Introduction to Analytical/Execution Questions
Analytical/execution PM interview questions test your ability to set goals, reason about data, and make defensible decisions under uncertainty.
According to our research among top companies, the most common PM interview question today is some variation of “Define the north-star metric/success metrics for X.” A common scenario is for the question to start by defining metrics, then add complexity by asking a conflicting-metric follow-up.
How analytical has changed
First, agentic AI products have introduced new sources of metric noise: bots, AI assistants, and automated workflows can now inflate your engagement numbers in ways older frameworks weren't built to catch.
Second, AI products require metric categories that classic engagement funnels don't cover. Quality, trust, and helpfulness don't map cleanly to clicks or time on page, and interviewers at companies like OpenAI, Apple, and Perplexity are now actively testing whether you understand that difference.
What to expect
Typically, you'll spend the first 20 to 25 minutes on an anchor question where you're asked to define success for a product or experiment. Then, with 5 to 10 minutes left, the interviewer introduces a follow-up that puts pressure on the metrics you just defined.
Asked at Google DeepMind • 
Asked at Meta Follow-up questions
Most interviews end with one of these three:
- Root cause analysis. Your north star dropped. Diagnose it. E.g., “For Instagram Reels, if prioritizing Reels would cannibalize Stories, what would you do?"
- Conflicting metrics. One metric is up, another is down. What do you do? E.g., "You're tracking metrics, and you're seeing the percentage of users engaging with notifications 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 ship decision. Your north star is up, but a guardrail metric is down. Do you ship? E.g., “The LLM fails occasionally right now, what do you launch?”
What interviewers are looking for
In general, interviewers ask analytical/execution questions to assess your ability to:
- Diagnose a broken metric or ambiguous data signal systematically, rather than jumping to the first plausible explanation
- Choose the right metric for what's being built and defend that choice, including knowing where classic engagement metrics can mislead for AI products
- Translate your analysis into a concrete, actionable plan, not just a directional call
- Communicate your reasoning clearly enough that the interviewer can follow along in real time and engage as a thinking partner
- Show genuine curiosity about the problem, rather than executing a memorized framework
A solid answer names the tradeoff, lists considerations on both sides, and reaches a conclusion, even if that conclusion comes with a hedge like "I'd probably lean toward shipping, but I'd want to keep an eye on the guardrail metric."
A senior+ answer goes further on every dimension. Hypotheses are drawn from a fuller picture of what could be wrong, including AI-specific failure modes. The metric choice is defended against real alternatives and paired with a guardrail. The plan is fully specified: who gets it first, what the rollout sequence is, and the exact condition that would trigger a reversal. The interviewer is treated as a thinking partner throughout, and the questions asked reveal genuine curiosity rather than just what's needed to move to the next step.
The next lessons in this module break down the concepts you need to know to ace analytical/execution rounds and exactly how to implement them.