Contents
Introduction
English version of an article first published in Polish on 25 June 2026. Read the Polish original.
Performance Max is not a magic "make me money" button. It is an engine that runs only as well as the fuel you pour into it. It is the most automated format in Google Ads: one campaign serves ads in Search, on YouTube, across the Display Network, in Gmail, Discover and Maps, while Google's AI decides on bids, audiences and creatives.
It sounds like a marketer's dream come true, and this is where the trouble starts. "Automatic" does not mean "maintenance-free". I have seen accounts where PMax delivered a ROAS of 5×, and accounts where it quietly burned half the budget on brand traffic and junk conversions. The difference was not in the algorithm. It was in what the algorithm had been fed.
Below I show how to handle Performance Max like a professional: what happens under the bonnet, why conversion measurement matters more than budget, and how to regain control in the places where Google would gladly take it away from you.
What you will learn:
- What Performance Max is and which Google channels it covers from a single campaign
- How the algorithm makes decisions and what it really needs to work well
- Why the quality of conversion data decides the result more than the size of the budget
- How to build asset groups and audience signals so the AI does not wander
- How to regain control: brand exclusions, feed management, brand safety
- How to report on and optimise PMax despite limited visibility
What Performance Max is
Performance Max is a campaign type in which one campaign covers all Google channels at once, and an algorithm steers serving and bidding. Instead of building one campaign for Search, another for YouTube and another for the Display Network, you give Google one goal, a full set of assets and a budget, and it distributes them wherever its predictions say a conversion is easiest to get.
Under one roof, PMax combines:
- Search: text ads on queries related to your offer
- YouTube and Discover: video and image ads in high-reach channels
- The Display Network and Gmail: remarketing and reaching new audiences
- Maps and Shopping: local traffic and products from the Merchant Center feed
Put plainly: in Search you choose the keywords, in Performance Max the algorithm chooses them. The whole art is to give it good hints and good data.
How Performance Max uses AI
The AI in Performance Max works in real time and, at every auction, answers one question: does this user, in this channel, at this moment, have a chance of converting, and how much is that worth paying for. On that basis it sets the bid (Smart Bidding), picks the channel and assembles the creative from your assets.
In practice the algorithm does four things at once:
- Predicts conversion for every auction, based on signals about the user and the context
- Bids automatically towards the goal: maximum conversions or maximum conversion value (tCPA / tROAS)
- Selects the channel and the audience, wherever the prediction shows the highest chance of a result
- Tests creatives: it combines headlines, descriptions, images and video and promotes the combinations that work
One thing is key: the algorithm learns from your data. It has no knowledge of its own about your business. It has only what you measure and what you feed it. That is why the whole "magic" of PMax stands or falls on the quality of conversion data.
The algorithm (AI)
takes over for you
- bidding and bids (Smart Bidding)
- choice of serving channel
- choice of audience and moment
- assembling creatives from assets
You
stays on your side
- goals and conversion value
- assets and audience signals
- brand exclusions and the feed
- measurement, reports, oversight
Conversion data is the fuel
The most important decision in Performance Max is not about budget or assets. It is about what you count as a conversion and how much it is worth to you. The algorithm optimises exactly towards the signal you give it. If the signal is noisy, the optimisation will be noisy along with it.
Garbage in, garbage out. In PMax that is not a slogan, it is how the thing works. The algorithm chases exactly what you tell it to chase.
The algorithm chases cheap, high-volume, often worthless enquiries
The algorithm looks for real customers, not form fillers
Measure value, not just volume
For e-commerce, set bidding towards conversion value (tROAS), not towards the bare number of transactions. Otherwise the algorithm treats a sale worth 50 PLN and one worth 5,000 PLN as the same "one" and naturally goes where conversions are cheaper and more numerous, which usually means less valuable.
For leads, upload offline conversions
Service businesses make the same mistake: they optimise towards "form submitted". So the algorithm starts collecting cheap, high-volume, often worthless enquiries. The solution is offline conversion import: you pass back from the CRM which leads turned into real sales, and only then train the campaign on that. At that point PMax starts looking for customers, not form fillers.
