Why one PMax for the whole store can hold growth back

A lot of the stores that come to us have one big Google Ads campaign for the entire range. It has run for years, nobody touches it, and in the report it looks fine. At Elektro Sláma it carried 74% of the budget, at Vše pro pejska 81–96%, and at Letsport the inherited campaign still carries 77% of spend today.

That campaign is almost always Performance Max. The problem isn’t that it’s a bad campaign. The problem is that one campaign has one target and one budget — and the algorithm spends them where selling is easiest, not where you make money. This article is about when to split PMax, when to keep it whole, and how to do it without the account falling apart for a month.

What PMax is and what changed since 2025

PMax (Performance Max) is a Google Ads campaign type that buys ads across every Google placement — Search, Shopping, YouTube, Gmail, Discover and the Display network — from one budget and towards one target. You supply the target, the budget, the products and the assets. The algorithm decides who sees the ad, where, and at what price.

For a store, it pulls products from the XML feed through Merchant Center, plus the copy, images and videos you upload. It replaced Smart Shopping for e-commerce and is now the most common campaign in the accounts we take over.

For a long time it was called a black box. During 2025 Google added a channel performance report, a full search terms report, campaign-level negative keywords, age exclusions and device targeting. You can see far more of PMax today than you used to.

One thing hasn’t changed: the target and the budget are still set for the whole campaign, not for individual products or asset groups. Everything else in this article follows from that.

How a Performance Max campaign decides where the money goes

The algorithm gets one number — a target ROAS or a target CPA — and looks for auctions where it expects to hit it. It can’t tell whether a thousand crowns of revenue came from a jacket with a 45% margin or a bowl with a 15% margin. To the algorithm, a thousand is a thousand.

Second: it spends where it has the most data. A product that sells often gets more impressions because the algorithm is surer of it. Bestsellers soak up the budget and the rest of the range never gets a turn, even when it would earn more. Move the sliders to see what one shared target does to two categories.

So one campaign with one target makes money on part of the range and quietly loses it on another. Added up for the account it looks like a decent average, which is why you can’t read the campaign’s overall ROAS on its own. What decides is each category’s break-even ROAS: 100 divided by its margin in percent.

The same in ACoS: a 25% target is comfortable for a category with a 45% margin and loss-making for one with a 15% margin. One campaign can only have one of them.

Google says “consolidate”. Why that isn’t always right

Google’s advice has been the same for years: fewer campaigns, more data in each. The algorithm learns from conversions, and if you split them across five campaigns, each gets a fifth. That’s true, and for small accounts it holds completely. At MyDeko we pulled six campaigns into one PMax, search and brand — and cost per purchase fell 72%.

But consolidation fixes the quantity of data, not its quality. It doesn’t fix a campaign that mixes products with different margins, different seasons or different customers. The algorithm then has plenty of data — it just optimises for the wrong number.

Consolidate until campaigns have enough conversions. Split as soon as one average target starts hurting part of the range.

The rule we build account structures by

5 signs it’s time to split PMax

Check them against your own account. You need the campaigns view, the products report and a margin table.

1. One campaign carries more than 70% of the budget and nobody can say which categories in it make money and which only spend.
2. Categories differ in margin by more than 10 percentage points. Their break-even ROAS then differs by hundreds of percent, and one target fits neither.
3. Part of the range has its own season — winter clothing, back to school, Christmas. Averaged over the year it gets lost, and the budget arrives late.
4. Spend sticks to a handful of products. The products report shows most of the budget on bestsellers and cheap items, while a big part of the range gets no impressions at all.
5. A category needs its own creative. Copy about dog-jacket sizes doesn’t fit bowls and toys — and one asset group can’t be written to fit everything.

If two or more apply, splitting usually pays. If only the fifth applies, a new asset group is enough and you don’t need to split the campaign.

How to write copy and images for that asset group from your customers’ own words is covered in How to turn one customer review into five ad concepts.

