Product metrics PM interview questions
Contents:
Why metrics dominate the PM loop
The interviewer has thirty minutes and one job: figure out whether you can translate a product decision into a number and back. A PM who knows metrics only from textbooks is helpless on the actual job — they cannot tell you which metric matters right now, how to break down a drop, or where to validate a hypothesis. So the loop probes three things in tight rotation: your vocabulary (do you know what the terms mean), your causal reasoning (do you know what moves the metric), and your decomposition reflex (can you split an aggregate into components on demand).
The questions show up in three formats, and you should be able to spot which one you are being asked within the first sentence. Definitional questions sound like "what is 7-day retention?" — they test recall and precision. Analytical questions sound like "metric X dropped 15% — where do you look?" — they test structured thinking under uncertainty. Metric-design questions sound like "we're launching feature Y, what do you measure?" — they test product judgment. Most candidates blur the three and answer every question with a long monologue; the strong ones name the format and stay inside it.
A clean answer is usually under ninety seconds, has at least one number in it, and ends with "and that's why I'd track X over Y." If you cannot land that structure, no amount of framework name-dropping will save you.
North Star and the metric hierarchy
The North Star is the single metric that captures the core value your product delivers to a user. Spotify uses minutes listened. Airbnb uses nights booked. DoorDash uses deliveries completed. A good North Star sits close to user value, not to revenue — revenue is a lagging consequence of value, and optimizing for it directly produces dark patterns.
Three properties separate a real North Star from a vanity metric:
| Property | Why it matters | Bad example |
|---|---|---|
| Tied to user value | Survives strategy pivots | Sign-ups (value not yet delivered) |
| Measured at least weekly | Enables fast iteration | Quarterly NPS |
| Movable by product, not marketing | PM team has agency | Paid traffic volume |
Under the North Star sits a three-level hierarchy. L1 is the North Star itself — one metric. L2 holds five to seven drivers: new users, activation rate, retention, frequency, revenue per active user. L3 is operational — funnel conversion at each step, time-to-first-action, feature adoption. Expect to draw this tree for a product the interviewer picks on the spot, so rehearse on three or four products you actually use. If you can sketch the hierarchy for Notion, Linear, and Uber Eats in under two minutes each, you will pass this question almost every time.
Engagement metrics
DAU, WAU, and MAU count unique active users in a day, week, or month. The trap is the definition of "active" — is a login enough, any event, or only an event inside the core flow? Always force the interviewer to define it before you compute anything; if they refuse, define it yourself out loud and proceed. Sloppy candidates compute DAU off session_start and miss that half the sessions are background pings.
Stickiness = DAU/MAU measures what share of your monthly base shows up every day. Social networks run 50%+, productivity tools sit around 20-30%, and rarely-used utilities live at 5-15%. A favorite trick question: "stickiness went up — good or bad?" The honest answer is it depends on which side moved. If DAU climbed faster than MAU, yes. If MAU collapsed and DAU stayed flat, your ratio improved because you lost casual users — that is churn dressed as a win.
Active days per user is a subtler cousin of stickiness. Instead of a ratio, it counts how many distinct days each monthly user was active. A product moving from 4 to 7 active days per user has genuine habit formation; one where DAU/MAU ticks up because new sign-ups all happen on Tuesday does not.
Load-bearing trick: Whenever you cite a ratio metric, immediately state which numerator and denominator moved. Ratio-only answers are how strong candidates lose easy points.
Activation and retention
Activation is the moment a new user has experienced the core value for the first time. Slack's classic activation was 2,000 messages sent in a team workspace. Facebook's was 7 friends in 10 days. For an interview question on a hypothetical product, pick the action most predictive of D7+ retention and defend it with a sentence about why that action signals real value reception.
Retention is the share of users who come back N days after first visit. Two flavors trip people up:
- Classic retention — the user returned on exactly day N
- Range retention — the user returned on any day between N and N+M
If the interviewer does not specify, ask. Defaulting to classic is safe but the question is the actual signal of seniority. The retention curve plots day 0 through day 30 or 90: a healthy product shows a steep early drop followed by a flat plateau — the "smiling curve." If your curve never plateaus, you have no habit loop and the product leaks users forever.
Cohort retention groups users by their acquisition week or month and tracks each group separately. A classic interview probe: "if we broke activation in February, what does the cohort table show?" The February cohort drops below earlier cohorts starting at D1 and never catches up — and you would see that pattern even if aggregate DAU looked stable because new traffic masked the bleeding.
Monetization and unit economics
ARPU = Revenue / Active users measures average revenue per active user, useful for comparing segment profitability. ARPPU = Revenue / Paying users strips out non-payers, so a rising ARPPU with a flat ARPU means each payer spends more but you cannot convert more of the base to pay.
LTV (Lifetime Value) is total expected profit from a single user. The interview-ready formula:
LTV ≈ ARPU × (1 / churn) × marginBe ready to decompose. "What happens to LTV if we cut churn in half?" Answer: LTV doubles, all else equal — and that is exactly why churn is the highest-leverage lever in most subscription businesses.
