Select and prioritize metrics with guardrails
Company: Meta
Role: Data Scientist
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Technical Screen
A new Groups Stories feature aims to increase meaningful engagement without harming the broader ecosystem. Build a metrics framework and pick a minimal gating set: 1) define three top-line goal metrics (success metrics) and why they are normalized (per DAU, per viewer, or per poster); 2) list feature-only metrics (treatment-only CTR/creation rate) for sizing if launched; 3) ecosystem metrics to catch second-order effects (sessions, retention, revenue); 4) explicit cannibalization reads (impact on Feed, Marketplace, Pages); 5) guardrails (latency, crash, report rate, cancellations) and sanity checks; 6) specify a multiple-testing control method (e.g., Holm–Bonferroni vs BH) and when you’d prefer each; 7) show how you’d back-of-the-envelope the LTV and ARPDAU impact from a +1% DAU movement. Prioritize the three metrics that must be green to roll out to 100%, and explain why others are secondary.
Quick Answer: This question evaluates a candidate's ability to design a robust metrics framework for product experiments, covering selection and normalization of success metrics, feature-only sizing and ecosystem metrics, explicit cannibalization reads, guardrails, multiple-testing control, and back-of-the-envelope LTV/ARPDAU impact estimation.