Product teams often use experiments to determine how a new feature or change will affect metrics that they want to influence. However, they must also ensure that this release does not significantly harm other company-wide metrics. Airbnb released a highly configurable Guardrails system that flags experiments that might negatively impact a key company-wide metric prior to launch. The system uses three guardrails: an impact guardrail (ensures that the global average treatment effect is not more negative than a preset threshold), a power guardrail (ensures that the Impact Guardrail has sufficient false-positive escalation rate and power), and the stat sig negative guardrail (ensures that experiments with a statistically significant negative impact on top metrics are always escalated).