What is Performance Max (PMax) and when to use it

Performance Max (PMax) is a Google Ads campaign type that buys on a single budget across every inventory at once - Search, Shopping, YouTube, Display, Discover, Gmail and Maps - and decides on its own where, to whom and with which creative to appear. You do not pick a network, keywords or bids: you give it a conversion goal, a feed or a set of assets (headlines, descriptions, images, video) and some audience signals, and the system distributes.

That is the deal: you hand over targeting and bidding in exchange for coverage and for Google optimising against your goal with all of its signal. It works well when what you feed it is good; when it is not, PMax does not fail loudly, it fails quietly, spending on the cheapest traffic available: your own brand.

What Performance Max is exactly, and how it differs from Shopping or Search

A standard Shopping campaign buys one inventory with one feed. A Search campaign buys one inventory against keywords you choose. PMax buys all of them at the same time, with one budget and one bidding strategy, and reallocates between them as it sees fit from one day to the next.

The practical consequences of that are the ones worth remembering:

• There are no keywords. There are audience signals, which are hints rather than limits, exactly like the Advantage+ audience on Meta.
• There are no per-network bids. Only Maximise conversions or Maximise conversion value, with an optional target CPA or target ROAS.
• Assets, not ads. You upload headlines, descriptions, images, logos and video into asset groups, and Google assembles the combinations. If you do not supply video, it generates one automatically from your other assets.
• It takes priority. When the system expects to perform better, PMax outranks standard Shopping and broad or phrase match Search in your own account. It can drain traffic from campaigns you already had under control.

That last point is the one most people find out about too late: PMax is not just another campaign added to the account, it changes how the campaigns already there deliver.

Problem number one: brand cannibalisation

This is the single most repeated finding when we audit an account where PMax is running "really well". Campaign ROAS is spectacular, the business is not growing, and opening the search term insights shows why: a huge share of the spend has gone to queries containing the brand name, which would have converted anyway without paying, or which your brand campaign was already covering at a far lower cost.

It happens because the system looks for the cheapest route to an attributed conversion, and that route is always demand that already exists.

How it is fixed: exclude brand queries from PMax (with a campaign-level or account-level negative list, depending on how your brand campaign is set up) and leave brand in its own Search campaign, where CPC is low and control is total. After the exclusion, panel ROAS drops: that is not a deterioration, it is the real number you had hidden before. What settles whether you won or lost is not campaign ROAS but business MER and the volume of new customers.

What data PMax gives you, what it withholds, and how to get the rest

The opacity is real, but smaller than its reputation. It is worth knowing exactly where the ceiling is so you do not waste hours hunting for a report that does not exist:

• Search terms: there is no classic report. There are search term insights, which group queries into themes. They are read campaign by campaign with no daily breakdown: useful for spotting brand and noise, not for precise measurement.
• Placements: the PMax placement report gives you impressions and little else. You know which sites and YouTube channels you appeared on, but not what you spent or what converted on each. Good for excluding obvious rubbish, not for optimising on performance.
• Performance by channel: not broken out in the standard interface. You can approximate it with asset reports and listing group detail, but do not expect a clean split of spend across Search, YouTube and Display.
• Performance by product: this one is genuinely good. For ecommerce, the product-level report is the best real reading of the campaign you are going to get.

What is worth setting up from day one: your own UTM tags on the final URL (do not rely on auto-tagging alone) so you can read the traffic in your analytics and in your backend. The truth about PMax is almost never in the Google panel: it is in crossing its spend against actual orders.

When it is worth using

It fits well when:

• You are ecommerce with a healthy product feed. The feed is the single biggest lever: titles, attributes, images and availability decide more than anything inside the campaign.
• There is enough conversion volume for the system to learn. With a trickle of weekly conversions it never exits learning and results swing week to week.
• Conversion tracking is solid and, ideally, sending back values that reflect margin rather than revenue. PMax optimises towards the number you give it, so giving it revenue means it chases revenue, not profit.
• You want coverage of demand you are not reaching with keyword campaigns, and you have brand already fenced off.

Tread carefully when:

• B2B lead generation. If the optimised event is the form submission, PMax will find you the cheapest forms on the internet. Without offline conversion imports carrying qualification back, it works against you.
• You need channel-level reading to make decisions outside Google. It is not there.
• Small budgets split across several PMax campaigns. Consolidation is the format's advantage; slicing it cancels it out.
• Strict brand control. Auto-generated assets and auto-created video assemble combinations nobody signed off. They can be disabled, but it has to be done deliberately.

How to launch it without giving budget away

A launch protocol that avoids the false conclusions of the first few weeks:

• Exclude brand before spending a single euro, not after the first good report. Otherwise your baseline is contaminated from day one.
• Fix the feed first. Every hour spent on titles and attributes pays back more than any in-campaign setting.
• Turn off auto-generated assets and final URL expansion on the first run, so you know what is actually being shown and where traffic is landing.
• One campaign, segmented by asset group, rather than four campaigns competing against each other in the same auction.
• Write down the baseline before you start: MER, total orders, new customer share and brand Search cost. Those are the four numbers that judge the test.
• Let it run two to four weeks untouched. Every target change restarts learning, and a tROAS set too high at launch simply stops delivery.

Frequently asked questions