Performance Max: what you actually control and what you do not

Performance Max has settled in as the default choice for online shops and as a permanent source of argument: the client sees good numbers, the analyst sees that the campaign got to them by appropriating brand queries. Both observations are usually true.

This article does not list every box in the interface. It ranks the levers by how much they actually change.

1. Brand exclusions — without them none of this means anything

If PMax has no brand exclusions, it takes people searching directly for you. Those are the cheapest and most willing customers, so the campaign looks excellent. But most of those orders would have arrived anyway — they would have overflowed into a brand search campaign or into organic results.

What to do:

• Set brand exclusions at campaign or account level, including misspellings and variants.
• Keep a separate brand search campaign. It handles brand queries cheaply and, more importantly, visibly.
• Consider excluding competitor brands too, if you do not want to target them deliberately.

Only after this step does the PMax report start to say anything. Why it matters so much is in brand vs. non-brand.

2. Feed quality — the biggest lever for shops

For an online shop a substantial part of PMax runs on the product feed. Disapproved products, missing identifiers, wrong prices or unintelligible titles mean part of the range never appears at all. The budget then ends up on a handful of items that happen to work well in the feed.

An hour spent on the feed usually does more than a week of reconfiguring the campaign. Detail in product feeds and Merchant Center.

3. Asset group structure — your map of the range

One asset group for a whole shop tells the system “I do not mind what you sell”. Sensible division follows margin and commercial priority, not the categories in your shop:

• High-margin products separately from low-margin ones. Otherwise budget flows to where things sell most easily.
• Bestsellers separately from the long tail, because they warrant a different return target.
• New and seasonal lines separately, so established items do not drown them.

Every group needs enough conversions for anything to be evaluated. Ten groups at three orders a month is worse than three groups at ten.

4. Assets: copy, images, video

In PMax you do not decide where the ad appears, but you do decide what the system assembles the message from. That is more control than it seems:

• Upload your own video. If you do not supply one, the system generates it from your images and text, and the result usually does not match how you want to come across.
• Give the assets range. Different headline lengths, different angles. Five variations of the same thing are one variation to the system.
• Watch the asset ratings and fill in what is missing rather than endlessly tweaking the best performers.

5. Search themes: a hint, not keywords

Search themes tell the system what your business is about when it has no way of inferring it — for new campaigns, new product lines, or queries your pages do not contain. The limit has changed over time (from an original ten, through twenty-five, to today’s fifty per asset group), so check the current figure directly in Google Ads Help.

6. Negative keywords and lists

These are now available at campaign level and as shared lists. Use them with the same discipline as in search campaigns: go through the search terms report regularly, exclude the nonsense, and find the queries that deserve their own search campaign and landing page. The method is in negative keywords.

7. What you do not control (and it is better to know)

• The split of budget between channels. How much goes to search, to YouTube, to display is not yours to set.
• Full placement transparency. Reporting has improved, but the detail is nowhere near what you know from classic campaigns.
• Which asset combination the system assembles. You supply material, not a finished ad.

PMax or a search campaign?

When PMax and a classic search campaign compete for the same query, the one with the better relevance signal usually wins. The larger analysis the industry cites (an Adalysis study of roughly 3,300 campaigns) reports that where both were eligible, search campaigns consistently had a higher conversion rate. We were unable to locate the original study in primary form, so treat it as an indication rather than proof.

The practical conclusion from the accounts we run is the same though: the highest-intent queries belong in their own search campaign, where you can see them and write a specific ad and page for them. Let PMax do volume and discovery across the rest of the range.

How to evaluate PMax

1. Take brand out of the campaign and out of the reasoning. Compare the non-brand result only.
2. Watch new customer cost, not just ROAS. PMax has its own new-customer goal setting, worth switching on.
3. Go through the search terms report. If your brand is in there, the exclusions are not working.
4. At larger budgets, verify incrementality. Platform ROAS for a campaign mixing remarketing with discovery almost always overstates.

What to take away

Performance Max rewards preparation and punishes laziness differently than before: you do not notice it through bad numbers, but through numbers that look good and mean nothing. Exclude brand, fix the feed, give the system honest assets, and keep your most valuable queries in your own search campaign. That is ninety per cent of the work.