-ups you should expect on any Meta product sense question are "how would you improve this?" and "how would you measure success?" They appear almost universally.

But the trade-off follow-ups are where candidates actually differentiate themselves. The approach that works: clarify the metric definition first, enumerate possible explanations for the conflict, then prioritize which explanation to investigate and why.

Meta's AI Product Sense Round

This is the round no one else has covered in full. Meta began rolling it out in late 2025, and by early 2026, it's becoming standard in the on-site loop for AI-track PM roles.

The format is 30/30. You spend the first 30 minutes running a traditional product sense case: clarifying questions, user groups, pain points, solution. You stop at the solution.

Then the interviewer sends you a link to Meta's internal Llama-based chatbot, and you spend the remaining 30 minutes prototyping whatever you just described.

The tool looks roughly like Vercel's v0. You type prompts, it generates a preview of the app in a side panel. Generation takes five to seven minutes on the first pass. After that, you iterate through follow-up prompts to refine features, add pages, and adjust the UI.

One candidate who went through this in January 2026 described the experience: "I put in my whole prompt notes, my assumptions, my pain points and user groups, and my solution. I told it I wanted a listing page of volunteer opportunities and an onboarding page. It went to work for five to seven minutes and spit out a preview."

The follow-up questions came in three distinct buckets, depending on the interviewer's background:

Technical AI questions: Token optimization, latency, inference compute, retrieval. One candidate was asked: "What's a better way you could have built this prototype with consideration of compute power, token optimization, and latency?" and "Don't you think specifying a pie chart will eat more tokens than letting the LLM decide the chart type?" His friend, who got the map app prompt, was asked mostly about retrieval: where he'd source geographic data, how he'd optimize the LLM for data retrieval, how he'd integrate directional features.

Prompting strategy questions: Why you prompted the way you did, whether there was a more efficient approach, how you'd handle edge cases the LLM surfaced. The interviewer in one case kept intervening mid-prompt: "Oh, let's rethink this."

Product questions: More traditional product sense follow-ups about the prototype itself. User onboarding, data incentives ("How would you incentivize a user to give Meta additional data so we can improve the recommendation algorithm?"), standalone app vs. integrating into Facebook, mobile vs. web considerations.

A Meta PM who conducts these interviews told one candidate directly: "There's really no guidelines as to what we should be asking. We're just judging you based on how you prompt." The follow-ups you get depend heavily on whether your interviewer is an AI specialist or a generalist PM.

The biggest mistake candidates made in this round was spending too much time on UI polish.

One candidate described it: "I was trying to build a really nice prototype versus actually considering backend functionality." He was iterating on dashboards and charts while the interviewer wanted to hear about data sourcing and retrieval. Another error: starting the prototyping phase with only 20 minutes left instead of 30, which eliminated any buffer for the technical follow-ups.

How to prepare for the AI round: Practice on v0 (Vercel) and Lovable. Do three to five full prototypes with a 30-minute timer. Learn the basics of token optimization, latency trade-offs, and retrieval-augmented generation, not at an engineering depth, but enough to answer "why did you prompt it this way?" and "how would you make this more efficient?" Use the loading time productively: go back to your notes, think about out-of-scope features you'd add, and prepare to answer "if this were a production app, what would you do differently?"

Meta Interview Rubric

The general product sense evaluation criteria (problem framing, creativity, user empathy, structured thinking) apply here.

See our Meta PM interview guide.

A few Meta-specific signals came through repeatedly across all four candidates:

User segmentation clarity. One candidate's advice: "Be extremely clear about how you've segmented users and why you've prioritized one segment over another. Provide reasoning behind that prioritization and be extremely clear and succinct." Multiple interviewers pushed on this.

Metrics depth, especially secondary metrics. The Google director who interviewed at L7 got feedback that he didn't spend enough time on secondary metrics for the Facebook events product, even after presenting engagement as the north star. His reaction: "I've launched hundreds of features to billions of users, and they still didn't like my metrics answer." At Meta, listing a north star metric isn't enough. You need guardrail metrics, counter metrics, and the reasoning for each.