From my practice: A client in the services sector had a "great" cost per lead: 38 PLN. The problem? 80% of those leads never picked up the phone. After connecting offline conversions with lead quality, the cost "rose" to 120 PLN, but the number of real customers tripled. The algorithm finally knew who to look for.
Asset groups and audience signals
You build the structure of Performance Max around asset groups and audience signals. These are the two places where you really influence how the algorithm understands your offer. The rest is its own decisions.
Asset groups = themes
An asset group is the equivalent of an ad group: a set of headlines, descriptions, images and video around one theme. Do not throw the whole offer into one bag. Split it into groups by category, margin or intent. A fashion shop will create separate groups for "footwear", "jackets" and "accessories"; a service business, one for each main service.
And provide a full set of assets, video included. If you do not upload a video, Google will generate one automatically from your images, and it usually looks the way automatically generated video can be expected to look. It is better to supply your own, even a simple one.
Audience signals = a hint to start from
An audience signal is not rigid targeting. It is a hint about who the algorithm should start looking for. Put in your customer lists, custom segments built from competitor and intent keywords, and remarketing data. This shortens the learning phase, because the algorithm does not start from zero. Over time it will go beyond that signal anyway, and that is fine: it is what prospecting is about.
How to regain control of the campaign
Performance Max hands a lot over to the algorithm, but not everything. There are a few levers with which you really steer the campaign. Professional PMax management is largely about using them deliberately.
- Brand exclusions: so that PMax does not collect cheap brand traffic; in some accounts they are enabled through Google support
- Negative keyword lists: they cut off queries you do not want to appear on
- Feed management: for e-commerce this is a powerful lever; custom labels in Merchant Center let you separate products by margin or turnover
- Brand safety and placement exclusions: so that ads do not end up in random apps and on unwanted channels
- Final URL expansion: turn it off if you do not want PMax to send traffic to arbitrary pages of the site
Traffic on your own brand is the easiest "conversion" in the world. If you do not exclude the brand, PMax will proudly take credit for sales you would have had anyway.
Reporting despite the "black box"
Performance Max has a reputation as a "black box", but the box has a few windows. You just need to know where to look. You will not get a full breakdown of every penny on every keyword, but there is more data for sensible optimisation than it seems.
- Insights: they show trends, seasonality and growing query categories
- Search terms insights and search categories: they indicate what the budget is really being spent on
- Results at asset group level: which themes and which assets carry the result, and which hold it back
- Scripts and reports: the community has written scripts that pull out the placements and search terms the interface does not show by default
The most important thing I keep under constant watch is cannibalisation of Search. PMax can "poach" conversions from search campaigns and report them as its own. Without segmentation it is easy to believe that PMax is working wonders, when it has merely taken the credit.
Performance Max vs Search: when to use which
This is not an either-or choice. Search and Performance Max play different roles and work best as a pair, with clearly separated jobs.
Search campaign
- Full control over keywords and intent
- Best for hot, purchase-ready queries
- Handles brand traffic
Performance Max
- Maximum reach and prospecting
- E-commerce with a product feed
- Scaling once Search has hit its ceiling
My default setup: a strong, controlled Search campaign for purchase-intent and brand keywords, and next to it PMax for scaling and reaching new audiences, with the brand excluded so that the roles do not overlap and the reports do not lie.
The most common Performance Max mistakes
Most PMax failures are the fault of the setup, not the algorithm. These are the mistakes I see most often:
- Wrong conversion setup: optimising towards micro-conversions, or towards volume alone instead of value
- One bag for everything: no split into asset groups by margin or category
- No video of your own: leaving the creative to the machine and being surprised by weak results on YouTube
- No brand exclusion: PMax takes over brand traffic and inflates its own results
- Nervous changes: frequent adjustments to budget and goals reset the learning phase
- Judging after a week: PMax needs time and data; a verdict after 7 days is reading tea leaves
Start
campaign launch
Learning phase
~6 weeks · collecting data
Stability
optimal serving
Frequent changes to budget and goals reset the learning phase. Let the campaign collect data in peace.
FAQ: common questions
What is Performance Max?
