Performance Max: What You Can and Cannot Control
Performance Max asks you to hand over a lot. You give Google goals, a budget and creative, and its automation decides where and to whom the ads appear. The decision for you is not "use it or not". It is which levers you keep and which results you accept on trust.
Why "let the algorithm run" is not a plan
The common advice is to feed PMax assets and wait. That ignores two facts. The system optimises toward the conversion signal you give it, so a weak signal produces weak decisions. And it reports at a level that is coarser than a Search campaign, so you see less of what happened.
Automation is good at finding patterns. It does not know your margin, your stock problems or which customers you want. Those stay with you.
The control map
Think in three layers: inputs, guardrails and measurement.
Goal
The system decides: Conversion action and value you choose
How bids move per auction
Budget
The system decides: Daily budget and any target
Where inside the campaign it is spent
Creative
The system decides: Assets in each asset group
Which combination is shown
Audience
The system decides: Signals you suggest
Who it actually reaches
Guardrails
The system decides: Negatives, brand exclusions, URL settings
Matching inside those limits
Google's documentation lists campaign-level controls such as negative keywords, brand exclusions, search themes and final URL expansion settings. Which ones are available can change, so confirm in the current documentation and in your own account.
Inputs worth your time
- Conversion signal. Optimise for purchases with a value, not for every small action. See conversion actions setup.
- Asset groups. Build one group around one product theme or audience, not a mix.
- Audience signals. Treat them as hints that start the learning, not as limits on delivery.
- Product feed quality. For stores, the feed is a main input. See the feed quality checklist.
Conversion signal
Decide what counts as success and what it is worth.
Asset groups
One theme per group keeps the message clear.
Audience signals
Hints for learning, not hard limits.
Product feed
Titles, prices and availability shape what can show.
Guardrails
- Brand exclusions. Without them PMax may serve on your own name and take credit for demand that already existed. See brand vs non-brand.
- Negative keywords. Check the current documentation on what your campaign type supports.
- Budget. A budget cap is the most direct control you have.
A decision path
Illustrative decision, not real results.
Signal
Is purchase tracking accurate and valued?
YesMargin
Do you know break-even return?
YesFeed
Is the product data clean?
YesBrand
Are brand exclusions or a brand campaign in place?
YesReview
Is there a date to judge it against break-even?
Yes
If any answer is no, fix it first. PMax amplifies what it is given, good or bad.
Trade-offs and when this changes
- Low data. With few conversions, automation has little to learn from. Search or Shopping campaigns may give you more visibility. See Search vs Shopping.
- Need for transparency. If you must know exactly which queries and placements spent money, PMax shows less than Search.
- Large catalogue. Automation can help you cover many products.
Reading the results without overclaiming
Judge PMax over a period long enough to smooth daily noise, and avoid changing several things at once. If you change the budget, the assets and the conversion goal in one week, you will not know which change mattered.
Ask three questions of any report. Did total store sales rise, or only the platform figure? Did new customers arrive, or did returning ones get relabelled? Did the margin survive after ad cost? A good platform number that fails these questions is not a good result.
Keep a record of each change with its date. When performance moves, that record is the only way to connect cause and effect, because the system will not explain its own decisions in detail.
What you cannot measure
You cannot fully separate what PMax created from what it captured. Its return may include people who were already going to buy. Compare against a baseline period and your store's total sales rather than only the platform figure. See store sales vs ad platform numbers.
Your next action
Write down the five answers above for your account. If more than one is no, delay the scale-up and fix those first. The ads management service can review your setup.