Growth, Activation and Retention Questions
Driving product-led growth across the user lifecycle: onboarding, activation, engagement, retention, and churn prevention. Covers designing growth features, activation funnels and onboarding experiments, building long-term engagement loops, and improving product adoption, value realization, and lifetime value. Also covers diagnosing why users churn and spotting early warning signs, win-back and re-engagement programs, lifecycle messaging and in-product nudges, referral and viral loops, prioritising growth experiments and roadmaps, and weighing short-term gains against user trust. Assesses systematic thinking about how users discover, adopt, stick with, and keep getting value from a product.
Your mobile app loses 60% of new users at account setup, the second onboarding step. You are the PM and have six weeks with design. How do you decide what to change, and how do you know it worked?
Sample Answer
Direct answer
A 60% loss at step 2 means that of every 100 users who start, only 40 get past account setup. In six weeks I would spend the first week diagnosing why, using the funnel data and a handful of user sessions; weeks 2 to 5 shipping one to three targeted changes behind an experiment; and week 6 reading the result. "It worked" means more people complete step 2 without those extra completions dropping out at the next step or turning into low-quality accounts.
Week 1: decide what to change (diagnosis)
Rank causes by data, not opinion:
- Split the 60% by cut: platform and OS version, device, country, acquisition channel, new-versus-returning email, time on the step. A drop that is 85% on one Android version is a bug, not a design problem.
- Break step 2 into its sub-steps: which field or screen do users leave on? (Instrument field-level focus, errors and submit events.)
- Check errors: validation failures, password rules rejecting input, email verification delays, third-party sign-in failures.
- Watch 10 to 15 recorded sessions of people who abandoned, and run short interviews with 5 of them.
- Look at what step 2 asks for that the user cannot yet justify: phone number, payment, permissions, long password rules, profile photo.
Then list candidate changes and score each on users affected, expected completion gain, confidence and effort, and choose what fits six weeks of design and build.
Weeks 2 to 5: ship in a testable way
- Prioritise removal over redesign: drop or defer fields that are not needed now, offer social or passkey sign-in (signing in with a fingerprint, face or device PIN instead of a password), show progress, save partial input.
- Use an A/B test (a random half of new users get the change) rather than before/after, since traffic mix changes week to week.
- Illustrative sizing: baseline step-2 completion of 40%, detecting +5 points (45%) at 5% significance (accepting a 1-in-20 chance of calling a win when nothing changed) and 80% power (an 80% chance of detecting a true +5 point lift) needs 1,531 users per arm. The standard two-proportion formula is n = (z_a + z_b)^2 x [p1(1-p1) + p2(1-p2)] / (p1-p2)^2, where z_a = 1.96 (for 5% two-sided significance) and z_b = 0.8416 (for 80% power) are cut-offs on the normal curve. Substituting: (1.96 + 0.8416)^2 = 7.849; p1(1-p1) + p2(1-p2) = 0.24 + 0.2475 = 0.4875; (p1-p2)^2 = 0.0025; so n = 7.849 x 0.4875 / 0.0025 = 7.849 x 195 = 1,530.6, rounded up to 1,531. With 500 new users a day, that is about 6 days of enrolment for both arms combined, so one test fits inside the window and a second can follow.
Week 6: how I know it worked
- Primary: step-2 completion rate, treatment versus control.
- Downstream check: the share of those users who complete the next activation step (the next action on the way to first value, such as creating a first project), not just step 2. If the extra completers stall one step later, the problem moved, it did not shrink.
- Guardrails (other measures that must not get worse while the main one improves): spam or fake-account rate (a looser step attracts abuse), support tickets about sign-in, time to complete.
- Report a confidence interval (the range of lifts consistent with the data, for example +2 to +8 points), not only a p-value (how surprising the result would be if the change did nothing), and break the result out by platform.
