Product metrics interview questions
Metrics questions come in two kinds: choose the number a product should steer by, or work out why a number moved. Both test whether you can tie a metric to the value a user actually gets.
Below: what the round tests, the method in short, and 40 practice questions, each with a first move and the outline of a strong answer. All free, no sign-up.
What the interviewer is scoring
- Whether your metric counts real value, not opens, sign-ups or anything that only goes up.
- Whether you check that a change is real and find where it lives before you guess at causes.
- Whether you name guardrails, and what you would do next with the answer.
Watch a mock interview from this round
Grace’s first try, Ava’s pushback, then the stronger answer with its scorecard. The question: Daily active users on a messaging app dropped 10% week over week. How would you investigate?
Read what Ava and Grace say
- Ava:
- I'm Ava. Today I'm interviewing Grace, on a drop in daily users.
- Grace:
- I'm Grace. I'm preparing for product manager interviews.
- Ava:
- Grace, daily active users on our messaging app dropped 10% this week. How would you investigate?
- Grace:
- Probably a competitor launched something. I'd ship a new feature to win people back.
- Ava:
- Slow down. How do you know the drop is even real? Try again.
- Grace:
- Right. First, is it real? I'd compare the dashboard with raw server logs. If the logs are flat and only the dashboard fell, it is a tracking break, and I stop. I'd ask if the definition of a daily user changed, and check the pipeline ran in full. Say all three are clean, so the drop is real. Next, size. Ten percent against a normal weekly swing of two is a real move. It was a step on Tuesday, not a slide. Messages sent fell ten percent too, while sign-ups held. So existing users went quiet. Then I slice, by platform, country, app version, and new versus existing users. It lives on Android, version 8.2, in every country. Now suspects. Inside: Tuesday's 8.2 release, a notification experiment, a pricing change. Outside: an Android update, a competitor, a holiday. A competitor or a holiday would hit iPhone too, so they drop out. The experiment ran on both platforms, so it ranks last. The release fits the day, the platform and the version, so it ranks first. The Android update is second. To settle it. My likeliest cause is the 8.2 release. The proof: compare 8.2 with 8.1 on the same Android version. If only 8.2 is down, it is the release. If both are down, it is the update. If it is the release, I roll back behind a flag today. I tell the Android lead, so the fix has an owner, and support, so they can answer users. I call it fixed when daily users are back within two percent of last week for three days.
- Ava:
- That's the order. You proved it was real before you chased a cause, and you said what fixed looks like. That would score well.
The method in short
Metrics: pick the metric: Value Moment
- Name the valueWho gets what from the product, in one sentence. Name both sides of a marketplace.
- Find the momentThe action that shows a user just got that value, not that they opened the app.
- Make it a numberDecide what you count, over what window and with what qualifier. Match the window to how often the value happens.
- Stress-test itDoes it rise only with real value? Can a team game it? Does it lead revenue and retention? Will it move within weeks?
- Guard itAdd one to three guardrails that catch the damage if a team chases it too hard.
Example
A recipe app: a saved recipe is a promise and a cooked one is the value, so steer by recipes cooked per week, guarded by unsubscribes from its reminder emails.
Most common mistake
Reaching for revenue, sign-ups or plain daily users by reflex. They lag, only ever go up, or count opens that gave nobody anything.
The method in short
Metrics: diagnose a change: Change Drill
- Check it’s realRule out logging, pipeline and definition changes before you hunt for causes.
- Size itMeasure how large it is, when it started, whether it was a step or a slide, and which other numbers moved.
- Slice itSplit by platform, region, user type and the metric’s parts until you find where the change lives.
- Line up suspectsInternal changes (releases, experiments, prices) and external ones (season, competitors, OS updates) that fit the timing and the slice. Rank them.
- Settle itName the likeliest cause, how you would prove it, what you would do, and whom you would tell.
Example
Sign-ups down 12% since Tuesday: logging is unchanged, the drop is all on Android, and Tuesday’s release changed the sign-up screen, so roll it back behind a flag and confirm with tomorrow’s numbers.
Most common mistake
Guessing at causes (a competitor, the weather) before checking the drop is real and finding where it lives.
