Brand vs. non-brand: the costliest mistake in your report
Brand queries rescue the account average and hide how you are really doing. In a public set of 225 geo tests, brand search came out at a median incremental ROAS of 0.70× — below the line of profitability.
- Brand has a high conversion rate and low CPC, so in a combined report it lifts the average and hides non-brand.
- In a dataset of 225 DTC geo tests, brand search had a median incremental ROAS of 0.70× — below break-even.
- Switching brand off blindly is not the answer though; there are cases where defending it pays. A test decides, not an opinion.
When someone shows a report where the account has a ROAS of 9, the first question is never “how did you do that”. It is “how much of that is brand”. In nine cases out of ten the answer is “most of it”, and in eight of those nine nobody had split it out.
Brand queries are searches for your name, your product name or your domain. Someone searching “acme laptop” already knows where they want to buy. An ad on such a query is cheap, converts excellently and looks like the best campaign in the account. The problem is that it measures something other than you think.
Why brand distorts everything else
Imagine an account where the brand part spends €800 and brings €16,000 in revenue (ROAS 20), while the non-brand part spends €3,200 and brings €9,600 (ROAS 3). Together it looks like this:
| View | Spend | Revenue | ROAS | What follows |
|---|---|---|---|---|
| Whole account | €4,000 | €25,600 | 6.4 | “It works brilliantly, let’s add more” |
| Brand only | €800 | €16,000 | 20.0 | People who already know you |
| Non-brand only | €3,200 | €9,600 | 3.0 | Actual acquisition of new customers |
If the company runs a 25 % margin, break-even ROAS is 4. The whole account therefore looks profitable, but acquiring new customers is loss-making. And the decision to “add budget” will go almost entirely into non-brand, because brand has limited volume. The added money will lose — and in a combined report it will look like a mild dip in the average.
The second problem: cannibalising organic
Some of the people who click a brand ad would otherwise have clicked the organic result right below it. That one is free. The platform still takes credit for the conversion, because the click happened on the ad.
How large that share is cannot be derived from attribution — it has to be measured by experiment. The largest public dataset we found comes from the vendor Stella and covers 225 geo tests of DTC brands between August 2024 and December 2025. The median incremental ROAS of brand search in it was 0.70×, below the break-even line of 1.0, and the lowest of all channels tested.
It is one vendor’s dataset, DTC brands only, predominantly the US market. It cannot be transferred directly to B2B or to a shop with little brand awareness. As an indication it is strong though, and it matches what we see in geo tests with clients: platform ROAS for brand tends to be a multiple of the incremental figure.
So switch brand off?
Not blindly. There are situations where brand advertising has a demonstrable benefit:
- Competitors bid on your name. When a competitor appears above your organic result, some people go to them. Here brand defends revenue you already have.
- Weak organic position. New brands, generic names (a company called “Tiles Ltd”), or strong news and directory results above you.
- You need control of the message. A current promotion, a new product, crisis communication. You do not control the organic snippet; you do control the ad.
- Ad extensions. Sitelinks, prices and contact details take up space the competition would otherwise get.
In all four cases the benefit is defensive, not growth. Brand does not belong in a report about acquiring new customers even then.
How to measure it
There is only one reliable answer: an experiment.
- Geo test. Pick a group of regions, switch the brand campaign off there and leave it running elsewhere. After a sufficient period compare total revenue in the test and control regions — not campaign performance, but total revenue from those areas.
- A time-based test is weaker but better than nothing: alternate weeks with brand on and off, watching seasonality and concurrent campaigns.
- Watch the total, not the channel. If brand advertising disappears and organic clicks rise by ninety per cent of that volume, you know what you were buying.
How to set such a test up, how long it must run and how large a difference you can even detect is covered in incrementality and geo experiments.
How to separate brand in the account
- A dedicated search campaign for brand queries only, on exact and phrase match.
- Negative keywords in non-brand campaigns. Brand name, domain, common misspellings and hyphenated forms.
- Brand exclusions in Performance Max. Otherwise PMax appropriates brand and you will not see it in the report. See the article on PMax.
- Your own brand-term list in analytics, so you see the same split in GA4 and Search Console.
- Report in three rows, always: brand, non-brand, total. Never total alone.
Edge cases
Not everything is black and white. A few situations you will meet:
- Brand + product (“acme running shoes”). We count it as brand, because the person already knows the brand.
- Competitor brands. Unambiguously non-brand, and usually an expensive discipline with a low conversion rate. Measure it separately from both.
- A brand that is also a common word. Only a careful negative list and patience help. No automation will split this correctly.
- Supplier brands in your shop. If you sell shoes of brand X, then “X shoes” is non-brand for you, because the person is not looking for you.
What to take away
Brand is not a fraud, but it is an accounting item that belongs somewhere other than performance. Until you have it separated, you do not know whether your advertising brings new customers or merely invoices for the ones you already have. Separating it is an afternoon’s work; what comes out usually changes the budget for a whole year.
Common questions
Should I switch off brand campaigns?
Not blindly, but you should verify what they contribute. Brand advertising is often largely non-incremental, because some of those people would have clicked the organic result just below the ad. In a public dataset of 225 geo tests of DTC brands, brand search had a median incremental ROAS of 0.70×, below the break-even line. At the same time there are cases where defending brand makes sense: when competitors bid on your name, when your organic position is weak, or when you need control of the message. A geo test should decide, not an opinion.
How do I tell how much of my results come from brand?
You need brand queries in their own campaign and excluded from the others, including brand exclusions in Performance Max. Then always read the report in three rows: brand, non-brand and total. The figure for the whole account on its own says nothing — the brand part has a high conversion rate and a low cost, so it lifts the average and hides how you are doing at winning new customers.
What if competitors bid on our brand name?
Then a brand campaign has defensive value and usually pays. People searching your name may click a competitor’s ad above your organic result. Measuring it works like anything else: switch the brand campaign off in selected regions and watch what happens to total revenue from those areas, not just to the campaign’s own numbers. Even then brand belongs in the report separately — it is a defence of existing revenue, not acquisition of new customers.
- How to Run a Google Ads Incrementality Study, StellaSource of the median incremental ROAS of 0.70× for brand search from 225 geo tests (DTC brands, August 2024 to December 2025). One vendor’s dataset, DTC only, predominantly the US market.
- Incrementality Testing with Geo Holdout Experiments, SegmentStreamGeo holdout methodology including an example brand test with a positive result (confidence interval +22 % to +49 %). Shows the outcome is not universal — it depends on the brand and the market.
- Explaining incrementality, PPC LandGeneral explanation of the difference between attributed and incremental results.