The same method on a B2B SaaS onboarding target
The same sizing logic applies when the goal arrives as a quarterly target. Suppose the product is B2B SaaS (software sold to businesses) and the goal is to cut onboarding drop-off from 45% to 25%. Of every 100 starters, 55 complete now and the target is 75, so the plan must add 20 completers per 100. I would not split that by feel. I would start from the 45 who leave and use the diagnosis (funnel plus sessions) to say where they leave. Illustratively: 20 leave at an empty first screen, 15 leave because they cannot tell which step is required, and 10 leave at the "invite your team" and billing steps. Each bet is then stated as a hypothesis with an assumed recovery rate for the leavers it addresses (illustrative assumptions that the experiments must confirm):
- Import sample data or the user's existing data so the first screen is not empty: recovers 45% of the 20, which is +9.
- A progress checklist with the one required step first: recovers 40% of the 15, which is +6.
- Defer the "invite your team" and billing steps until after first value: recovers 50% of the 10, which is +5.
9 + 6 + 5 = 20 points, exactly the gap. I would run the bets as sequenced experiments and re-forecast after each: if the first gets only +4, I cut scope or add a bet rather than assume the quarter will still land.
Trade-offs and pitfalls
- Optimising one step in isolation: completion goes up, activation does not.
- Cutting a field that the business needs (legal consent, verification). Replace with progressive collection after sign-up.
- Mixing four changes in one release, so no one knows which worked. Ship the biggest single change first.
- Not segmenting: a 60% average can be 15% on iOS and 95% on one old Android build.
Users with below-median engagement are 30% of active users but 10% of revenue. Propose three prioritised, data-driven recommendations to improve their retention. For each, give the expected direction of change in churn and LTV, the data you would need to validate it, and a metric for a pilot.
Sample Answer
Direct answer
These users are 30% of active users but only 10% of revenue, so each earns roughly a third of what the average active user does: 10% of revenue / 30% of users = 0.33, where the average user is 1.0. The other 70% of users bring 90% of revenue, so 90 / 70 = 1.29 times the average. The other users therefore earn 1.29 / 0.33 = 3.9x as much as the low-engagement group (worked below). I would not treat them as one group to "re-engage". I would split them by why engagement is low, and run three prioritised plays: (1) guide new low-engagers to a first valuable habit, (2) tailor reminders to what each user already does, (3) offer a lighter plan or reduced-effort path to casual-but-happy users. Churn (the share of customers who leave each month) should fall and LTV (lifetime value: the profit a customer brings before leaving) should rise slightly for each, but I would size them honestly because the revenue pool is small.
Size the prize (illustrative)
Take 100,000 active users and 1,000,000 USD monthly revenue.
- Low-engagement group: 30,000 users, 100,000 USD, so 3.33 USD per user per month.
- Everyone else: 70,000 users, 900,000 USD, so 12.86 USD per user per month.
- Ratio: 12.86 / 3.33 = 3.9x.
If churn among the low group falls from 8% to 7% a month, expected lifetime rises from 12.5 to 14.3 months (+14.3%). Because revenue per user is small, the gain per user is about 3.33 USD x 1.8 months = 6 USD. The value is in the number of users, not per user, and in what the group costs to serve and whether they become heavier users.
Three prioritised recommendations
1. Guided path to a first habit (highest priority). Many low-engagement users never found the product's core use. Show a short sequence to the one action that predicts staying.
- Expected direction: churn down, LTV up (via longer lifetime and a chance to move up an engagement tier).
- Data needed: which early actions separate those who later become engaged from those who do not (a cohort, meaning a group of users who signed up in the same period, split by first-week actions), plus time since signup (a low-engagement new user is a different problem from a low-engagement veteran).
- Pilot metric: share reaching the habit action within 14 days, with a random holdout (a randomly chosen group of eligible users who do not get the change, kept as the comparison).
2. Personalised, opt-in reminders based on past use. Light digests tied to each user's own history rather than blanket messages.