40 product metrics questions to practice
Pick one, answer it out loud on a timer, then open its page to compare with the outline. Start with the warm-ups if this round is new to you. Want a score? Paste your answer into the free answer grader, or answer it out loud with Ava.
Warm-up questions (10)
- What would be the north star metric for Spotify?First move: Start from the value a listener gets, not from revenue. Ask what single behavior shows people are getting that value often.
- What metric would you pick as the north star for a meditation app?First move: Ask what outcome users want, which is usually feeling calmer or sleeping better, then find a behavior that predicts it.
- What is the north star metric for Duolingo, in your view?First move: Separate habit from learning. Daily use is easy to measure, but the real value is getting better at a language.
- You're launching saved searches on a home-listing site. What goal would you set for the first quarter?First move: Start with what the feature should change for users (coming back when new homes appear), then pick a goal that is ambitious but grounded in a baseline.
- What metrics would you track for a newly launched podcast app in its first 90 days?First move: Early on, you are testing whether people find value at all. Focus on activation and retention before growth.
- Explain what a p-value means to a stakeholder who has never taken statistics.First move: Use a plain analogy: how surprised should we be by this result if the change had no effect at all?
- Break down revenue for a subscription streaming service into a metric tree.First move: Start with revenue equals subscribers times average revenue per subscriber, then break each part down further.
- What metrics matter most for a B2B project management tool?First move: Remember the buyer and the user are different. Measure whether teams use it, because that drives renewals.
- You shipped a new onboarding flow for a photo-editing app. How do you know it is working?First move: Define what a successful onboarding outcome is, such as editing and saving a first photo, and compare users who got the new flow with those who did not.
- How would you measure the success of a loyalty program in a coffee chain's app?First move: Ask what the program should change: more visits per customer, bigger orders, or keeping customers from going to competitors.
Standard questions (22)
- Choose a north star metric for Airbnb and defend it.First move: Remember the two sides of the marketplace. A good north star must reflect value for both guests and hosts.
- How would you measure the success of Instagram Stories?First move: Ask what Stories is for (sharing casual moments, keeping people connected) and build a tree from that goal down to creator and viewer metrics.
- How would you measure the success of comments in Google Docs?First move: Start from why comments exist: to help people collaborate on a document without meetings. Measure whether collaboration actually gets done.
- Daily active users on a messaging app dropped 10% week over week. How would you investigate?First move: Before guessing causes, check whether the drop is real: data pipeline, definition changes, and seasonality.
- Uber trips in one city fell 15% yesterday. What do you do?First move: Narrow it down by side of the market: did riders stop requesting, or did drivers stop accepting?
- Checkout conversion on an e-commerce site dropped after a release. Walk me through your diagnosis.First move: The timing points to the release, but confirm it. Compare users on the new version with users still on the old one.
- Food delivery orders are flat, but average delivery time rose by 8 minutes. How do you dig in?First move: Break delivery time into its stages: restaurant prep, waiting for a courier, and the trip itself.
- Search usage on an online marketplace fell 20%, but only on mobile. Diagnose it.First move: The mobile-only pattern is the clue. Look at what changed on mobile: an app release, an OS update, or a redesign.
- Set a quarterly goal for a ride-sharing app's driver retention team.First move: Define what retention means for drivers (active in a given week) and what the team can realistically influence in one quarter.
- An A/B test shows a 3% lift in clicks but a 1% drop in purchases. Do you ship?First move: Go back to the goal: clicks are a means, and purchases are closer to the outcome. Then check whether the purchase drop is statistically real.
- Your A/B test is significant on day two. Should you stop it and ship?First move: Explain the risk of peeking: checking early and often makes false positives more likely. Then mention novelty effects and weekly cycles.
- How would you A/B test a new pricing page?First move: Clarify what changes on the page (layout or actual prices), because testing different prices raises fairness and trust issues.
- Ads revenue would rise 5% if you added one more ad to the feed. Should you?First move: Ask about long-term effects. A short test may show revenue gains while hiding slow damage to engagement.
