When Performance Max is worth using, how to stop it claiming sales you would have had anyway, and the settings and tests that keep your Google Ads budget under control.
Performance Max (PMax) is Google's most automated campaign type. A single campaign can show your ads across Search and Shopping results, YouTube, Display, Discover, Gmail and Maps, while Google's systems decide where each euro goes. Used well, it finds customers your other campaigns miss; used as a default, it can quietly absorb traffic you were already winning more cheaply and report it back as its own success.
Using PMax without losing control comes down to two things. First, give it what it needs to learn: a reliable conversion signal, enough volume and, for retailers, a good product feed. Second, set its boundaries before it spends: keep brand searches in your brand campaign, protect your best Search keywords, limit where it sends visitors, and judge it by what it adds to your total results rather than by its own report.
This matters because PMax can look excellent in its own dashboard while your total sales barely move. The sections below cover when it suits your business, how to stop it taking credit for sales you would have had anyway, how to build asset groups that give it good material, and how to test what it really adds.
You give PMax a goal (conversions or conversion value, with or without a target), a budget and one or more asset groups. An asset group is a set of headlines, descriptions, images, videos and logos, plus audience signals that describe your likely customers; retailers also connect their product feed from Google Merchant Center. Google then decides, auction by auction, which channel, placement and combination of assets to use.
There are no keywords and no fixed placements, and audience signals are hints rather than limits: PMax will show ads to people outside them if it predicts a conversion. That flexibility is the point of the product, and also its risk. PMax pursues the conversions you count wherever they are cheapest to get, and that isn't always where new customers come from.
Give a new PMax campaign four to six weeks before judging it, and avoid frequent changes to budget, targets or asset groups during that time. Like any automated bidding, it needs a stable period to learn, and constant editing keeps it permanently in that phase.
PMax tends to work best when a few conditions hold. Use them as a checklist before you launch, and be honest about any you can't tick:
PMax is usually a weaker fit for business-to-business (B2B) companies with long sales cycles and a few dozen enquiries a month. Optimising towards raw form submissions across Display and Discover can bring in a lot of cheap, low-quality leads. If you use it for lead generation, import qualified leads from your customer relationship management (CRM) system, bid towards those rather than every submission, and check lead quality every week.
Budget matters too. A modest budget spread across many asset groups gives each one too little data, so start with fewer, broader groups and split them only when each has enough conversions to stand on its own. Retailers who mainly want new buyers can also use the customer acquisition setting, which tells PMax to bid more for new customers or to pursue only them, provided it can tell new and existing customers apart, usually from a customer list you upload.
The most common problem is cannibalisation: PMax picking up searches your other campaigns were already winning. Google's rule is that when a search exactly matches a keyword in one of your Search campaigns, the Search campaign takes priority. When it doesn't, the two compete and whichever has the stronger ad rank serves, which leaves room for PMax to take searches you used to win through phrase and broad match.
Brand searches are the biggest leak. People searching for your company name would mostly have found you anyway, and they convert well, so when PMax serves on them its reported results look far better than its real contribution. Use brand exclusions to keep those searches in your brand Search campaign, and negative keywords to keep PMax away from other searches you don't want it to buy.
Treat Performance Max as a contract with Google's algorithm, and read the terms before you sign.
A worked example shows why this matters. Say PMax reports €40,000 in revenue from €8,000 of spend, a return on ad spend (ROAS) of 500%. Over the same weeks, your total revenue from Google Ads rose by only €6,000 compared with the period before launch, while your brand campaign's revenue fell by €20,000: most of what PMax claims was moved, not added.
The figures are illustrative; the check is not. Compare total results across all campaigns before and after any PMax launch, not just the campaign's own numbers. Treat Performance Max as a contract with Google's algorithm, and read the terms before you sign.
Asset groups decide what PMax can show and, to a large extent, where it can show it. A group with three headlines and one image gives the system little to work with, and if you don't add a video, Google may generate one from your images, leaving you with little say over how your brand looks on YouTube.
Build one asset group per distinct theme, such as a product category, a service line or a clear customer segment, each with its own landing page. Fill every asset type Google asks for: the full set of short and long headlines and descriptions, horizontal, square and portrait images, your logo, and at least one real video, even a simple one filmed in-house. Retailers can also limit each asset group to the products that fit its theme, and leave out products they don't want to advertise at all.
Use audience signals to point the system at your best customers: an uploaded customer list, collected and used in line with GDPR (the EU's data protection rules), your website visitors, and custom segments built from searches your buyers make. Search themes let you add the searches you know matter. All of these are starting points, not fences.
Then check two settings that decide where traffic lands and what your ads say. By default, PMax may send people to pages other than the one you set as the landing page, a setting called final URL expansion, and may write its own text variations. For lead generation, paid clicks landing on blog posts or job pages are a common leak, so either switch URL expansion off or exclude the pages that shouldn't receive ad traffic.
PMax reports less than a Search campaign, so use everything Google does show: the search terms or search categories behind your results, how individual assets perform and, where available, how results split across channels. Check where your ads appeared on YouTube and partner sites, and use account-level placement exclusions, which apply to PMax, to remove placements that don't fit your brand.
Reports can't answer the question that matters most: what would have happened without PMax? For that you need a test. Google Ads offers experiments for PMax, and a simpler alternative works in larger markets: exclude a set of comparable regions from PMax for four to six weeks and compare total conversions, from every channel, with the regions where it kept running.
If total results barely differ, PMax is mostly reshuffling credit, and its budget may work harder elsewhere. If they clearly differ, you have evidence for scaling it, which is worth more than any figure in its own dashboard.