- Expected direction: churn down modestly; a risk is notification fatigue (users tuning out or switching off messages because there are too many), so watch opt-outs.
- Data needed: per-user activity patterns, notification opt-in and opt-out history, response rates by time and channel.
- Pilot metric: weekly active rate of the treated group, with unsubscribe rate as a guardrail.
3. A lower-effort plan or pause for casual users. Some people are satisfied but use the product rarely. A cheaper or pausable plan keeps them as customers instead of cancelling.
- Expected direction: churn down, ARPU down, LTV up only if the retained months outweigh the price drop. This one has the trickiest direction, so test first.
- Data needed: price sensitivity (usage versus plan price), cancellation reasons, exit survey.
- Pilot metric: retained paying users at 90 days and revenue retained per user, not ARPU (average revenue per user) alone.
Prioritisation logic
Order by users reached x expected change x confidence / cost. Number 1 is cheap and broad, so it goes first. Number 3 changes revenue per user, so it goes last and gets a careful test.
Trade-offs and pitfalls
- Low engagement does not always mean churn risk; some users pay for insurance-like value (they rarely use the product but keep it for peace of mind, like a backup tool or a safety net feature). Check churn rate before assuming a problem.
- Raising engagement for its own sake can create noise; tie each play to retention.
- Compare against a holdout; users selected for being low-engagement may bounce back on their own (regression to the mean: a group picked because it is at an extreme low tends to move back toward average even with no intervention, so a pilot would take credit for a recovery that was coming anyway).
A mid-market customer finished implementation two weeks ago, but fewer than one in five of their licensed users has done anything meaningful and the time to a first valuable action is long. How do you diagnose it, and what do you change in the product or onboarding to fix it?
Sample Answer
Direct answer
Treat it as a funnel from licence to value and find where the 80% went missing before changing anything. Break "fewer than one in five licensed users has done anything meaningful" into stages (provisioned, invited, logged in, took the first valuable action), then split each stage by team, role and setup path. The cause is usually one of four things: users never got access, users never heard why to use it, the first valuable action needs a setup step nobody finished, or the licences were bought for people who do not need the product. The fix follows the stage that leaks most: product changes for friction, onboarding changes for awareness and setup.
(Terms: provisioned means the account is created and a licence assigned to the person. An event is a logged record that the product writes each time a user does something, such as sharing a report; "events fire" means those records are actually being written. Enablement is the training and help that teaches a team to use the product. Right-sizing licences means matching the number of paid seats to the people who will use them. Mid-market means mid-sized customers, roughly a few hundred to a couple of thousand employees. A Solutions Architect or implementer is the vendor-side person who sets up the customer's account. Activation means a user reaches the first action that delivers real value. Time to value is the elapsed time from getting access to that action. SSO is single sign-on, logging in with the company identity provider. A champion is the customer-side person who pushes adoption internally.)
Step 1: Define and instrument before diagnosing
Agree on the "meaningful action" with the customer's stated goal (for example "created and shared a report", not "logged in"). Confirm the events fire for this account, meaning the product really records each action for these users (a mid-market customer on SSO or an unusual setup can silently drop events, which would show up as users who did nothing when they did). Pull time to first value as a distribution, not an average.
Step 2: Build the licence-to-value funnel (illustrative account: 200 licences)
| Stage | Users | Drop from previous stage |
|---|---|---|
| Licensed | 200 | none |
| Invited or provisioned | 160 | 40 (20%) |
| Logged in at least once | 120 | 40 of 160 (25%) |
| Took the first valuable action | 38 | 82 of 120 (about 68%) |
38 of 200 is 19%, matching "fewer than one in five". The biggest relative leak is login to first valuable action (about 68%), and the biggest absolute loss across the early stages is split between access (80 users never logged in) and value (82 users logged in then stopped). So I would work both, starting with the one with the faster fix.
