Your brand search isn't the hero your dashboard thinks it is

I pulled ninety days of Google search-ad performance for a consumer brand I look after, sorted by return, and the picture was almost comically lopsided. The campaigns bidding on the brand's own name looked like superstars. Everything else — the generic product terms, the competitor terms — was quietly setting money on fire.

If you only read the dashboard, the move is obvious: pour everything into brand, kill the rest. That's exactly the wrong lesson, and it's the kind of confident-but-backwards call that platform reporting nudges you toward every single day.

Why brand search always looks like a hero

When someone types your brand name into Google, they were already coming to you. The ad just stands in the doorway and takes the credit. Google's own attribution is delighted to hand brand campaigns the full value of that sale, because the platform is the one being paid to count it.

The marketing-science firm Recast puts it bluntly in their breakdown of brand-search incrementality: last-click attribution and Google systematically over-credit brand terms, because the visitors were "already looking for your brand, so the credit should be mostly assigned to the initiatives that drove that search." They point to the famous experiments — P&G cutting brand-search spend with no change in sales, eBay finding the same — where pausing brand ads changed almost nothing, because organic search caught the demand anyway.

So that gorgeous return on brand is partly real and partly a mirror. Some of it defends against competitors squatting on your name, which is worth paying for. But a big slice is revenue you would have banked for free. A dashboard can't tell you which slice is which. Judgement can.

The number that looks best on the report is usually the one doing the least new work.

The expensive terms were expensive for a boring reason

The losing campaigns — competitor names, broad product phrases — weren't a lost cause. They were just badly built. The competitor campaign had been running on broad match, which let Google interpret "elektrischer Lattenrost" and a dozen rival brand names as a vague suggestion rather than an instruction. It spent freely and converted into nothing of value.

The fix wasn't to switch off non-brand. It was to give non-brand a reason to exist. I rebuilt those campaigns around specific buyer situations — the person searching because of reflux, post-surgery recovery, a new baby, a snoring partner — and pinned ad copy that spoke to exactly that moment instead of generic product specs. Tight phrase match instead of broad. A shared negative-keyword list so the brand's own automated campaigns stopped bidding against the new ones in the same auction.

That last point matters more than it sounds. If you launch sharp new search campaigns while a Performance Max campaign is roaming the same keywords, you're paying to compete with yourself.

Why I didn't just leave it all to Performance Max

Performance Max will happily serve search inventory with text assets, and Google would love for you to let it. But you can't pin a specific use-case message to a specific query, and historically you couldn't even see which individual headlines and assets were doing the work — Google has only recently started opening up that asset-level reporting. For a use-case angle and competitor defence, that's not good enough. I want to control the message and read the result. So those got dedicated, old-fashioned search campaigns, and PMax kept the broad commercial demand.

This is the whole game with AI-driven ad platforms, and it's the same point I made when I pointed an AI agent at a real Meta account: the automation is genuinely good at execution and genuinely bad at knowing what it's optimising for. It will confidently scale the thing that looks profitable on its own scorecard. Someone with the reps has to decide which numbers are real — that's the domain expertise that's actually the leverage, not the tool. It's also the same lie of omission I wrote about when your analytics dashboard leaves things out: the report isn't wrong, it's just answering a narrower question than you think.

What to actually do with this

Don't kill non-brand because the dashboard told you to. Ask whether it's losing because the demand isn't there, or because the campaign was built lazily — they look identical in the spreadsheet and they could not be more different. And treat your heroic brand return with suspicion. The cheapest experiment in all of marketing is pausing brand search in one region for two weeks and watching whether sales actually move. Mostly, they don't budge as much as the report implied.

The platform optimises for what it can measure. Your job is to remember everything it can't.

Sources & further reading

External Recast — The Conundrum of Brand Search: Is It Incremental? Google Ads Help — About asset reporting in Performance Max Forbes / Augustine Fou — When big brands stopped spending on digital ads, nothing happened

Related posts I pointed the Meta MCP at a real ad account — here's what it found in ten minutes The death of generic AI: why deep domain expertise is the only real leverage left Your analytics dashboard is lying to you by leaving things out