Performance Max works when you have stable conversion tracking, at least a couple of dozen conversions a month and a clear goal, ideally an online shop with a product feed. It burns money in new accounts without data, in lead generation where nobody measures lead quality, when it's allowed to grab your brand searches, and when a small budget is spread across six channels.
What is Performance Max, in plain English?
In a regular Search campaign you choose the keywords, write the ads and set the bids. In a Performance Max campaign (often called PMax) you hand most of that over. You upload headlines, descriptions, images, video and, if you have one, a product feed from Merchant Center. Google then builds the ads itself and shows them across all its channels at once, wherever the algorithm thinks a conversion is most likely.
What matters is what it optimises for. Performance Max optimises only for the conversions you've told it to count. If a purchase is your conversion, it looks for buyers. If a submitted form is your conversion, it looks for people who submit forms, whatever they write in them. An algorithm knows nothing about quality, only what you measure, and Performance Max is the clearest example of that I know.
When does Performance Max work?
In my accounts, when PMax delivers, three conditions are almost always met.
E-commerce with a product feed. With a product feed from Merchant Center, Google knows exactly what you sell, at what price and whether it's in stock. The Shopping ads in search results often bring in most of the results, and the other channels add to them. For an online shop with a few hundred products, PMax is now usually a better choice than building Shopping campaigns manually.
Stable conversion tracking and enough conversions. The algorithm learns from every conversion. Too few, and it has nothing to learn from. My rule of thumb is at least a couple of dozen conversions a month per campaign, preferably more. Google itself often cites 30 conversions in 30 days as a benchmark for Smart Bidding in general. Below that, results get erratic: one good week, two bad ones, and no way of telling why.
A clear goal with a value. PMax works best when you can say "a purchase is worth €110 on average" or "a qualified quote request is worth €270 to us" and bid towards a target return on ad spend (tROAS) or a target cost per conversion (tCPA). That gives the algorithm something to compare against when it decides how much a click is worth.
When does Performance Max burn money?
The same things that make it good make it expensive when those conditions are missing.
New accounts without data. An account that's had tracking for three weeks and recorded eleven conversions gives the algorithm nothing to go on. So it guesses, and you pay for the guesswork. Start with a regular Search campaign, build up conversions for a few months and switch later.
Lead generation without measuring lead quality. This is the most common way I see B2B and service companies burn money. The form counts as a conversion. PMax quickly works out that the cheapest form fills come from Display and Discover, and delivers plenty of them. Leads go up, cost per lead goes down and the report looks fantastic. Then sales starts calling round and finds students, job seekers and people who thought they were buying something else. Junk leads count as success because nobody told the algorithm otherwise. The fix is to feed quality data back: mark qualified leads in your CRM and import them as offline conversions, or at least count a booked meeting rather than a submitted form. That's the kind of thing I help with in analytics and measurement.
Cannibalising brand searches. Without exclusions, PMax happily grabs searches where someone is already typing your company name. They convert brilliantly (the person was looking for you, after all) and the campaign takes the credit. PMax looks like a genius when all it really did was pick up fruit that had already fallen. Add a brand exclusion (more on that below).
Small budgets spread too thin. Around €270 a month split across Search, YouTube, Display, Gmail, Discover and Maps leaves no channel with enough to learn from. As a rough rule of thumb, I want a daily budget of at least ten times the expected cost per conversion before I start a PMax campaign. Below that, a plain Search campaign is almost always better.
How do you set up Performance Max properly?
Asset groups
An asset group is a bundle of headlines, descriptions, images and videos built around one theme, such as a product category or an audience. Create one per clear theme, not one for everything. Fill each group with as many variations as you can: several headlines, images in landscape, square and portrait formats, and at least one video of your own. If you don't upload a video, Google makes one from your images, and it's rarely something you'd want your name on.
Search themes
Search themes are how you hint to the algorithm which searches are relevant. They aren't keywords: Google may go beyond them if it thinks that will bring more conversions. But early on, when the campaign has no data, they point it in the right direction. Add what your customers actually type, not your internal name for the product.
Brand exclusions
Create a brand list in Google Ads with your company name and its common misspellings, and add it as an exclusion on the PMax campaign. Your brand searches then go to a separate Search campaign, where they belong and clicks are cheap, and PMax has to find new customers instead of taking credit for existing ones. It's the single most important setting if you want to read the results honestly. If you've taken over an account where PMax is already running, I cover what else to check in my article on the Google Ads account you inherited.
Channel insights, and what you still can't see
Google has opened up the reporting in recent years. There's now a channel report showing how much went to Search, Shopping, YouTube, Display and the rest, plus search term insights and results per asset group. Check them every month. If most of the budget ends up on Display and Discover while lead quality drops, you know where the problem is.
But be clear about what you still don't get. You can't set bids per channel, and you have far less control over where the money goes than in a Search campaign. That's the price of letting Google drive. Sometimes it's worth paying, sometimes not.
Why does Google recommend Performance Max to everyone?
Open Google Ads today and you'll be nudged to start a Performance Max campaign, whatever you sell and whatever state your account is in. The recommendations in the interface, the calls from Google's support team and often agencies too all push it. Not because it suits everyone, but because it's easy to scale, moves budget into Google's own channels and makes you more dependent on its automation.
That doesn't make PMax bad. It means you should make the call yourself, based on your own data, not because a blue button says "recommended". And I'll be honest about my own uncertainty: even in accounts where PMax works well, I can rarely say exactly why, because you can only see so much. I can see that it works, not always how. If that makes you uneasy, a regular Search campaign with a Shopping campaign alongside it is still a perfectly good way to advertise on Google. You can read how I work with both, so that you end up running the account yourself, on the Google Ads page.
What to do next
- Count your conversions over the last 30 days. Fewer than a couple of dozen? Stick with Search and build up data.
- Check that your conversion tracking measures something with real value (a purchase, a booked meeting, a qualified quote request), not just a submitted form.
- Create a brand list and add it as an exclusion before you launch the campaign.
- Build one asset group per theme with your own video, and add search themes.
- Block out half an hour on the first Monday of every month to review the channel report and asset group results.
- Want to know where your account stands today? The quick analysis is five questions and takes two minutes.