Illustrative findings for this account after the cuts: of the 160 invited, 120 logged in, and 30 of the 40 who did not log in share one email domain (a subsidiary whose invites were filtered as spam). Of the 82 who logged in and stopped, 51 (62%) stopped at data import, 19 at permissions and 12 at an empty first screen (51 + 19 + 12 = 82). And 30 of the 38 users who took the first valuable action sit in two departments, finance and operations. That points to a fixable import step plus a fit problem for the other teams, which is the diagnosis the rest of the plan follows.
Step 3: Cut each leak to find its cause
| Question | Cut | What it tells you |
|---|---|---|
| Did invites reach people? | Invite sent vs accepted, by email domain and team | Provisioning or SSO misconfiguration, spam filtering, wrong user list |
| Do some teams adopt and others not? | Actions per user by department and role | A fit problem (licences given to people who do not need it) vs an enablement problem |
| Where do logged-in users stop? | Event path of the 82 who stalled: last step reached | The specific step with friction (data import, permissions, empty-state confusion) |
| Is the first value action blocked by setup? | Integration or data-connect status vs activation | The product needs data before it can show anything |
| Is there a champion? | Admin activity, who sent invites | No internal owner means no internal push |
Step 4: Changes, in the order I would make them
- Remove the blocker found in Step 3. For example, if users stall at data import, ship sample data and templates so the first action works with zero setup, and make import asynchronous with a progress screen.
- Role-based onboarding. Ask the role on first login and show that role's shortest path to its first valuable action, instead of one tour for everyone.
- Admin adoption view. A page for the customer admin showing adoption by team, who has not logged in, and a one-click reminder. This gives the champion a lever.
- Targeted enablement. A short live session for the two teams with the highest potential, scheduled around their real workflow, not a generic training.
- Right-size licences. If one department will never use it, say so and re-allocate seats, which protects renewal trust.
Success measure and pitfalls
Target (illustrative): 14-day activation of licensed users from 19% to 45% within 60 days, plus median time to first valuable action. Why 45% and not more: if every licensed user is invited (200), 75% of them log in as now (150) and 60% of those take the first valuable action once import is fixed (90), that is 90 of 200 = 45%. Today those three rates are 80%, 75% and about 32%, so the target assumes fixing access fully and roughly doubling the login-to-action rate, not that every user adopts. Pitfalls: counting logins as adoption, fixing the account with a manual rescue rather than the product cause (if three similar accounts show the same stall, it is a product defect, not an account problem), and a Solutions Architect or implementer closing "implementation complete" without checking an activation criterion. What would change my call: if the stalled users are all in teams that were never meant to use it, the answer is licences and scope, not onboarding.
Explain RICE and ICE scoring for prioritising growth experiments. Rank these three using RICE: A) reach 50k, impact 1.2, confidence 0.6, effort 5; B) reach 10k, impact 2.0, confidence 0.8, effort 3; C) reach 200k, impact 0.2, confidence 0.5, effort 8. When would ICE be preferable, and what would make you distrust the ranking?
Sample Answer
Direct answer
RICE is a scoring formula that ranks ideas by reach times impact times confidence, divided by effort; ICE drops reach and replaces effort with ease. Applying RICE to the three ideas gives A = 7,200, B = 5,333 and C = 2,500, so the ranking is A, then B, then C. ICE is preferable when reach is similar across ideas or cannot be estimated, or when you need a fast, rough ordering. I would distrust the ranking when its winner depends on a single soft number, especially impact.
What the terms mean
- Reach: how many users the change touches in a stated period (for example per quarter).
- Impact: the size of the effect on the target metric for each person reached, on a consistent relative scale (the question's 1.2, 2.0 and 0.2 do not name their scale, so I assume one shared scale such as the common 0.25 minimal, 0.5 low, 1 medium, 2 high, 3 massive, on which 1.2 is a little above medium, 2.0 is high and 0.2 is just under minimal).
- Confidence: a 0-1 factor for how sure you are about reach and impact; low evidence should pull the score down.