- Engineering can either cut app load time in half or build the top-requested feature this quarter. How do you choose?First move: Estimate the impact of each on the same metric, so you are comparing like with like rather than gut feel.
- How would you measure the health of a two-sided marketplace like Etsy?First move: Measure both sides and the match between them. A healthy marketplace has buyers finding what they want and sellers making sales.
- What metrics would you use to measure the success of LinkedIn's 'People you may know'?First move: Go beyond connection requests sent. Ask whether the new connections create value, such as messages or feed engagement.
- How would you measure the success of Amazon's 'Buy it again' feature?First move: Ask what the feature should do: make reordering faster for items people already buy regularly. Measure that, not just clicks.
- Define success metrics for a video calling app's new noise cancellation feature.First move: Combine technical quality and user perception. Noise cancellation that measures well but cuts off voices will fail.
- What is a guardrail metric, and which ones would you set for a test that changes news feed ranking?First move: Define a guardrail as a metric you do not want to harm while chasing the main goal. Then list ones a ranking change could hurt.
- Weekly active users are up, but revenue is down. How do you explain this to leadership?First move: Find the link between users and revenue: maybe new users are lower value, or existing paying users are leaving.
- A feature in your app is used by only 2% of users. What would you look at before deciding to remove it?First move: Ask who the 2% are before judging the number. A small group of high-value or highly loyal users can make a little-used feature worth keeping.
- Which would you rather increase for a dating app: matches or conversations? Why?First move: Avoid the vanity metric trap. Ask which one is closer to the user's real goal, which is meeting someone.
Stretch questions (8)
- Sign-ups for a SaaS free trial doubled overnight. Is that good news?First move: Treat a sudden spike with the same suspicion as a drop. Check where the sign-ups came from and whether they look like real buyers.
- Average session length on YouTube went up, but daily active users went down. What's happening?First move: This might be a mix shift. If light users left, the average rises even when nobody watches more.
- How would you A/B test a change in a marketplace where buyers and sellers affect each other?First move: Name the problem: test and control groups share the same supply, so changes leak between them. Then suggest designs that limit the leak.
- A feature increases engagement but also increases reports of harassment. Do you keep it?First move: Do not treat it as a simple trade. Ask who is harmed and how badly, then look for a version that keeps the engagement without the harm.
- Notifications drive 30% of daily opens, but the uninstall rate is rising. What do you do?First move: Check whether the two are actually connected before cutting notifications. Then look for which notifications earn their place.
- How would you measure whether Google Maps' traffic predictions are good?First move: Separate accuracy of the prediction from usefulness to the user. A slightly wrong estimate that still helps someone choose the best route can be fine.
- Listings on a second-hand marketplace grew 40%, but sales grew only 5%. What's going on, and what would you do?First move: Supply outgrew demand or matching got worse. Check whether buyers can find the new listings and whether those listings are good.
- Your team hit its main goal, but customer complaints doubled. What went wrong in how the goal was set?First move: Name the trap: a single metric without guardrails invites gaming or tunnel vision. Then explain how you would reset the goal.
Common questions about this round
What is a north star metric?
A north star metric is the one number a product team steers by, chosen because it rises only when users get the product’s core value. For a recipe app that could be recipes cooked per week, not recipes saved. A good one is hard to game, moves within weeks and leads revenue and retention.
How do I answer “a metric dropped, what do you do”?
Work in order. Check the drop is real (logging, pipelines, a changed definition). Size it: how big, since when, a step or a slide. Slice it by platform, region and user type until you find where it lives. Line up the internal and external changes that fit that timing and slice. Then name the likeliest cause, how you would prove it and what you would do.
What is a guardrail metric?
A guardrail is a number you refuse to harm while you chase the main one. If a team steers by recipes cooked per week and uses reminder emails to get there, unsubscribes from those emails is a guardrail: it shows the damage if the reminders get pushy.
Get all 200 questions as a PDF
Every question in the bank, all seven round types, with its first move, a strong opening, the answer outline and the common mistake.
From PMInterviewKit
One method for each interview round, 88 short lessons to read or watch, 66 worked answers and mock interviews with Ava. The first lesson of every module is free. The full kit is $129 once.