Performance Max (PMax) is a Google Ads campaign type in which one campaign serves ads across all Google channels at once (Search, YouTube, Display, Gmail, Discover and Maps), while Google's AI decides on bids, audiences and creatives. You provide the goal (conversions or their value), the assets and the signals, and the algorithm decides who sees the ad, where and at what price.
How is Performance Max different from a Search campaign?
A Search campaign runs only in the search engine and gives you full control over keywords, while Performance Max covers all Google channels and hands the steering to the algorithm. You choose Search when precision and purchase intent matter; PMax when you want reach, new audiences and (especially) e-commerce sales from a product feed. In practice they work best together, with clearly separated roles.
Is Performance Max worth it for a small business?
Yes, on one condition: conversion tracking has to be set up correctly and there has to be a sufficient stream of data for the algorithm to learn from. For a business with a few conversions a month, PMax often has nothing to optimise on and burns through the budget. For a shop with a feed or a service business with regular leads it can be very effective, even on a modest budget.
How much budget does Performance Max need to learn?
Google says the learning phase usually lasts up to about 6 weeks and needs a stable stream of conversions. In practice that means at least somewhere between a dozen and a few dozen conversions a month. Stability matters more than the amount itself: frequent changes to the budget, goals or assets reset the learning. It is better to set a sensible budget and let the campaign collect data in peace than to tweak it every few days.
How do I exclude brand traffic from Performance Max?
The most reliable options are brand exclusions, available at campaign level, and negative keyword lists. In some accounts these have to be enabled through Google support. It matters because without them PMax happily collects cheap clicks on your own brand and takes credit for sales that would have come in anyway. Brand traffic is usually better handled by a separate, controlled Search campaign.
Will Performance Max replace a Google Ads specialist?
No, it only shifts where the work is. The algorithm takes over bidding and serving, but someone has to look after what it is fed: correct measurement and conversion values, asset quality, audience signals, exclusions, the product feed and the interpretation of reports. Unsupervised, PMax optimises towards whatever you give it, mistakes included. Less clicking around the interface, more strategy and data analysis.
Worth remembering
- PMax is not "set and forget". You delegate bidding and serving to the algorithm, but oversight of data and assets stays on your side.
- Without correct measurement of conversions and their value, the algorithm optimises in the dark. Measurement is the foundation, not an extra.
- Separate brand traffic so that PMax does not take credit for sales that would have come in anyway.
- Feed the campaign a full set of assets and a strong audience signal, but leave the algorithm room to test.
- For lead generation, upload offline conversions with lead quality. Otherwise the AI chases cheap, worthless enquiries.
- Report at the level of asset groups, insights and search terms. The "black box" has windows, you just need to know where to look.
Summary
Performance Max rewards those who treat it as an engine to supervise, not as an autopilot. The algorithm takes bidding and serving off your hands, but in return it demands clean conversion data, well-thought-out asset groups and deliberate exclusions. Give it that, and it can scale results like no format before it.
PMax is not smarter than your data. It is exactly as smart as what you feed it.
If your Performance Max "sort of works" but you do not know why, start with an audit of conversion measurement. In 90% of cases that is exactly where the difference between burning budget and scaling lies.
Sources and further reading
Official Google documentation worth basing campaign decisions on:
Is your Performance Max burning budget?
I will check conversion measurement, structure and exclusions, and show you where the money is leaking. No sugar-coating.
See the Google Ads serviceHow this text was written
I will not pretend it was written entirely by hand, by candlelight. For the research, the proofreading and the code of this page I used a language model. It is a good tool for digging through documentation and catching the typos you no longer see on the third read of your own text. The English version was also prepared with a model, from my Polish original.
What the model did not do: it did not choose the topic, form the arguments or decide what this data means for your site. That stays with me, as does the last word on what goes into the text and what comes out of it.
Where the model helped
- digging through sources and organising data
- proofreading and the English version of the text
- code, page layout and infographics
What stays with me
- the choice of topic and what is worth saying about it
- the arguments, the reading of the data and the conclusions
- fact-checking and the decision on what goes into the text