Your Performance Max ROAS Is Taking Credit It Didn't Earn
Your Performance Max campaign shows the highest ROAS in the account, and you have started treating that number as proof the channel works. The mistake is trusting a single fused return for a campaign that is quietly taking credit for sales that were already coming to you. This article proves that your PMax ROAS is inflated by branded searches and remarketing impressions served to people who had already decided to buy, and it shows you how to separate the revenue PMax actually created from the revenue it merely photographed.
That distinction has a name. It is called incrementality, and it is the only number that pays your bills. The reported figure on your dashboard is not a lie. It is a blend, and the blend hides whether Performance Max is bringing you new customers or just standing in the doorway taking the receipt for ones you already had.
Your Best-Performing Campaign Is Four Campaigns Wearing One Coat
Performance Max is not a campaign type the way a Search campaign or a Shopping campaign is. It is a wrapper. When you launch it, Google takes your budget and spends it across nearly every surface it owns: Search results, Shopping listings, YouTube, Gmail, Display banners, Maps, Discover. One budget, one bidding goal, one reported return.
That sounds efficient. It is efficient, for Google. The trouble for you is that those surfaces do completely different jobs and convert at completely different rates, and the system reports them as a single result. A branded search, where someone types your company name because they already decided to buy, converts at a very high rate and costs almost nothing. A cold Display impression to someone scrolling a news site converts almost never. Average those two together and the cheap, already-decided branded clicks pull the whole figure up. The return you see is propped up by the traffic that needed Performance Max the least.
Adalysis, which builds auditing software for Google Ads, studied this across a large body of accounts covering roughly 3,300 Performance Max campaigns. The sample size is the part worth standing behind as a hard number. The pattern they surfaced is the one practitioners keep running into: Performance Max leans heavily on the cheapest, highest-intent traffic it can find, and a meaningful slice of the conversions it claims come from queries that were already going to convert. The campaign is good at finding people who were already raising their hands. That is the first thing to internalize. Your reported ROAS is an average of four or five very different traffic types, and the average is dominated by the easiest one. This is the same dynamic that lets a quarter of every dollar leak out of a Google Ads account without anyone noticing, because the headline metric looks healthy the whole time.
Owners Get Fooled Because Google Designed the Report to Fool Them
The mechanism is specific, and falling for it is not your fault. When someone searches your brand name, they are already yours in every meaningful sense. They saw your truck, got a referral, remembered an old purchase. The search engine's job at that point is simply to hand them to you. That click is some of the cheapest, highest-converting traffic anywhere in digital marketing.
By default, Performance Max is allowed to compete for those branded searches, and it does. It serves your ad against your own name, the click converts at a sky-high rate because the person was already sold, and PMax records the conversion as its own. On the report, that looks like the campaign generating a fantastic return. In reality it intercepted a sale that was already on its way and stamped its name on the receipt. Five Nine Strategy has written about exactly this attribution problem, the way PMax positions itself to absorb credit for demand that already existed. The campaign is not creating that demand. It is taking the credit for it.
That hits your wallet and not just your pride. Branded search is the single cheapest way to convert a customer. If you were running a dedicated brand-search campaign before, paying a few cents a click for people typing your name, PMax may now be eating that same traffic at a different bid while denying you the clean visibility you used to have. You are paying to capture customers you already had, and the report calls that capture a triumph. The same flattering logic shows up when Google's automated bidding quietly reallocates your money, a risk worth understanding before you hand over more budget, which is why smart bidding can put your budget at risk even when the dashboard looks fine.
The Data Shows Relocation Dressed Up as Growth
If you wanted to verify this yourself, your instinct would be to open the search terms report and see which queries PMax converted. Good instinct. Performance Max has historically given advertisers very limited visibility into the actual search terms and placements it spends on. Optmyzr, a respected platform in the paid-search world, has documented this reporting opacity at length. You get categories, groupings, some search-term themes. What you often do not get is the clean, query-level detail that would let you say with confidence how much of your spend went to branded terms versus genuinely new demand. In a normal Search campaign you can see the exact words people typed. In your best-performing campaign, that window is frosted over.
Go Fish Digital published a study on a furniture brand, Joybird, where Performance Max was cannibalizing branded search. The pattern they untangled is the textbook version of everything above. PMax was absorbing branded queries, claiming the conversions, and inflating its own apparent performance while the dedicated brand campaign told a more complicated story. Once they separated brand from non-brand and stopped letting PMax feast on the cheap branded traffic, the true picture of what the campaign contributed to new demand came into focus.
Performance Max was not always adding revenue. In places it was relocating it, moving a sale that would have closed through cheap branded search into the PMax column where it looked like incremental growth.
