Braze Content Optimizer: What It Tests and When to Use It

Braze Content Optimizer is a Canvas step that mixes variants of message components, such as subject lines and calls to action, and shifts sends toward the combination that performs best for your audience. It suits journeys with a steady daily flow of entries. Give it clearly distinct variants and an event that matches your goal, then leave it alone while it learns.

Key Takeaways

• Content Optimizer finds one winning combination for your audience; it does not personalise per user.
• Use it in Canvases with steady daily entries, because if everyone enters at once it behaves like a static A/B test.
• Write clearly distinct variants, and include your business-as-usual content as one of them.
• For email, optimize for clicks where they fit the goal.
• Every step stays in Learning for at least seven days and no more than 15, and publishing changes resets it.
• Deactivate variants about an hour before an offer ends, because delayed sends can still receive them.

What does Braze Content Optimizer test?

Content Optimizer finds one winning combination for your audience by generating variants and mixing their elements. Braze is explicit that it does not personalise per user. It runs as a step inside a Canvas, and it's available for email, push, SMS, MMS and RCS.

For email it can generate up to 625 combinations per step, and for push up to 125. Each user receives one message per entry into the step, and a re-entry starts again with no memory of earlier variants.

Each component needs its Liquid tag placed in the base message. The Canvas can't launch until every selected component is in place, which is a useful guard against a half-built test going out.

When should you use Content Optimizer instead of another test?

Braze offers several ways to compare content, and they answer different questions. Choosing the wrong one wastes the traffic you test with.

We'd put one question first. If you already know the two or three messages you want to compare, a standard test is simpler to read. Content Optimizer earns its place when you're unsure which subject line, header and call to action belong together, because it tests the combinations for you.

Our guide to A/B testing email campaigns covers the standard route in detail.

Which Canvases suit Content Optimizer?

Choose the journey before you write a single variant. Braze advises using Content Optimizer where users enter gradually and regularly, such as recurring, always-on, event-triggered or API-triggered Canvases with consistent daily volume.

The reason is mechanical. If every user enters the step at once, Content Optimizer has no time to learn from early results, and the step behaves more like a static A/B test. For spiky journeys, Braze suggests entry controls to spread entries over several days.

A welcome series or a post-purchase journey usually fits well, because new users arrive every day. A one-off product launch email usually doesn't, unless you deliberately stagger it.

How should you design the variants?

Make each variant a different idea, not a reworded copy. Braze's own advice for a step that can't find a winner is to make variants more clearly distinct in tone, structure or content.

Braze also notes that testing at least 10 total combinations generally gives better results. If your volume is low, averaging under approximately 3,000 sends per day, it suggests slightly fewer variants, since too many combinations can slow learning.

Include your business-as-usual content as one variant for each component. That gives you a direct comparison against what you would have sent anyway, and it's the number a stakeholder will ask about first.

Pick the optimization event to match the goal. For email, Braze notes that steps optimizing for clicks tend to outperform steps optimizing for opens, and its click count excludes bot clicks and recognised unsubscribe clicks. Content Optimizer counts qualifying unique events registered within seven days of each send.

What happens while the step is learning?

Every step stays in the Learning state for at least seven days and no more than 15. After that, Braze moves it to Optimizing, or to Action Recommended if no clear winner emerged.

Publishing changes resets the optimizer, and Braze advises against updating variants during Learning. Treat the Learning window as frozen, and collect ideas for the next round instead of editing the live one.

Retiring a variant needs lead time. A user whose send is delayed by quiet hours, Intelligent Timing, rate limiting or prioritisation can still receive a variant you deactivated after they were assigned it. Braze recommends deactivating about an hour before an offer ends rather than at the exact moment.

In what order should you set up a Content Optimizer step?

Work in this order, because each choice limits the ones after it.

1. Pick a Canvas with steady daily entries, or add entry controls to smooth a spiky one.
2. Decide the goal and choose the matching optimization event, clicks for most email.
3. Choose the components to test, and write distinct variants for each, including business-as-usual.
4. Check the combination count against your daily volume.
5. Place every component's Liquid tag in the base message.
6. For a first run, use Experiment Paths so only part of the audience enters the Content Optimizer branch.
7. Launch, and leave the step untouched until it leaves Learning.

The Experiment Paths step is worth the extra setup. It gives you a holdout that shows whether the optimized branch beat the journey you already had.

Which failures should you watch for?

Name the common failures so a reviewer can spot them in a setup review. A launch-day blast is a step placed in a send where everyone enters at once, so it never gets to optimize.

A Learning-state edit is a change published during the first week, which resets the optimizer and restarts the clock. Near-twin variants are options so similar that no winner can emerge, which tends to end in Action Recommended.

A ghost variant is one you deactivated that still reaches users whose sends were delayed. Plan deactivations with lead time, and expect a tail of the old variant after you switch it off.

What should you know about AI-generated variants and plan limits?

Content Optimizer uses OpenAI only when you ask it to generate variant suggestions. Braze states it does not use OpenAI to choose which variant each user receives or to allocate traffic, and you can type variants yourself instead.

If your organisation reviews which tools receive message content, raise the suggestion feature with your security team before first use. Typed variants avoid the question entirely.

Plan tier also matters. On one tier, a workspace can run one active Content Optimizer step at a time, and your Braze account manager can help if you need more. Check your own plan before you design several tests in parallel.

Where does this connect to the rest of the programme?

The content you test is shaped by the patterns in high-converting email campaigns and by the personalisation in hyper-personalisation in Braze. Reading the results alongside your other sends is covered in email campaign reporting.

For messages that run across several channels, multichannel engagement campaigns covers the journey design around a step like this.

As a Braze implementation partner, CustomerIK recommends choosing the journey before the copy. A Content Optimizer step in the right Canvas, left alone through Learning, does far more than a clever variant in the wrong one.

Frequently Asked Questions

1. Does Content Optimizer personalise content for each user?

It doesn't. It finds one winning combination for the audience. Braze points to Decisioning Studio Go or Pro for per-recipient personalisation.

2. How long does Content Optimizer take to find a winner?

Every step stays in Learning for at least seven days and no more than 15. Braze then moves it to Optimizing, or to Action Recommended if no clear winner emerged.

3. Can I edit variants while the step is learning?

You can, but publishing changes resets the optimizer. Braze advises against updating variants during the Learning state.

4. Should email steps optimize for opens or clicks?

Braze notes that email steps optimizing for clicks tend to outperform those optimizing for opens. Use clicks where they fit your goal.

5. Does Content Optimizer send my content to OpenAI?

Only when you request AI-generated variant suggestions. Variant selection and traffic allocation use Braze's own algorithm, and you can enter variants manually.

Sources

• Braze, Content Optimizer. Retrieved 16 Oct 2026.
• Braze, Content Optimizer step. Retrieved 16 Oct 2026.
• Braze, What's new in Braze, checked for Content Optimizer changes. Retrieved 16 Oct 2026.

CustomerIK is a Braze implementation partner working across onboarding, technical integration, marketing operations and customer data management.

If you're planning your first Content Optimizer test and want the Canvas checked before launch, let's talk.