When Google rolled Performance Max out as a full Shopping replacement in 2022, my first reaction — and the reaction of most operators I respect — was fury. The pitch was "trust the algorithm." The fine print was "we're merging six ad formats into one opaque bucket, removing most of your reporting, and please don't ask which SKUs are working." Three and a half years later, I've run PMax on roughly twenty accounts across e-commerce, lead gen, and B2B. My position has softened. But I want to tell you exactly where, and where it hasn't.
Let me start with the concession. Performance Max is, in 2026, the single best-performing campaign type in Google Ads for a specific kind of account. If you have a clean product feed, a high-conversion-volume business (roughly 30+ conversions per week per asset group), and conversion tracking that actually reflects business value — PMax will outperform any hand-crafted Shopping + Search + Display configuration I could build you. That was not true in 2022. It became cautiously true in 2024. It is now confidently true in 2026, assuming all of the above conditions are met.
The problem is that most accounts don't meet those conditions
And this is where Google's marketing and the on-the-ground reality diverge. Google's official line — echoed by every Google rep I've ever been assigned — is that PMax works for "every business." That is not true, and it has not been true at any point. Here's what PMax actually does badly, and still does badly in 2026:
It cannibalizes brand search. If you don't explicitly exclude your own brand terms from PMax, it will quietly serve on them, attribute those conversions to itself, and make the campaign look dramatically better than it actually is. I've seen accounts where 40% of PMax "conversions" were brand traffic that would have converted anyway. Setting up brand exclusion lists is now the very first thing I do on any PMax account. Google made this easier in a late-2024 update, but it's still opt-in. Which tells you everything about their incentives.
It flattens your customer acquisition into a single averaged-out bid curve. PMax decides — on its own — how much budget goes to Shopping vs. Display vs. YouTube vs. Search. If your margins vary wildly across those surfaces (and they do), you're trusting Google's black box to optimize against your ROAS input, which it interprets as aggregate. A customer acquired on YouTube via a remarketing pixel fire is not the same as a customer acquired on a non-brand Search query. But in PMax, they look identical to the algorithm.
It punishes low-volume accounts mercilessly. Below roughly 30 weekly conversions per asset group, PMax's machine learning has nothing to learn from. It will spend, and it will generate data. But the data is noise, and the spending is not guided by anything meaningful. I see small DTC brands — spending $8K/month, getting 10-15 weekly conversions — being sold "PMax as the primary channel" by their agencies. It's malpractice. At that volume, a well-structured manual Search campaign with tight Shopping feeds will outperform PMax by 30-50% every single time.
The rules I now treat as non-negotiable
After three and a half years, here's the structural playbook I use. Every item below has been validated across multiple client accounts. Every item exists because I watched Google's defaults cost someone money.
One. Brand gets its own campaign type, every time. Standard Search. Exact match on your brand terms. Aggressive bidding on top positions. This campaign is run entirely separately from PMax, and PMax has your brand terms in its exclusion list. Full stop. The number of accounts where this single change produces a 15-25% efficiency improvement is almost comical.
Two. New customer acquisition is a signal, not a setting you can trust. Google added the "new customer acquisition" bid modifier in 2023 with great fanfare. In practice, it's useful — directionally. It is not a precise dial. I still run at least two separate asset groups segmented by acquisition intent (prospecting-heavy vs. retention-heavy) rather than trusting a single PMax campaign to get the balance right.
Three. Asset groups are not ad groups. This is the mistake I see most often. Operators come from legacy Google Ads thinking and treat asset groups like ad groups — small, hyper-segmented, one-per-product-category. Don't. Asset groups in PMax need to be large enough to clear the conversion-volume learning threshold (again, ~30 conversions/week minimum). I typically run 3-6 asset groups per PMax campaign, organized around customer intent or product margin tier — never around product category or SKU type.
Four. Search themes are still a trap. Google introduced Search themes as a way to give PMax more signal about the queries you want to show up on. On paper: helpful. In practice: they routinely broaden your targeting past anything reasonable. I use search themes sparingly — and only in asset groups where I've set up an aggressive negative keyword list at the account level.
Five. You need to manually audit placements. PMax will happily spend meaningful budget on YouTube content that no self-respecting brand would want to advertise next to. The placement report has improved since 2023 (where it was essentially useless) but you still have to check it monthly and exclude aggressively. I've found kids' content, AI-generated spam channels, and the occasional genuinely offensive video in the placement reports of clients whose previous agencies never bothered to look.
What Google has actually gotten right
I want to be fair here. Google has made real improvements to PMax since 2022, and credit is due:
- Asset-level reporting finally arrived in late 2024. You can now see which individual creative assets are carrying performance and which are dragging it. This is what Meta has had forever, and it was embarrassing how long PMax took to get there. But it's here now.
- Account-level negative keywords for PMax came in 2023 and are now table stakes. If you're not using them, you're effectively letting PMax show on search queries you would never manually bid on.
- Customer lifetime value inputs have gotten genuinely useful. If you can reliably feed PMax your 90-day CLV (or a reasonable proxy), the bidding engine can optimize toward longer-term value instead of first-purchase ROAS. This is a real unlock for subscription businesses and anyone with meaningful repeat purchase rates.
- The brand exclusion list, as I mentioned above. Still opt-in, but at least it exists now.
My recommendation, by account type
If you're running an e-commerce brand with $50K+/month in spend, a clean product feed, 50+ weekly conversions, and properly configured conversion tracking — use PMax as your primary Shopping + prospecting channel, alongside a brand Search campaign. This is the use case it was actually built for, and it's the use case where it performs exceptionally well.
If you're running a lead generation business — services, SaaS, local, B2B — your situation is more nuanced. PMax for lead gen got dramatically better in 2024 when Google finally let it optimize toward offline conversions reliably. I now use it for about half my lead gen clients, but never as the whole stack. Standard Search handles bottom-of-funnel intent queries; PMax handles everything else. The split is maybe 40/60 or 50/50 depending on the business.
If you're spending under $20K/month, I'd tell you to avoid PMax entirely. Run standard Search, standard Shopping, and maybe a YouTube Demand Gen campaign if you have the creative for it. At that spend level, PMax's learning requirements will eat you alive.
The uncomfortable truth
Here's what nobody at Google will say out loud but every experienced operator knows: Performance Max was built to reduce the skill gap. Google wants paid media to be accessible to advertisers who can't or won't hire someone who knows what they're doing. For those advertisers, PMax is a genuine improvement — because the counterfactual is them setting up bad manual campaigns and losing even more money.
For advertisers who do hire someone who knows what they're doing, PMax is a tool. An increasingly good tool. But it's not the whole toolbox, and any operator who tells you otherwise is either selling you something or hasn't looked critically at their own account data.
If your agency is running your account with PMax as 80%+ of spend and no structural rigor around it — book an audit. You're probably leaving 20-40% of your returns on the table, and it won't show up in the PMax reports because PMax is, by design, the thing reporting on itself.