Skip to content
  • platform
  • tutorials
  • auto tune setup

Tutorial: Set Up Auto-Tune

Auto-Tune monitors your live experiments and surfaces recommendations to improve performance — without taking control away from you. This tutorial walks through enabling it, understanding notifications, and managing the optimization loop.

Before you start

Section titled “Before you start”
  • You need a template with an active experiment that’s accumulating traffic.
  • Your template should use Internal Experimentation with Persistent User Bucketing enabled.
  • Auto-Tune evaluates variants after they reach 1,000 sends.

Bucketing setting

How Auto-Tune works

Section titled “How Auto-Tune works”

Auto-Tune runs in the background on live experiments. It:

  1. Monitors variant performance against your key metric
  2. Identifies statistically significant winners and losers
  3. Sends you a Slack notification when action is needed
  4. Suggests what to archive and what new variants to approve

Nothing changes automatically — you’re always in control of what goes live.

Step 1: Get notified via Slack

Section titled “Step 1: Get notified via Slack”

When an experiment is eligible for Auto-Tune, you’ll receive a Slack notification summarizing:

  • Which experiment is running
  • The key metric being evaluated
  • Whether variants are over- or under-performing

Click the notification to go directly to the experiment in Compact Mode in Studio.

Auto-Tune Slack notification

Step 2: Review performance in Compact Mode

Section titled “Step 2: Review performance in Compact Mode”

Compact Mode gives you a summarized view of the experiment:

  • The control and all active variants
  • Which variants are outperforming (green)
  • Which variants are underperforming (red)
  • Why certain variants are flagged for Auto-Tune actions

This view lets you assess the situation at a glance before making any changes.

Step 3: Archive underperformers

Section titled “Step 3: Archive underperformers”

Auto-Tune flags underperforming variants with a red “Archive” button.

When you archive a variant:

  • It’s removed from active traffic
  • If other approved variants exist, they continue running
  • If no other variants are available, the control is shown instead

Underperforming variants flagged

Step 4: Approve new variant suggestions

Section titled “Step 4: Approve new variant suggestions”

To keep the optimization loop going, Auto-Tune generates draft variants based on patterns from your past tests.

These suggested variants:

  • Are labeled Auto-generated
  • Appear with a green “Approve” button
  • Are never published automatically — nothing goes live without your review

Review the suggestions and approve the ones you want to test.

Auto-Tune suggested variants

Step 5: Let it run

Section titled “Step 5: Let it run”

After you’ve archived losers and approved new variants, Auto-Tune continues working in the background. Over time:

  • Your content improves automatically as performance data accumulates
  • Testing maintains momentum without manual oversight
  • If performance declines due to fatigue, seasonality, or temporal effects, Auto-Tune detects and responds
  • Every message stays evergreen and continuously optimized

Day-to-day management

Section titled “Day-to-day management”

Once Auto-Tune is running, your ongoing workflow looks like this:

When you get a notificationWhat to do
Underperformer flaggedReview and archive if you agree
New variants suggestedReview, edit if needed, and approve
A variant is winning consistentlyConsider shipping it as the new control
No notifications for a whileThings are running smoothly — no action needed

What’s next

Section titled “What’s next”