Growth Engineer Interview Questions

Option B is strongest: it structures the answer into four explicit steps, explains why weak activation definitions matter (misleading metrics), provides a concrete activation event example (teammate invitation accepted), names the specific properties to track, defines cohorted vs. raw activation rate with an explanation of why raw % can be misleading, and names both analytics tools and SQL-level analysis. Growth vocabulary: Activation event — the specific user action that signals "aha moment" / perceived value. Activation funnel — stages from awareness to activated. Cohorted activation rate — % of users from a given signup period who activate within a defined window. Aggregation bias — distortion in aggregate metrics caused by changes in the composition of the population. Options C and D are accurate but lack the cohort bias explanation and the event taxonomy detail.

Option B is strongest: it names all six components explicitly with the rationale for each, explains why session-level randomisation is wrong (same user sees both variants), explains the early stopping problem (peeking inflation of false positive rate), introduces confidence intervals vs. p-values as analysis tools, and adds novelty effect as a post-analysis concern. A/B testing vocabulary: Minimum detectable effect (MDE) — the smallest change worth detecting; used to calculate required sample size. Statistical power — probability of detecting a true effect when it exists (typically 80%). p-value — probability the observed result occurred by chance given no true effect. Peeking — checking results mid-experiment and stopping early when significant; inflates false positives. Novelty effect — temporary lift from users trying something new, not a true sustained improvement. CUPED — Controlled experiment Using Pre-Experiment Data; reduces variance. Options C and D are accurate but lack the randomisation unit explanation and the peeking problem justification.

Option B is strongest: it names three explicit design dimensions, explains why irrelevant notifications are the #1 churn cause, provides the specific feedback loop mechanism (suppress after 3 dismissals), and introduces the crucial insight that open rate is a misleading metric — downstream retention is the true north star. The feedback loop design shows systems thinking beyond "send and measure." Notification vocabulary: Frequency cap — maximum number of notifications per user in a time window. Send-time optimisation — personalising delivery to when the user historically engages. Opt-out rate — % of users unsubscribing from a notification type; a leading indicator of churn. Re-engagement notification — a notification sent to a lapsed user to bring them back. Notification fatigue — user burnout from excessive or irrelevant notifications. Options C and D are accurate but lack the feedback loop design and the retention-as-north-star framing.

Option B is strongest: it names three concrete confounding explanations for a single correlation, distinguishes randomised vs. quasi-experimental approaches with specific examples of each, introduces holdout groups as a separate tool, and names three danger patterns (survivorship bias, novelty effect, Goodhart's Law) with definitions. The Goodhart's Law reference shows economic sophistication. Causality vocabulary: Selection bias — users who adopt a feature may already be predisposed to the outcome. Confounding variable — a third factor that influences both the treatment and outcome. Difference-in-differences — comparing the change in outcome before/after treatment for treated vs. control groups. Holdout group — a permanently withheld control group for long-term causal measurement. Goodhart's Law — "when a measure becomes a target, it ceases to be a good measure." Options C and D are accurate but lack the Goodhart's Law insight and the three confounding explanations.

Option B is strongest: it grounds the story in a specific context (B2B SaaS, 40% drop-off), provides the quantified diagnosis (60% drop-off at teammate invite step), explains the causal reasoning behind the solution (users don't have something worth sharing yet), gives concrete engineering implementation details (Lambda, SQS, DB query), and reports both the primary result and guardrail metric outcomes — showing mature experimentation practice. Growth engineering vocabulary: Activation rate — % of users reaching the defined activation event. Onboarding funnel — step-by-step path from signup to activation. Feature flag — toggleable code path allowing controlled rollout. Contextual nudge — a prompt triggered by user behaviour, not scheduled timing. Guardrail metric — a metric that must not degrade when optimising a primary metric. pp (percentage points) — absolute change in a percentage metric. Options C and D are accurate summaries but lack the technical implementation depth (Lambda/SQS) and the guardrail metric reporting.