CAC (Customer Acquisition Cost) comes in two variants — blended across all traffic, or paid-only across only paid channels. Force the interviewer to pick before you do arithmetic. The healthy LTV/CAC ratio sits around 3:1; below that the unit economics break, above 5:1 you are likely underinvesting in growth. Payback period — months until CAC is recovered from gross margin — should clear in under 12 months for SaaS and faster for consumer.
| Metric | Healthy range | Red flag |
|---|---|---|
| LTV/CAC | 3:1 to 5:1 | < 1.5:1 or > 8:1 (underinvesting) |
| CAC payback (SaaS) | 6-12 months | > 18 months |
| Gross margin (SaaS) | 70-85% | < 50% |
| Net revenue retention | 100-130% | < 90% |
Decomposing a metric
This is the single most-tested skill across every PM loop you will sit. Any metric can be factored into components, and the candidate who can do it live on a whiteboard wins.
Revenue = Users × Conversion × ARPU
ARPU = ARPPU × (Paying / Total)
DAU = New + Returning − Churned
Retention = Activated_today / Cohort_sizeDecomposition serves two purposes. First, diagnosis: if revenue fell, was it the user base, the conversion rate, or the spend per converter? You cannot fix what you cannot localize. Second, experiment targeting: improving conversion versus improving ARPU lead to different roadmaps with different timelines, and the right answer depends on which component has the most headroom.
Practice decomposing DAU, retention, revenue, and funnel conversion three levels deep until you can do it without thinking. This is the kind of pattern drilling that compounds — one hour of decomposition reps beats ten hours of reading metric definitions.
The "metric dropped" case
The most common analytical question. Setup: "DAU fell 15% week over week. What do you do?" Spend five minutes structured like this:
Clarify (30 seconds). Which DAU exactly? Defined off what event? Sudden cliff or gradual decline? First time we've seen this drop or recurring? All segments or one cohort? Without clarification you are answering a different question than the one being asked.
Hypothesize (60 seconds). A finite list, named explicitly: shipping bug, marketing spend cut, seasonality, regression in latest release, competitor launch, platform change (App Store, Google Play, iOS update). State all six even if some feel unlikely — the structure scores higher than the specific guess.
Plan the checks (2 minutes). Cheapest first. Deploy logs and release notes → split by platform and country → split by new vs returning → check tracking SDK version → fraud and bot filters. Each check should kill one hypothesis or sharpen it.
Decide on findings (60 seconds). If technical, fix and run a postmortem. If marketing, sync with growth. If external and temporary, do nothing visible — log it. If external and structural, escalate to strategy.
Gotcha: Never propose a solution before you have a cause. "Let's redesign onboarding" without diagnosis is the fastest way to fail this round, no matter how sharp the redesign idea is.
Common pitfalls
When candidates first learn metrics, the biggest mistake is reporting means on long-tailed distributions. Revenue per user, time on platform, order value — these all have fat right tails where the mean is dragged by whales. Median (or a trimmed mean) tells the real story. Reporting "average order value is $87" when the median is $24 misleads the entire room, and a senior interviewer will catch it.
Another trap is conflating classic and rolling retention across teams. If the company calculates retention differently than your last shop, your numbers will silently disagree by ten or twenty points. Ask. Do not assume. This is also why the very first metric you cite in any answer should come with its exact definition spoken aloud.
Ignoring seasonality is a third killer. A 5% DAU drop over the winter holidays is normal for B2B SaaS and abnormal for streaming. Mention seasonality in the first sixty seconds of any "metric moved" analysis or you will be asked about it anyway.
A fourth pitfall: collapsing the dashboard into a single number. "DAU is up" without guardrails could be bots, paid spam, or a retention crash masked by new acquisition. Always read two or three metrics together — North Star plus a guardrail plus a counter-metric. PMs who hand the CEO a single up-and-to-the-right line invariably get burned at the next board meeting.
The last trap is quoting LTV with no horizon. "LTV is $400" is meaningless. LTV at 6 months? At 24? At infinite-horizon discounted to today? The horizon changes the number by 3-5x for most products, and the candidate who omits it sounds like they have never shipped a real LTV model.
Related reading
- North Star metric for PM
- How to choose a North Star metric
- DAU explained for PM
- Activation framework for product managers
- PM case interview guide
If you want to grind through PM metric scenarios like these daily, NAILDD is launching with 500+ product cases drilling exactly this format.
FAQ
Which metrics must a junior PM know cold?
DAU and MAU, retention at D1/D7/D30, funnel conversion, activation, ARPU, LTV, and CAC. That is the screening minimum — without fluent answers on these you will not clear the recruiter call, let alone the onsite. The bar is not depth on all of them, but instant recall of the definition plus one concrete benchmark range per metric.
What's the difference between retention and churn?
They are two sides of one coin. Retention is the share of users who stayed; churn is the share who left. If 30-day retention is 30%, 30-day churn is 70%. Most teams use both terms in the same conversation and switch between them by audience — engineering hears retention, finance hears churn — so be ready to translate on the fly.
Is it better to know many metrics shallowly or a few deeply?
Deeply, by a wide margin. A candidate who can decompose DAU, retention, activation, and LTV three levels deep in under thirty seconds each will outscore one who name-drops twenty metrics with vague definitions. Interviewers probe depth by asking follow-ups, and shallow knowledge collapses within two questions.
How should I prepare in the week before the loop?
Pick five products you use daily and write out the metric hierarchy for each on paper — L1 down to L3. Then for each one, write the decomposition tree for DAU, revenue, and retention. Finally, run three "metric dropped" cases out loud with a friend timing you. That single week of drilling beats two months of reading framework articles, because the loop tests live structured thinking, not memorized definitions.
Is this an official source?
No. This article is based on publicly available materials and patterns reported by candidates. Specific metric definitions vary between companies — always confirm the company's own definitions during the interview before quoting numbers.