- Effort: person-months (one person working for one month; or weeks) of total work across product, design and engineering.
- ICE: impact, confidence and ease (each often scored 1-10), combined by multiplying or averaging. It has no reach term and no effort in time units.
The calculation
RICE=EffortReach×Impact×Confidence
A=550,000×1.2×0.6=7,200B=310,000×2.0×0.8≈5,333C=8200,000×0.2×0.5=2,500
| Idea | Reach | Impact | Confidence | Effort | RICE | Rank |
|---|---|---|---|---|---|---|
| A | 50,000 | 1.2 | 0.6 | 5 | 7,200 | 1 |
| B | 10,000 | 2.0 | 0.8 | 3 | 5,333 | 2 |
| C | 200,000 | 0.2 | 0.5 | 8 | 2,500 | 3 |
C has the largest reach but the tiniest per-person effect, so breadth alone does not win. A beats B by 35% (7,200 / 5,333 = 1.35), which is smaller than the error in most impact estimates.
ICE on the same three ideas (illustrative scores)
ICE scores each idea 1-10 for impact, confidence and ease (the inverse of effort), then multiplies. Rescaling the inputs above, with impact 1.2 to 6, 2.0 to 10 and 0.2 to 1, confidence 0.6, 0.8 and 0.5 to 6, 8 and 5, and ease 5, 7 and 2 for efforts of 5, 3 and 8:
| Idea | Impact | Confidence | Ease | ICE | Rank |
|---|---|---|---|---|---|
| A | 6 | 6 | 5 | 180 | 2 |
| B | 10 | 8 | 7 | 560 | 1 |
| C | 1 | 5 | 2 | 10 | 3 |
ICE puts B first where RICE put A first, because ICE ignores that A reaches five times as many people (50,000 against 10,000). That flip is the cost of dropping reach, and it shows why ICE is only safe when audiences are comparable.
When ICE is preferable
- Reach is about the same across ideas (all changes hit the whole signup flow), so the reach term only adds noise.
- You cannot size reach credibly yet, for example an early product with unstable traffic.
- Speed matters more than precision, such as a weekly backlog triage in which a team scores ideas by gut and then refines the top few.
- Caution: ICE's 1-10 scores are subjective and drift between people; calibrate by scoring a few past experiments with known outcomes.
What would make me distrust the ranking
I check how far an input must move to flip the order (a sensitivity check):
| Change | Threshold that flips the order | Reading |
|---|---|---|
| A's impact (others fixed) | Below about 0.89 (from 1.2) | A swaps with B. A 26% overestimate is enough |
| A's effort | Above 6.75 person-units (from 5) | A swaps with B. A 35% underestimate is enough |
| B's impact (others fixed) | Above 2.7 (from 2.0) | B overtakes A. Confidence is already 0.8 and cannot rise beyond 1 to compensate: B at confidence 1.0 scores 6,667, still behind A |
| C's impact | Above about 0.43 to pass B, about 0.58 to pass A (from 0.2) | Needs 2.1 times today's value to pass B and 2.9 times to pass A, so C stays last for any plausible error |
So the A-over-B ordering is fragile (impact and effort are the least reliable inputs), while C's last place is robust. Other reasons for distrust: impact scored on different scales by different authors, reach counted over different time windows, overlapping audiences (A and B both reach the same users, so benefits do not add), confidence inflated for ideas someone loves, and effort estimates that leave out design, QA and analytics work.
Recommendation
Do A first only if a cheap check supports its impact estimate; otherwise run the fastest test of A's impact assumption, because that single input decides the A-or-B call, and keep C off the list. Scores order a backlog; they do not replace testing.
A freemium product converts 10% of users to paid within 30 days. Paid ARPU is 8 USD per month, monthly churn among paid users is 10%, and referred users convert 25% more often. Propose three experiments to improve LTV, justify the expected magnitude of each, and name the metrics you would monitor.