That is the trap stated plainly, and it is worth sitting with. If you judge Performance Max by its own self-reported ROAS, you will conclude it is your best channel and pour more budget into it. If you judge it by whether your total revenue actually grew when you added it, you may find a very different answer. Those two judgments can point in opposite directions, and only one of them pays your bills. Incrementality is the revenue that exists because of a campaign that would not have existed without it. Not the conversions a campaign reports, but the conversions that genuinely would not have happened otherwise. A campaign can report a 600 percent return and deliver almost no incremental lift if most of those conversions were people already going to buy. No dashboard hands you that number. You have to go get it.
How to Find Your Real Lift This Week
You do not need to rip Performance Max out, and you do not need a data science team. You need to stop trusting its self-assessment. Work through this order, starting now.
Start with brand exclusions, because this is the single highest-impact move and you can do it this week. Ask Google to apply brand exclusions to your Performance Max campaign so it stops competing for searches of your own company name. Many accounts can request that a brand list be excluded so PMax keeps its hands off branded queries. Once branded traffic is fenced off, the campaign has to earn its return on genuinely new demand, and you will often watch that flattering number come down to something closer to the truth. That drop is not a loss. It is the mask coming off.
Next, measure the share of conversions that were branded. Before and after the exclusion, look at how much of the reported performance was riding on people searching your name. Use whatever brand-search visibility you can pull together, including your dedicated brand campaign if you run one. You are answering one question: how much of this great result was just you capturing customers you already had. While you are there, watch your brand-search campaign for cannibalization. If you run a separate, cheap branded-search campaign, watch what happens to its impression share and volume when PMax is active. If PMax is quietly eating that traffic, the brand campaign will starve while PMax fattens. That is the Joybird pattern, and catching it is just a matter of looking at the two together instead of one at a time.
Then run the honest test: a holdout or a pause. Turn Performance Max off for a defined window, two to four weeks depending on your volume, and watch total revenue, not just the PMax line. If total sales barely move when the campaign goes dark, much of what it reported was never incremental. If total sales clearly drop, the campaign is genuinely contributing and you have your answer. For accounts with enough volume, a geographic holdout, running PMax in some regions and not others, gives you the same read with less disruption. Finally, re-decide the budget on incrementality, not reported ROAS. You may keep the campaign exactly where it is. You may shift spend back into the cheaper branded search it was cannibalizing. You may scale it because the test proved it really is bringing new customers. The point is that you are now deciding with a true number instead of a flattering one.
Measure Lift, Not the Number That Looks Good in the Report
The thing to track from here forward is total revenue against total spend through your holdout windows, not the self-reported return sitting at the top of the PMax tab. Run a pause test or a geographic holdout on a schedule, quarterly is reasonable for most accounts, and compare the business with the campaign on versus off. That comparison is the closest thing to truth you can get without a measurement vendor.
The vanity traps are easy to name once you know them. Reported ROAS is the loudest one, because it counts every branded and remarketing conversion as a PMax win even when those buyers were already yours. Conversion volume is the second, because it climbs as the campaign absorbs more existing demand and tells you nothing about whether the demand is new. Even a rising overall account ROAS can mislead you if PMax is simply relocating cheap conversions into a column that flatters itself. Watch the aggregate instead. Total leads, total revenue, and what those move to when you flip the campaign off. The aggregate cannot be gamed by attribution quirks, and a healthy total through a holdout is worth more than any single green number on a dashboard.
Frequently Asked Questions
Will excluding branded searches tank my Performance Max ROAS?
Almost certainly the reported number will drop, and that is the point rather than a problem. The branded traffic was inflating the figure because those buyers were already yours. When you fence it off, what remains is closer to the return the campaign earns on genuinely new demand. A lower, more honest number you can act on beats a high one you cannot trust.
How long should I pause Performance Max to test whether it is incremental?
Two to four weeks is a reasonable window for most accounts, longer if your sales cycle is slow or your conversion volume is low. Watch total revenue across the whole account during the pause, not just the PMax line. If overall sales barely move, much of what the campaign reported was never incremental. If they clearly fall, the campaign is contributing real lift.
Can I see the exact search terms Performance Max is spending on?
Not fully. Google gives you search-term themes, categories, and groupings rather than the clean, query-level detail you get in a normal Search campaign. Tools like Optmyzr have documented this opacity, and it is why a holdout test matters more than trying to audit the queries directly. When you cannot inspect the inputs, you measure the output by turning the campaign off and watching what happens.
The campaign that looks like your best performer might really be your best performer. It might also be a very confident mirror, showing you the customers you already had and calling them growth. The only way to tell the difference is to separate brand from non-brand, turn the thing off for a couple of weeks, and watch what your business actually does. So before you approve the next budget increase someone hands you with that green ROAS attached, ask whether you have ever once seen what your revenue does when the campaign is switched off.
