Guardrail Metrics
Guardrail metrics are secondary metrics you monitor alongside your primary metric to catch unintended side effects.
Why guardrails matter
Your experiment might boost revenue (your primary metric) but tank visits or spike churn. Guardrails ensure you don't celebrate a win on one number while another is quietly collapsing.
Example: You raise prices at test locations. Revenue goes up 8%—great. But visits drop 15% and churned members spike. Guardrails catch this before you roll out a net-negative change.
Adding guardrails
On the test detail page (Overview tab), you'll see a Guardrail Metrics section with a dropdown to add metrics. The dropdown excludes your primary metric (no point guardrailing what you're already measuring).
Available guardrail metrics include any metric your organization tracks: revenue, transactions, visits, enrollments, churned members, and more.
You can add or remove guardrails at any time, even after analysis has run. When you change guardrails, ProofPod re-runs the analysis to include them.
Reading guardrail results
Each guardrail metric gets its own results card showing:
- The estimated effect on that metric
- Whether the change is concerning (warning or critical)
- Confidence intervals
Impact on recommendations
Guardrail results directly affect recommendations:
- Critical guardrail failure can trigger a Kill recommendation, even if the primary metric looks good
- Warning-level guardrail issues appear in the recommendation checklist but don't override the primary metric's signal
- Clean guardrails (no concerning changes) contribute to a Scale recommendation's confidence
Start with 2–3 guardrails that represent the most important "do no harm" metrics for your business. Too many guardrails can make it hard to ever reach a Scale recommendation.