Performance Max needs adult supervision

Performance Max is the most consequential and least accountable product in Google Ads.

It works. That is not the controversy. The controversy is that its reporting makes almost any result look like a win, because brand search, remarketing and genuinely new demand are blended into a single ROAS number you cannot decompose.

Run it without controls and you will scale a campaign that is mostly taking credit for demand you already had. Run it with controls and it becomes one of the most efficient tools in the account.

Here are the five controls we install before we let it scale.

1. Separate branded search

This is the first and largest distortion. Branded queries convert at a high rate no matter what; when they run inside Performance Max, the campaign reports them as prospecting wins.

Move brand to its own campaign with exact and phrase match, and exclude brand terms from Performance Max where possible. Now you can see what the campaign does with genuinely new demand.

2. Map asset groups to products, not to your org chart

The most common structural failure is a single asset group representing “all products”. Performance Max then optimizes toward whatever converts cheapest, which is rarely your strategic product.

One asset group per product line or margin tier. Feed it the right products, the right creative and the right landing page. This single change often moves contribution margin more than any bidding tweak.

3. Feed it margin, not revenue

Google optimizes what you tell it to optimize. If you send revenue, it will find high-revenue, low-margin orders. If you import contribution margin as the conversion value, it will find profitable ones.

This requires a measurement rebuild — server-side tracking, offline conversion imports, a true margin calculation — and it is worth every hour.

4. Watch search themes, not just reports

Performance Max will happily expand into irrelevant queries if nothing stops it. Review the search themes report weekly, maintain a shared negative list, and set account-level brand exclusions.

The goal is not to micromanage. It is to catch expansion into categories where you have no business converting.

5. Test incrementality on a schedule

Platform-reported ROAS is a hypothesis, not a fact. Test it:

• Geo holdout: pause or reduce PMax in matched markets for four weeks, hold other channels constant, compare conversion rates.
• Spend-down experiment: use Google Ads Experiments with a pre-registered hypothesis and a fixed decision rule.

Do this quarterly. You do not need a perfect test — you need an honest one.

A quick self-check

Answer these five questions about your account:

1. Can you report Performance Max ROAS excluding branded traffic?
2. Does each asset group map to a distinct product line or margin tier?
3. Are conversion values based on margin, not revenue?
4. Do you review search themes weekly?
5. Have you run an incrementality test in the last six months?

Three or more “no” answers means you are scaling on a number that is partly fictional. The fix is not to turn PMax off. It is to make it accountable.