How you handle interviewer guidance. One candidate described interviewers as "like weather. You're not sure if you're going to get a sunny day or a rainy day." Some interviewers actively guide you toward a particular framework. Others give zero feedback. At the L7 level, the director candidate said one interviewer responded to his check-in with: "No feedback. You decide what you want to do." You need to be prepared for both extremes.

Connecting everything to Meta's mission. The recruiter told the L7 candidate explicitly: every round, bring it back to Meta's mission of connecting people. This isn't optional. It's expected.

Common Mistakes

Poor time management. One candidate started prototyping with 20 minutes left in a 60-minute AI round instead of 30, which meant he couldn't adequately address the technical follow-ups. The L7 candidate had a smarter approach: he set a timer divided into thirds for each call and checked in with the interviewer at each interval. That check-in itself demonstrated structured thinking and gave him a natural moment to recalibrate.

Insufficient metrics depth. This came up at every level. The L7 candidate, a Google director, was told his metrics answer on the Facebook events question wasn't deep enough. If someone with that background gets dinged on metrics, you should assume you'll need to go further than you think. Prepare north star, secondary, guardrail, and counter metrics for every practice question.

Not asking enough clarifying questions. The candidates who performed best asked specific, scoping questions early. One asked whether "volunteering" meant social volunteering, disciplinary community service, or workplace group volunteering. Another asked whether a contractor solution should live on an existing Meta platform. These questions don't just show thoroughness; they prevent you from spending 20 minutes building in the wrong direction.

Ignoring interviewer signals. One candidate described an interviewer who was "biasing me towards one way or another without necessarily guiding me towards the solution they were looking for." When an interviewer redirects you, follow the redirect. Fighting it costs time and signals inflexibility.

In the AI round: over-indexing on UI. Building a beautiful prototype when the interviewer wants to discuss token optimization and retrieval strategy. The prototype is a vehicle for conversation, not a deliverable.

Not spacing out interview rounds. One candidate recommended 2 to 3 weeks between rounds, noting you could push it to a month. Since Meta uses team matching rather than role-specific hiring, there's less urgency than you'd expect. Use that buffer.

How to Prepare

One candidate said it plainly: "You'll find all the questions online. It's mostly about preparation."

Read Meta product manager interview questions.

Practice at least 20 to 30 questions. Walk through them on your own, talk through them out loud, or write your answers out. Even without a mock partner, the reps matter. You can find Meta product sense questions in our questions database and practice the full format in our PM interview course.

Use a timer. The L7 candidate divided every call into thirds and checked in with the interviewer at each mark. This kept him on track and gave the interviewer confidence he was managing the time well. For the AI round, strictly enforce the 30/30 split.

For AI rounds, practice prototyping under time pressure. Use v0 or Lovable. Do the full exercise: 30 minutes of traditional product sense on a prompt, then 30 minutes of vibe coding. Repeat three to five times. Get comfortable with the loading time (use it to refine your thinking) and learn how to prompt efficiently rather than iterating toward a pixel-perfect UI.

Network with Meta PMs. The L7 candidate talked to 10 Meta PMs before his loop. They gave him different viewpoints on the process, the evaluation criteria, and the cultural expectations. His advice: "There's no substitute for the work. You must practice. And leverage Meta. Go talk to them. They're willing to talk to you."

Study Meta's product ecosystem. Know what Facebook, Instagram, WhatsApp, and Messenger do. Know the mission ("connect people") and bring your answers back to it. Every interviewer expects this.

Prepare for metrics to be harder than you think. Even the Google director found this round punishing. Practice defining north star metrics, secondary metrics, guardrail metrics, and counter metrics for Meta products. Our mock on designing Facebook Movies covers the exact type of product sense question that appeared at L7.

Expect 100+ hours of prep for senior roles. The L7 candidate put in over 100 hours. That's not unusual for director-level big tech interviews. For IC roles the number is lower, but the candidates who did well still described weeks of dedicated preparation.

Space your rounds. Schedule 2 to 3 weeks between interview phases. You can push it to a month if you need more time. Meta's team-matching model means there's no single req expiring while you prepare.