Find leading indicators of long-term churn
turning a lagging metric into early signals.
cohort renewers vs churners, compare first-30-day engagement depth and breadth, validate correlations and check causality.
jumping to a model before defining the cohort and target.
What's really being asked
The interviewer wants to see if you can convert a slow lagging indicator into early leading signals that the team can act on while it still matters. It also probes your discipline around cohorts, leakage, and causality.
The full answer
Start by framing the target: label users who have reached the 12-month mark as renewed or churned, ensuring everyone in the analysis had a fair chance to renew. Build a feature set strictly from the first 30 days so signals are usable early. Good features include login frequency, breadth of distinct features touched, time-to-first-value, completion of an onboarding milestone, and counts of a core habit-forming action. Compare distributions between renewers and churners using effect sizes, not just averages, and segment by plan and acquisition source to avoid Simpson-style reversals. Then validate that promising signals hold on a fresh cohort before trusting them.
The mistakes people make
Immediately fitting a black-box model without exploring the data, leaking features that occur after day 30, treating correlation as proof of cause, or ignoring survivorship and tenure so churned and active users are compared unfairly.
What usually comes next
How would you confirm a signal is causal rather than correlated, perhaps via an onboarding experiment? How would you operationalize the indicator into a dashboard or alert? How do you handle class imbalance and seasonality across cohorts?
A concrete example
Suppose churners average two distinct features used in week one while renewers average five. You hypothesize that early feature breadth drives stickiness. You confirm the gap on a new cohort, then run an onboarding experiment nudging new users toward a second and third feature. If 12-month renewal improves for the treated group, you have a validated leading indicator the team can monitor weekly instead of waiting a year.
Interview question
Why must features for predicting 12-month churn be drawn only from a user's first 30 days?
- a.Because data older than 30 days is always unreliable
- b.So the signal is available early enough to act on and avoids leaking future outcomesCorrect
- c.To guarantee the model achieves higher accuracy
- d.To reduce the storage cost of the feature table
Why? this is the answer
Restricting to the first 30 days keeps signals actionable early and prevents leakage of post-period behavior. Storage cost is irrelevant, and limiting the window does not by itself raise accuracy.
Just read this? Test yourself on what you have been reading.
Read the original → amplitude.com
- #churn
- #cohort-analysis
- #leading-indicators
- #product-analytics
- #retention
Put your scrolling time to good use
Learn one idea, try a quiz and save useful cards for revision. Tezvyn makes it easy to learn and stay current in your tech field, a few minutes at a time.
The iPhone app is on the way
We are building it. Until it lands, nothing here is held back from you: every interview card, your saved cards, streaks and the job board all work in Safari, plus hundreds of free practice quizzes of thirty questions each. Sign in and it all carries over to the app the day it arrives.
Want it as an icon? Tap Share at the bottom of Safari, then Add to Home Screen. It opens full screen and the cards you have read stay available offline.
We are hiring for this. Every open role lists the topics its interview covers, so you can prepare for the real thing rather than guessing.
See open roles