Sample Answer
Direct answer
The baseline: 10% of free users convert, paid ARPU (average revenue per paying user per month) is 8 USD, monthly churn (share of payers who cancel each month) is 10%. That makes expected paying lifetime 1 / 0.10 = 10 months and LTV (lifetime value) per paying user 8 x 10 = 80 USD, or 8.00 USD per free user (0.10 x 80). I would run three experiments, one on each term: conversion, churn, and a referral push, and expect roughly +10%, +11% and +5% on LTV per free user respectively. Revenue basis, no margin applied, because the question gives no margin.
Baseline arithmetic (checked by running it)
- Lifetime = 1 / churn = 1 / 0.10 = 10 months.
- LTV per paid user = ARPU / churn = 8 / 0.10 = 80 USD.
- LTV per free user = conversion x LTV per paid user = 0.10 x 80 = 8.00 USD.
- Referred users convert at 10% x 1.25 = 12.5%.
Assumption most results hinge on: churn is constant at 10% per month. In reality churn is usually higher early, so a real cohort survival curve would refine every number below.
Three experiments, sized against the baseline
Experiment A: onboarding and paywall timing (raise conversion). Show the upgrade offer after the user's first successful result, not at signup.
- Expected effect: conversion 10% to 11% (a +1 point lift, +10% relative). I chose 1 point as a plausible target for a single UX change, not a measured fact.
- LTV per free user: 0.11 x 80 = 8.80 USD, +10.0%.
- Metrics: 30-day conversion rate (primary), time to first result, ARPU of new payers (to detect discount-driven conversions), 30-day churn of new payers.
Experiment B: a cancellation save flow and a pause option (cut churn). Offer pause or a lower plan at cancellation, and a day-5 value reminder for new payers.
- Expected effect: churn 10% to 9% per month. Lifetime 1 / 0.09 = 11.1 months; LTV per paid user 8 / 0.09 = 88.89 USD; per free user 0.10 x 88.89 = 8.89 USD, +11.1%.
- Metrics: monthly churn by tenure month, pause-to-resume rate, refund and complaint rate (guardrail), 90-day retained payers vs holdout.
Experiment C: referral prompt for happy users. Because referred users convert 25% more often, shifting the mix toward referrals raises blended conversion. Suppose the prompt raises the referred share of new signups from 0% to 20% (my assumption). Blended conversion = 0.8 x 0.10 + 0.2 x 0.125 = 10.5%. LTV per free user = 0.105 x 80 = 8.40 USD, +5.0%.
- Metrics: referral invitations sent, referred share of signups, referred conversion and churn (does the +25% hold, and do referred users stay as long?), cost of any incentive per referred payer.
- Caveat: referred users may differ (friends of enthusiasts), so the +25% may not apply to an expanded referral pool. Test it, do not assume it.
| Experiment | Lever | New LTV per free user (USD) | Uplift |
|---|---|---|---|
| A | Conversion 10% to 11% | 8.80 | +10.0% |
| B | Churn 10% to 9% | 8.89 | +11.1% |
| C | Referred share 20% | 8.40 | +5.0% |
The effects roughly multiply when all land: 1.10 x 1.111 x 1.05 = 1.283, about +28%, assuming they do not interact (a stretch).
How to run and judge them
- Randomise users to treatment and control; hold out the control to the end.
- A needs enough free users for a 1-point conversion lift to be detectable; B needs a cohort long enough to observe a 1-point churn change, which is slow, so use early survival at months 1 to 3 as a leading signal.
- Do not pool results: each experiment must keep its own control.
Trade-offs and pitfalls
- A cheaper paywall may buy conversions that churn faster; the combined LTV figure catches this.
- Churn cuts that rely on friction can raise complaints and chargebacks.
- Referral incentives cost money; subtract it before claiming an LTV gain.
Unlock Full Question Bank
Get access to all 46 Growth, Activation and Retention interview questions and detailed answers.
Sign in to ContinueJoin thousands of developers preparing for their dream job.