Google Ads · QUICK HIT
PMax asset edits as A/B tests
Google's help page for Performance Max A/B asset tests (Beta) now describes a toggle in the asset editor that turns your edits into a test instead of saving them straight to the asset group.
Jerry Quin3 min read1 source
Illustration: The Keywordless.
The short answer
What changed
Performance Max A/B testing assets is a Beta experiment that compares 2 sets of assets inside the same asset group. Google's help page now has a section called "Set up an experiment from the asset group builder": when you edit assets, a toggle at the bottom of the page, "Save asset group changes as an A/B test," turns the edit into a test.
Who has it
Performance Max advertisers. The page still labels the feature Beta and doesn't give a date or say whether every account has the toggle.
What I'd do this week
My call, not Google's: before you flip the toggle on a big edit, decide whether you can live with that asset group being locked until the test ends. Steps below.
How it works
The test runs inside one asset group. A selection of your existing assets is the control group ("assets A"), a selection of assets you want to compare against it is the treatment group ("assets B"), and anything you don't put in either group keeps serving "to 100% of your campaign's traffic."
The page also covers a separate route, through the Experiments page. This section starts from the asset group itself. Open the campaign, go to Asset groups, and select the pencil icon under "Assets." On the "Edit assets" page you change headlines, descriptions, images, sitelinks and other assets as usual.
At the bottom of that page is a "Save and test asset group changes" panel. Turn on the toggle and the button changes "from "Save" to "Create test"." Before you create it, you can check the control arm, treatment arm, experiment name, treatment traffic split, duration and success metrics. The page's last step names another option, Save with experiment, which it says "retains the prior ad version of traffic to measure its effect to the original ad settings."
The default to know about
One option comes pre-checked: "Apply these changes to the original asset group if the treatment arm wins." The page says it "will be turned ON by default" and "only takes effect when the experiment ends and if results show that the treatment arm performed better than the control arm."
The page doesn't say whether you still confirm anything at that point. Its section on acting on a test describes selecting Apply and confirming in a dialog.
The limits that come with it
- Locked assets: once the experiment starts, "you can't edit, add, or remove any assets in the asset group that's being tested until the experiment ends."
- One asset group per test: "You can only test assets within a single asset group per experiment."
- Asset limits: control and treatment assets both count toward the asset group's limits.
- Review: new treatment assets go through normal policy review, and disapproved ones won't serve in the test.
For tests set up from the Experiments page, the same page recommends running "for a minimum of 4 to 6 weeks." The new section doesn't give its own duration advice.
What I'd do this week
None of this comes from Google; it's how I'd handle it.
Use the toggle for edits you'd want to measure anyway, like a new set of images or a rewrite of your headlines. Skip it for fixes that have to go live, like a wrong price in a description, because the asset group locks once the test starts.
Look at the pre-checked apply option every time. If a client or a brand team has to sign off on creative, I'd uncheck it so applying a winner stays a choice you make by hand.
Pick the asset group carefully. With one asset group per test and a lock on that group, I'd start with one that has steady traffic and no promotions planned during the test.
If you don't see the panel at the bottom of the editor, don't plan around it yet.
Jerry Quin
Jerry Quin is the pen name of a paid media practitioner who has managed Google Ads accounts at an agency for 10 years. He writes about Google Ads, Microsoft Ads, ChatGPT Ads, and the rest of paid media and digital marketing, from what's rolling out to what he'd actually do about it.