How do you define churn for a subscription service?

WHAT THIS TESTS: The interviewer wants to see if you can turn a vague business request into a rigorous, queryable data definition. Churn sounds simple until you have to write the SQL or build the model. The question probes your ability to identify event boundaries, handle edge cases like grace periods, and separate customer intent from system noise. It also reveals whether you think about the downstream impact of your definition on retention metrics, revenue forecasting, and experiment analysis.

A GOOD ANSWER COVERS: First, distinguish between voluntary and involuntary churn. Voluntary means the user actively cancels or opts out. Involuntary means the account lapses due to a failed payment, expired credit card, or system error without an explicit intent to leave. Second, pick a precise timestamp. Options include the moment of cancellation request, the subscription end date, the last successful login, or the close of a dunning grace period. Third, define the inaction threshold. For a subscription service, churn is not just a button click; it is the state of no longer having active access and not renewing within a defined window. Fourth, address reactivations and winbacks. A user who cancels on day one and reactivates on day three might not count as churned in a monthly cohort if your window allows for it. Fifth, align the definition with the business model. A monthly SaaS product might use end of billing period, while an annual contract might use contract anniversary plus renewal window.

COMMON WRONG ANSWERS: Treating every cancellation event as churn immediately. This ignores users on annual plans who cancel three months early but still consume and may revert. Using payment failure as churn without a dunning retry period, which inflates the number and makes recovery campaigns look futile. Defining churn by a single inaction like no login for thirty days without considering whether the subscription is still billing and the user is receiving value passively. Failing to mention that churn must be cohort-based or time-bound; a user who left in January cannot be counted in the February denominator without a clear rule.

LIKELY FOLLOW-UPS: How would you validate that your churn definition actually predicts revenue loss? What is the difference between churn rate and retention rate in a high-growth scenario with many new subscribers? How would you handle annual versus monthly subscribers in the same denominator? If a user downgrades from a premium to a free tier, is that churn? How do you account for seasonality or cohort effects when presenting churn to executives?

ONE CONCRETE EXAMPLE: Imagine a monthly streaming service that bills on the first of the month. A user clicks cancel on the fifteenth. The subscription remains active until the end of the month. A naive definition flags churn on the fifteenth. A senior definition flags churn at the end of the billing period, or even thirty days later if the user does not reactivate, because the revenue is still recognized and the user might reverse the decision during the remaining access window. If the user pays annually and cancels in month four, churn should arguably be recorded at the annual anniversary, not the cancellation click, unless the contract allows immediate termination.