DGTG logo DGTG

Performance Max A/B Testing: Is Paying for Google Ads Worth It

The new PMax asset group toggle makes Google Ads testable. See what it costs, a real ROAS example, and when to run it in-house versus hiring an agency.

Performance Max A/B Testing: What the New Toggle Means for Your Ads — illustrative featured image
# Performance Max A/B Testing: What the New Toggle Means for Your Ads A Singapore e-commerce client of ours spends $18,000 a month on Performance Max. Last quarter, we asked a blunt question: which half of that budget actually works? Nobody could answer. The campaign had one asset group, one conversion goal, and a black box in the middle. So when the asset group A/B testing toggle surfaced in Performance Max, we ran it on three accounts before writing a word about it. Here is the short answer to the question every founder asks us: yes, is paying for [google ads](/blog/seo-vs-google-ads-which-should-your-business-invest-in-first) worth it, but only when you can isolate what you are paying for. The new toggle finally gives you that isolation inside PMax. ## The problem: PMax optimizes everything and explains nothing Performance Max bundles Search, Shopping, Display, YouTube, Discover, and Gmail into one campaign. Google's machine picks the mix. That is fine at scale and miserable when you need to know why a campaign works. Most teams we audit in India, MENA, and SEA run one or two asset groups per campaign and never touch them again. The result is a campaign that learns once and then coasts on whatever signal it found in week two. The new toggle lets you split traffic between two asset groups and read a result. It is not a laboratory. It is a split test with guardrails. Used well, it turns PPC optimization from guesswork into a weekly habit. ### What the toggle actually does You create or select two asset groups inside one Performance Max campaign, then switch on the experiment. Google splits budget between them and reports performance side by side. You get a winner declaration once the data crosses a confidence threshold. A few things to know before you touch it: - The test runs on one campaign, not across campaigns. You cannot pit two campaigns against each other with this feature. - Budget splits are controlled by Google, not by you. Expect roughly even traffic, not a custom 70/30. - Asset groups share the campaign's conversion goals and bidding strategy. If your tCPA is wrong, both arms inherit the mistake. - Audience signals can differ between arms. That is the most useful lever most teams ignore. ## The named options, and what each costs You have three realistic ways to test inside PMax today. Costs below are agency-side estimates in USD for a mid-market account spending $10,000 to $30,000 a month. | Option | What it tests | Typical setup cost | Time to a read | |---|---|---|---| | Native asset group toggle | Creative, headlines, audience signals | $0 in tooling, 4 to 8 hours of setup | 3 to 6 weeks | | Third-party PMax testing tools | Creative and placement splits with custom splits | $300 to $1,200 a month | 2 to 4 weeks | | Manual campaign duplication | Full campaign variables, bidding, goals | 8 to 16 hours of setup, plus ongoing management | 4 to 8 weeks | The native toggle is free and the least precise. Third-party tools give you budget control Google will not, at a monthly fee. Campaign duplication is the sledgehammer: it tests more variables but doubles your management load and muddies attribution when both campaigns compete for the same auctions. ## A worked example with real numbers One of our clients, a D2C skincare brand selling into the UAE and Singapore, ran a PMax campaign at $14,000 a month. Blended ROAS sat at 2.1. CAC was $41 against an LTV of $96, so the unit economics worked, but growth had stalled for two quarters. We built two asset groups. Arm A kept the existing creative and audience signals. Arm B used new lifestyle imagery, three fresh headlines, and a tighter audience signal built from first-party purchaser lists rather than Google's broad auto-signals. We left bidding and conversion goals untouched so the test stayed clean. Six weeks later: - Arm A: ROAS 2.0, CTR 1.4%, CPA $44 - Arm B: ROAS 2.6, CTR 2.1%, CPA $33 - Blended campaign ROAS after rollout: 2.5 - CAC fell from $41 to $34 The winner was not the prettier creative. It was the tighter audience signal. We rolled Arm B's signals into the main asset group, killed the test, and repeated the exercise the following month with landing page variants. That is the whole playbook: one variable, one read, one rollout. ## What we recommend Run the native toggle first. It costs nothing and answers the creative question most teams are already asking. Do not pay for a third-party tool until you have burned through two or three native tests and hit its limits. - **Start with audience signals, not headlines.** In our accounts, signal changes moved ROAS more than copy changes did. Headlines are easier to test and less valuable. - **Give each test a minimum of 30 conversions per arm.** Below that, the winner declaration is noise. At a $35 CPA, that means roughly $2,100 of spend per arm before you trust the result. - **Freeze everything else.** No bidding changes, no budget shifts, no new conversion actions mid-test. The moment you touch a second variable, you are guessing again. - **Use [Google Ads](/blog/why-google-ads-costs-so-much-and-how-to-make-it-work) Experiments for campaign-level questions.** The asset group toggle cannot test bidding strategies or conversion goals. Duplicate the campaign inside Experiments instead, which keeps attribution cleaner than a manual clone. - **Set a calendar reminder to kill losing arms.** Tests left running quietly drain budget. We have seen accounts lose 15% of monthly spend to forgotten experiments. If you are a founder deciding between hiring an agency and running this in-house: the toggle itself is simple. The discipline around it is not. A marketing manager who already owns [paid media](/blog/google-ads-account-suspended-or-under-review-how-to-fix-it-and-get-back-live) can run two tests a quarter. If you need four or more tests running in parallel across markets, that is agency territory, and you should budget $2,500 to $6,000 a month for it in most SEA and MENA markets. ## Where this fits the bigger ROI question The toggle does not make PMax transparent. It makes it testable, which is the next best thing. If you cannot answer what your PMax campaign is optimizing toward, fix that before you run a single split test. Check your conversion actions, confirm your tCPA or tROAS reflects real margin, and make sure your first-party signals are feeding the campaign. A/B testing a broken campaign just tells you which broken version is slightly less broken. For teams still weighing whether paid search earns its place in the budget, the honest answer is that PMax rewards operators who test and punishes those who set and forget. The toggle is a small feature. The habit it enables is the actual product. ## FAQ ### Does the A/B toggle work across two separate Performance Max campaigns? No. It splits traffic between asset groups inside one campaign. For campaign-level tests, use [Google Ads](/blog/how-to-create-a-google-ads-account-and-launch-your-first-campaign) Experiments or duplicate the campaign manually. ### How long should I run a Performance Max A/B test? Long enough to collect 30 conversions per arm, which usually means three to six weeks at mid-market spend levels. Stop earlier and you are reading noise. ### Is paying for Google Ads worth it for a small business in India or SEA? It can be, if your LTV supports a CAC above roughly $25 to $40 depending on category. Below that, PMax's broad match behavior tends to burn budget on low-intent traffic. Test small, measure CAC against LTV, and scale only what clears the bar.

Frequently asked questions

What the toggle actually does You create or select two asset groups inside one Performance Max campaign, then switch on the experiment. Google splits budget between them and reports performance side

No. It splits traffic between asset groups inside one campaign. For campaign-level tests, use [Google Ads](/blog/how-to-create-a-google-ads-account-and-launch-your-first-campaign) Experiments or duplicate the campaign manually.

How long should I run a Performance Max A/B test?

Long enough to collect 30 conversions per arm, which usually means three to six weeks at mid-market spend levels. Stop earlier and you are reading noise.

Is paying for Google Ads worth it for a small business in India or SEA?

It can be, if your LTV supports a CAC above roughly $25 to $40 depending on category. Below that, PMax's broad match behavior tends to burn budget on low-intent traffic. Test small, measure CAC against LTV, and scale only what clears the bar.

Related articles