1. Product sense — "How would you improve X?" (2–3 rounds at most FAANGs)
2. Execution — "You launched X, metric dropped, what do you do?" (1–2 rounds)
3. Estimation — "How many pizzas are eaten in NYC in a day?" (1 round, sometimes bundled)
4. Strategy — "Should company X launch Y?" (senior PM+ mostly)
5. Behavioral — "Tell me about a time…" (every round has some of this)

CIRCLES is the most widely-used framework for product sense at FAANG. It stops you from rambling and makes your thinking visible to the interviewer.

1. C — Comprehend the situation: Restate the question, clarify scope.
2. I — Identify the customer: Who exactly are we solving for?
3. R — Report customer needs: What do they need? Rank by frequency & severity.
4. C — Cut through prioritization: Pick ONE need. Say why (impact, effort, strategic fit).
5. L — List solutions: Brainstorm 3+ options — some safe, some ambitious.
6. E — Evaluate tradeoffs: For each, name the pros/cons and pick.
7. S — Summarize: State your recommendation, metric to track, and what would change your mind.

Example: "How would you improve YouTube for kids?"

Watch how CIRCLES turns a vague question into a structured, offer-winning answer:

• C: "Are we optimizing engagement, or trust from parents?" — pick one.
• I: "Two personas: kids under 12, and their parents."
• R: "Parents want safety + limits. Kids want variety + autonomy."
• C: "I'll optimize for parents — long-term trust unlocks kids' engagement."
• L: "Three ideas: (1) content mode for age tiers, (2) creator badges, (3) time-limit gamification."
• E: "Age tiers is highest impact, medium cost. Badges is low impact. Time limits improve trust but risk kids leaving."
• S: "I'd ship age tiers first. Metric: 30-day retention among parents who set them up. Kill signal: kid DAU drops 15%+."

Execution questions typically say: "Metric X changed. What do you do?" Your job is to structure the diagnosis:

1. Clarify the metric: How is it defined? What's the exact time window?
2. Segment the drop: By user cohort, platform, geo, feature, entry point.
3. Form hypotheses: Product, tech, external (competitor, seasonality), measurement.
4. Prioritize by likelihood + impact: Which hypothesis is most likely and most impactful?
5. Investigate + verify: What data would confirm it? Session replays, logs, funnel drops.
6. Recommend action + measurement: What do you ship, and how do you know it worked?

Example: "DAU dropped 12% last week — what do you do?"

"First — is the drop real or a measurement bug? Check event pipeline health. Then segment: by platform (iOS/Android/Web), by geo, by cohort (new vs returning). Once I isolate where the drop is concentrated, I form hypotheses. If it's new-user acquisition, could be a marketing channel change. If it's returning users, could be a bug in a recent release. I'd pick 2 hypotheses, ask an analyst to pull the data, and have an answer within 24 hours."

Interviewers don't care that you get the number right. They care that you break the problem down clearly, make sensible assumptions, and sanity-check your answer.

The 4-Step Approach

1. Restate & clarify: "Are we estimating pizzas eaten by residents only, or including tourists? Slice-level or whole-pie?"
2. Structure the problem: Population × % who eat pizza × frequency × slices per meal
3. Estimate each variable: Show your reasoning ("NYC = 8M people, ~70% eat pizza, ~1x/week, ~3 slices/meal")
4. Sanity check: "That's 24M slices/day, or 3M pizzas — feels right for a city of 8M."

Every behavioral answer follows STAR: Situation, Task, Action, Result. Keep each part to 1–2 sentences except Action, which gets 3–4.

• Situation: Set the context in one sentence.
• Task: What was your specific responsibility?
• Action: What did YOU do (not the team)? What decisions did you make?
• Result: Metric-driven outcome. What did you learn?