When not to split PMax

Splitting costs data. Every new campaign learns from scratch and is more expensive until it has. These are our thresholds:

• Under 30 conversions in 30 days for the whole campaign, we don’t split. Tracking, feed and volume come first.
• The part you split off should have at least 30 conversions in 30 days on its own. With ten, it will learn for months and make the whole account more expensive in the meantime.
• Conversions must be real purchases with a value. At Papírnictví VojTech, 81% of “conversions” turned out to be add-to-carts. Splitting would only have copied that error into more campaigns.
• Don’t split right before the season peaks. The dog-clothing campaign at Vše pro pejska had five weeks of learning on real purchases behind it before Black Friday. That was what made the difference.
• A small account with similar margins across the range should stay in one PMax with search and brand. More campaigns add nothing but noise.

Calculator: split PMax or keep one?

Enter the numbers for the last 30 days. The calculator works out how many conversions the split-off part would get, how far apart your categories’ break-even lines are, and recommends a level of splitting.

The calculator uses a threshold of 30 conversions in 30 days and a margin gap of 10 percentage points. Those are our rules from practice, not Google’s. You don’t enter seasonality — if your category has a season, read “split gradually” as “split in time”.

How to split Performance Max without needlessly resetting learning

Splitting isn’t an overnight rebuild. It’s a series of small steps after which the algorithm still has something to learn from.

1. Split by margin, season or customer, not by the site menu. A category from the store tree only makes sense if it matches a margin or seasonal group.
2. Tag products in the feed with custom labels. Margin, season, bestseller. Without labels you can’t cut the range sensibly and new campaigns can’t pick what to sell.
3. Split off one category, not five at once. At Vše pro pejska clothing went first, beds four weeks later, hoodies in February.
4. Exclude it from the original campaign the same day. Remove it in the original PMax’s listing group, or the two campaigns compete for the same products.
5. Give it its own budget and target ROAS based on that category’s break-even ROAS plus a profit margin. Start slightly looser so it has something to learn from.
6. Add in steps. Budget up by 10% at most every 14 days, target ROAS by 15% at most and no more than once a week.
7. Judge after four to six weeks and judge the whole: store revenue and margin, not just the new campaign’s ROAS.
8. Don’t switch the original campaign off. With a smaller budget it spends on what it sells best and catches products that belong to no category.

Campaign, asset group or custom label

Not every split needs a new campaign. It depends on what exactly you need to separate.

A common path: first a label and a margin report, then an asset group with its own creative, and a separate campaign only once the category reaches the conversion threshold.

Data from our accounts: before and after

Four accounts we manage. In three we split PMax, in one we did the opposite and consolidated. Charts and the month-by-month story are in the case studies.

At Vše pro pejska the smaller budget also cut the original campaign’s own cost: from 21–25% in summer 2025 to 12–17% in 2026. A smaller budget forced it to spend on the products it sells best. At MyDeko the same principle worked the other way — an account with low purchase volume needed its data in one campaign, not spread over six.

What we got wrong

Three of these splits didn’t go smoothly. They belong here, because the rules in this article are built on them.

The new campaign competed with the old one

At Elektro Sláma ACoS dropped below target for the first time in April, to 19.1%. In May it went back to 30.9%. The new category campaigns were still learning and competed with the original PMax for the same products. The account only turned in June, once they had matured. Today we exclude the category from the original campaign on the same day the new one launches.

A category without its own demand

The bulbs campaign for the same client turned out to be the weakest in the account. It keeps a small budget and waits for data. Splitting off a category that doesn’t have enough purchases of its own doesn’t pay — which is why the calculator above counts the split-off part’s conversions before anything else.

A seasonal campaign launched late

We launched hoodies at Vše pro pejska on 11 February, at the end of winter. They finished at 26% ACoS. Clothing, launched on 15 October, had five weeks of learning before Black Friday and closed the season at 16.3% — the cheapest of all campaigns outside brand.

And one split is still an investment, not a saving: the climbing PMax at Letsport has run since 5 June on a fifth of spend at 56.2% ACoS. We keep it because it opens a category the account never had. It doesn’t contribute to profit yet, and we know it.

Checklist before you split

Ten questions to go through before you touch the account structure. Download the checklist and the decision table on the left — no email, no form.

Want to know how much of your budget one campaign carries and what in it actually earns? It’s the first thing we look at when we take over an account as part of Google Ads management. Or send us an enquiry below.

FAQ

Jiří Kopejska

Co-founder, Marketing ASAP

Co-founded Marketing ASAP. Takes over accounts where one campaign has carried most of the budget for years, and takes them apart by where the store actually makes money.