Primary vs. Guardrail Metrics in Experiments

This question tests your understanding of holistic product impact and risk management, not just optimizing a single number. A strong answer defines the primary metric as the main success criterion and guardrail metrics as critical health checks you don't want to see regress (e.g., latency, retention). A feature ships only if the primary improves *without* a statistically significant negative impact on guardrails. A common red flag is treating guardrail regressions as acceptable trade-offs rather than hard blockers for a ship decision.