Performance Max A/B testing in Google Ads lets you compare two sets of assets inside the same asset group, so you can test headlines, descriptions, images or video without rebuilding the campaign. The feature is in beta and still rolling out. Brands Win, a Singapore-based performance marketing agency for consumer brands, explains how ecommerce brands can set up and read the test.

Checked against Google Ads Help and Search Engine Land, 7 October 2026. The feature is still rolling out as a beta, so check Campaigns > Experiments in your own Google Ads account.

What Performance Max asset A/B testing is

It is a Google Ads experiment that compares a control set of assets (A) with a treatment set (B) inside one Performance Max asset group, while unselected “common” assets keep serving to 100% of traffic.

That is Performance Max asset testing inside one group, not a new campaign type.

Google’s help page About Performance Max optimization experiments: A/B testing assets (Beta) describes the three buckets: control group, treatment group and common assets. Search Engine Land reported on 2 October 2026 that advertisers can now test headlines, descriptions, images and videos within an asset group instead of swapping creative and guessing from before-and-after results.

That is different from choosing campaign type. If you are still deciding between Performance Max and Standard Shopping from your Merchant Center feed, use our Performance Max vs Shopping guide. This post assumes you already run Performance Max and want a cleaner read on creative changes inside it.

Why this matters for Singapore ecommerce brands now

Creative is one of the few levers you still control once Performance Max automation is bidding and assembling ads.

Beauty texture demos, supplement pack shots, lifestyle scenes, F&B plating and festive SKUs for 11.11 or 12.12 are typical tests. Brands refresh seasonal creative in Q4 and need a clearer signal before locking festive assets. Without a controlled split, learning-period noise and other campaign changes make a simple asset swap hard to judge.

Treat availability carefully. Google’s business announcement says the features are “currently rolling out”. Singapore is not called out as excluded in the Help pages checked on 7 October 2026, but access is account-based. Open Campaigns > Experiments in your Google Ads account before you brief a full creative batch around this flow.

How the experiment works (control, treatment, common assets)

Google splits assets in the chosen asset group into three categories.

Per Google Ads Help:

• Control group (assets A): existing assets you keep as the reference.
• Treatment group (assets B): other existing assets or new uploads you want to compare against control.
• Common assets: anything you do not put in A or B. These keep serving to 100% of traffic alongside whichever arm a user is in.

You set the traffic split yourself. The start date defaults to the next day. The Experiment Guidance System suggests an end date for statistical significance; Google recommends running for at least 4 to 6 weeks.

Locked assets matter. Once the experiment starts, you cannot edit, add or remove assets in that asset group until it ends. The editing pages become view-only so the test stays valid. Prepare every treatment headline, description, image and video before you schedule.

Other limits from the same Help page: one asset group per experiment; control and treatment assets both count toward asset-group limits; newly uploaded treatment assets go through policy review and disapproved assets will not serve.

Sibling types on About Performance Max optimization experiments: Asset testing cover adding assets to a feed-only campaign, or video on versus video off. Use those when that is the question; this post focuses on A/B testing two asset sets inside one group.

Set-up steps in Google Ads

Google’s Help page lists these steps for an “Any assets” experiment:

1. Go to Experiments within the Campaigns menu.
2. Select the plus button under the All experiments tab.
3. Under “What do you want to test?”, select Assets.
4. Select Assets provided by you, then Performance Max, then Continue.
5. Select Any assets under experiment type.
6. Choose the Performance Max campaign and the asset group to test.
7. In the Control arm card, review existing assets; in the Treatment arm card, select or upload the assets to compare.
8. Set the traffic split, name the experiment, confirm the schedule (start defaults to the next day; end date comes from the Experiment Guidance System), then select Schedule.

When the test finishes, Apply can add treatment assets and, by default, keep control assets (uncheck to remove them). End discards treatment assets and restores the original state.

Common blockers from the same Help page: shared budget; portfolio bidding or Smart Bidding Exploration; overlapping experiment dates; unsupported campaign features; an asset group with more than 15 assets that Google describes as “maxed”. Fix those before you schedule.

You own the Google Ads and Merchant Center accounts and pay Google directly. An agency works inside your account. That is how our Google Shopping and Performance Max work sits alongside Google Ads and SEM.

What to test (ideas for beauty, wellness, lifestyle and F&B)

Run one clear hypothesis per experiment.

Examples that fit Singapore consumer brands:

• Creator-style product demo video versus studio pack-shot images.
• Lifestyle scene versus product on a plain background.
• Festive or sale angle versus evergreen messaging.
• Short product-in-use clip versus stills, useful when Performance Max serves into YouTube video ads inventory.

Produce variants through performance creative for ads. On Meta we use a different test method; see how we test creatives on Meta if that is your other channel.

Claims still apply. HSA’s cosmetic products overview says advertisements cannot claim to modify a physiological process, such as reversing hair loss, or to prevent or treat a disease or medical conditions. ASAS’s Singapore Code of Advertising Practice expects factual claims to be capable of substantiation. A winning asset that makes a non-compliant claim is still non-compliant. This is not legal advice. Skip therapeutic, weight-loss or before-and-after promises in your treatment creative.

Reading results without fooling yourself

Wait for the recommended duration. Do not stop early because one arm “looks ahead” after a few days.

Judge against the campaign’s agreed performance target: conversions or conversion value at a similar cost per acquisition or return on ad spend, not vanity clicks. Cross-check Google Ads figures with store or booking data; see why platform numbers and sales reports differ.

After Apply, decide whether to keep both asset sets or remove the loser. Google’s business page also mentions a second success metric “soon”; treat that as not live until it appears in your account.

A 90-day way to use asset experiments

Ninety days is long enough to run more than one hypothesis without thrashing the account.

Our Growth System 90 programme covers paid media on Meta, TikTok and Google, marketplace channels such as Shopee and TikTok Shop, influencer marketing and content strategy over 90 days (Discover, Launch, Optimise, Review), against one performance target:

• Discover: list live Performance Max asset groups and the creative hypotheses worth testing first.
• Launch: schedule the first asset A/B with a seasonal or creator-style treatment, assets prepared before the lock.
• Optimise: Apply winners, then start the next hypothesis on another asset group.
• Review: keep or drop assets and brief the next creative batch.

FAQs about Performance Max A/B testing