Performance Max: built
to run itself - here's
how to take control
Performance Max was sold as the campaign you don't have to manage. The reality is that the accounts getting real results from it are the ones that manage it anyway - structuring it around commercial objectives and stripping out the placements and defaults that flatter in-platform numbers but do nothing for the top line.
Here's what to weigh up before you run it, and the levers that actually move the needle once it's live.
Performance Max will always look great in-platform. Whether it's actually growing the business is a different question entirely.
Quick reminder - what Performance Max actually is
In late 2021 Google rolled out Performance Max (PMax) - the all-in-one campaign solution designed to lower the barrier to entry for advertisers, letting anyone build and launch a campaign in minutes.
It was the first of its kind: a single campaign type showing across Search, Display, Shopping, YouTube, Maps and Gmail, and crucially, without keyword bidding.
It started life as a black box with minimal insight. Since then it has evolved into a popular choice for D2C advertisers - and Google has steadily handed back more of the controls it withheld at launch.
Before you run PMax
PMax isn't the right answer for every account. As a PMax specialist, here's a few things I'd weigh up before committing budget to it:
- It's smart bidding only - you need a minimum of around 30 conversions a month for it to work in your favour.
- It goes warm. It finds the easiest sales first, which makes it a very strong retargeting engine more than a prospecting one.
- Opinions on B2B are mixed. There's a lot of junk-quality volume. Some report success feeding it offline conversions to steer the algorithm, but even then reviews are split.
- It can work for a new brand.
- It needs proper segmentation and structure based on commercial objectives - split by margin, stock age, newness, and profit versus cash-flow goals.
- For Shopping, consider running Standard Shopping as a feeder at a lower tROAS with PMax set higher to capture that demand.
- If you already run Standard Shopping, judge PMax on its impact on top-line revenue and new customer acquisition, not on in-platform numbers (which will look great).
- It won't work if your feed isn't optimised.
How to optimise Performance Max
Once it's live, these are the levers worth pulling - most of them about taking back the control PMax tries to abstract away.
Cut the terms eating budget
The search term report is now available to advertisers, and alongside search themes you can exclude search terms like any regular campaign. That means cutting high-spend, non-converting terms that are draining budget without a positive return.
Your review cadence should match how long it takes someone to buy. If a customer goes from discovery to conversion in under a week, you can review weekly or more often. For a high-consideration, high-AOV product, every couple of weeks to monthly is enough. Review too aggressively on thin data and you risk over- or under-excluding, which feeds back negatively into smart bidding.
Trim what's dragging performance down
Another hygiene piece: exclude locations and devices that are pulling performance down.
Be conservative here, especially with geo. You can go as granular as post-code level, but that only works if you have enough smart bidding conversion history behind it - otherwise you're negating off a data set that's far too small.
See where the spend is actually going
Advertisers can now see the percentage of spend going to each format - what went on Display, what went on Shopping, what went on Search, and so on. This used to be available only via scripting.
For Shopping brands, if you can see PMax spending too much upper funnel (Display, YouTube, Gmail), you can rebuild it as feed-only or switch to a Standard Shopping approach to keep spend on the Shopping network.
You can also build dedicated upper-funnel campaigns with Demand Gen. That's my preferred approach - I'm not a fan of PMax as a true all-in-one.
Manage products by role and performance
Exclude poor performers - SKUs where back-end contribution is negative, or where in-platform reporting shows no orders attributed.
Dynamic performance bucketing works well here. Using scripts and custom labels to segment SKUs and campaigns by performance and role makes this far more manageable than manually pausing and moving products. For example, the SKUs you want to maximise profit on live in one PMax campaign, and any that drop below a back-end profit threshold (say 1 POAS) are removed automatically.
With PMax feed-only, auto-created videos will still show even if you opt out of auto-asset creation. Google still uses Display remarketing in the campaign, serving Shopping ads on the GDN plus video, because that video is used for dynamic retargeting. It's a trade-off you have to accept. And if you add your own video, it forces you to add the rest of the text assets too.
More granular control over placements
Launched in May 2026, you can now opt out of the Search and Display networks. It's currently unclear whether the feed-only auto-video still shows once you do.
Either way, if you want maximum granular control, this lets you keep your PMax campaign showing - in theory - only where you want it. Although if you're excluding this many placements, it's worth asking whether a different campaign type would serve you better in the first place.
Useful, but read the in-platform numbers with suspicion
Tin-foil hat on. Audience signals are part of how Google makes PMax performance look better than it is.
You can't target audiences directly in PMax - you only provide a signal that finds people "like" the ones you include. So it's best to upload first-party customer lists and GA4 lists of your existing customers.
The trade-off: Google will likely retarget those users to inflate in-platform sales and attribution in PMax's favour, encouraging you to spend more - when in reality those existing customers were probably going to buy anyway. If conversion volume is strong, it can be worth testing a completely blank audience signal to see how it performs.
Stop PMax cannibalising your brand terms
A final hygiene piece: make sure brand lists and exclusions are in place so PMax doesn't chase your brand terms and cannibalise existing Search and Shopping campaigns. I'd add competitor terms to that exclusion list too, to prevent wasted spend.
For maximum control and a lower cost of sale, set up manual-CPC branded Search and Shopping campaigns and exclude those terms from all generic campaigns. Otherwise Google inflates your non-brand prospecting campaigns with artificial performance from brand searches - people who were going to convert anyway - and trains the algorithm on the easy wins you don't want it chasing. Let branded campaigns do the brand job. Let non-brand find new customers.
Don't use the new customer acquisition goals. Having run this test, Google charged a premium on CPCs - I saw it as high as 4x my usual - for a small uplift in in-platform reported new-versus-returning customers (a distinction that isn't reliable anyway). Back-end tests showed no uplift at all.
The pattern underneath all of this
Almost every lever above comes back to the same thing: PMax defaults are built to make the campaign look good in-platform, not to grow your business. Taking control means structuring around commercial objectives, judging it on top-line and new customer impact, and being willing to strip out the placements and goals that only flatter the reported numbers.
And if you find yourself excluding networks, placements and goals one by one, that's usually the signal that a more deliberate campaign type would serve the objective better than PMax ever will.
Google Shopping Ads & Feed Optimisation
Want to go deeper on the feed?
PMax lives or dies on the feed underneath it. The Google Shopping Ads & Feed Optimisation Hub brings together custom labels, supplemental feeds, CSS partners and price competitiveness in one place.
Explore the hub →Further reading
The 5 Shopping feed mistakes I find in 9 out of 10 accounts
The most common feed errors holding back Shopping performance - and what to fix first.
Google Shopping Price Comparison
How competitive pricing affects your Shopping impressions and what to do about it without racing to the bottom.