Safe Rollouts

ProSafe Rollout is available on Pro and Enterprise plans.

Releasing a new feature always carries risk. Even small changes can introduce regressions. Safe Rollouts reduce this risk by releasing to a subset of users while monitoring key metrics for issues.

Overview

A Safe Rollout is a Ramp Schedule with guardrail monitoring turned on. It releases a feature to a subset of users and automatically monitors guardrail metrics for regressions. Guardrails can be any metrics that matter to your team, such as error rates, latency, or conversions.

The rollout runs as a short-term A/B test:

• The Control receives the existing value.
• The Rollout receives the new value.

GrowthBook analyzes the rollout's impact on your selected guardrail metrics and provides status updates: safe to ship, issues detected, or consider rolling back.

How to Add a Safe Rollout

1. Add a Targeting Rule with Monitoring

When adding a new rule to your feature, select Targeting rule, then choose Monitored Ramp-up as the release plan. The Show me shortcut on the rule type screen configures a fully monitored ramp for you.

note

Safe Rollouts are now configured as a monitored Ramp Schedule rather than a separate rule type. The monitoring, analysis, and rollback behavior described below is unchanged.

2. Configure the Safe Rollout

• Rollout value: the value users in the Rollout group receive. It is compared against the Control (the rule's existing value).
• Sample by: the attribute used to assign users to the Control and Rollout groups (for example, id or organization_id).

3. Define Metrics and Monitoring

• Data Source and Assignment Table: Confirm the data source and assignment table used for exposure tracking.
• Guardrail Metrics: Select one or more metrics to monitor for regressions. GrowthBook automatically rolls back and disables the rule if any of them regress significantly, so there is no separate auto-rollback toggle.
• Signal Metrics: Optional metrics that pause the rollout at the current step if they regress, without rolling it back. It resumes when they recover.

Under Advanced Settings, you can choose how GrowthBook responds to other issues, such as a sample ratio mismatch, no traffic, or multiple exposures: hold the ramp, roll back, or warn only.

4. Ramp Up Schedule

A Safe Rollout ramps traffic up through the steps of its underlying Ramp Schedule. By default, a monitored ramp steps the rollout through:

1% → 5% → 10% → 25% → 50%

Each monitored step splits its traffic 50/50 between the Rollout and the Control, so monitored steps top out at 50%. The early steps each hold for a short slice of the Duration, then the rollout holds at 50% for the rest of the window before releasing to 100% when the schedule completes. You can change the pace with Duration, or edit the steps directly in Advanced View.

5. Monitor the Rollout

Once published, GrowthBook begins monitoring the guardrail metrics. The rule shows a status badge that updates automatically as results come in:

Status	Meaning
X days left	The rollout is running. Monitoring continues until the duration completes.
Unhealthy	A health check failed, such as a sample ratio mismatch. Traffic may be imbalanced; check your implementation.
Guardrails Failing	A guardrail metric is regressing. The rule shows a Revert Now button so you can roll back.
Ready to ship	The duration finished with no failing guardrails. Safe to release to 100%.
No Data	No traffic detected. Check your setup.
Reverted	The rollout was rolled back to the control value.
Released	The rollout was shipped to all users.

info

Guardrails are analyzed for failure using frequentist sequential testing, allowing you to roll back as soon as statistical significance is reached without fear of false positives. Your safe rollout is automatically monitored for implementation errors with sample ratio mismatch and multiple exposures checks.

6. Take Action

At the end of the monitoring period (or sooner if issues arise), you can:

• Release the feature to 100% of users (if not already there).
• Revert the feature to the control value.
• Continue to monitor.

If a guardrail metric regresses, GrowthBook automatically rolls back and disables the rule to protect your users.

Understanding the Time Series Graph

The safe rollout monitoring interface includes a time series graph showing how your guardrail metrics perform over time. The graph displays two key values:

Metric Boundary

The Metric Boundary is the statistical boundary for whether a safe rollout is failing. When it crosses zero, we have enough statistical certainty that the safe rollout is harming this metric. Technically, it is the [lower/upper] bound of the absolute change confidence interval between the baseline and the safe rollout groups.

Threshold

The threshold for when a metric is considered failing. It's always set to zero — as soon as there's statistical certainty that a metric is being harmed at all (even by very small amounts), the safe rollout is marked as failing.

FAQs

How Does the Traffic Ramp Up Work?

A Safe Rollout ramps traffic through the steps of its Ramp Schedule. By default it steps through 1%, 5%, 10%, 25%, and 50% before releasing to 100%, keeping the initial blast radius small and widening as monitoring stays healthy. You set the pace with the rollout's Duration and can customize the steps in Advanced View.

What Happens if a Guardrail Metric Degrades?

If a guardrail metric crosses the significance threshold for a regression, GrowthBook rolls back and disables the rule automatically to prevent further harm, and the rollout's status shows the failing guardrail. For the other automated checks (sample ratio mismatch, no traffic, and multiple exposures), you choose the response under Advanced Settings: hold the ramp, roll back, or warn only.

What Should I Use as My Guardrail Metrics?

Choose metrics that represent key system health or business outcomes. This might include error rates, latency, or conversion rates—whatever signals that the change is working as intended and not causing harm.

Choosing too many guardrail metrics increases the chance of false positives. Aim for a focused set of critical metrics.

How Is This Different From an Experiment?

Safe Rollouts use the same analysis engine as GrowthBook experiments but are designed for operational decision-making, not learning. The primary goal is to ensure a safe release, not to measure long-term impact.

What Happens if There Isn't Enough Data?

Safe Rollouts are built to help you ship confidently and reduce the chance of negative regressions. If any guardrail fails, roll back the change.

If results are still inconclusive after the configured duration, ship — there's no clear evidence that the feature is harmful.

Safe Rollouts bias towards action. If you're more uncertain about a feature and want to learn about its impact, run a regular Experiment instead.