Your brand campaign has the best ROAS in the account. Ask who sent the searcher

Branded search posts the best return in paid search because it reaches people who already chose you. Something else created that demand. Keep a brand campaign where competitors bid on your name or your free listing is weak, report it apart from non-brand, and pause it in a controlled test before assuming it earns what it claims.

A homeowner sees a roofing company's wrapped pickup parked on the next street, notices the yard sign two doors down, and remembers the name. Three weeks later a storm takes some shingles. They type the company name into Google, click the ad at the top of the page, call and book an inspection that turns into a $14,000 replacement. In the ad account that is one conversion from a campaign that cost $1.80 a click. The brand campaign's reported return for the month is somewhere north of twenty to one.

Every part of that report is technically true and almost none of it describes what happened. The truck, the yard sign and the neighbor's recommendation did the work. The ad was the door the customer walked through, and the free listing directly beneath it was a second door into the same room. This post is about what a brand campaign actually buys, when that is worth paying for, and how to stop it from flattering the rest of paid search.

Why branded search always posts the best ROAS

Branded search means ads on queries that contain your business name. It posts the best return in nearly every account because of who is searching, not because of what the ad does: people who type a company's name have usually already decided to contact it, and the ad meets them one step from the sale.

Compare it with a non-brand query such as "roof repair near me". That searcher has a problem and no supplier; the ad has to win them against nine other companies and a map pack. The brand searcher has a supplier and is looking for the phone number. Conversion rates on brand terms are high for the same reason conversion rates on a company's own contact page are high. The intent was there before the page loaded.

So a brand campaign's conversion count is a measure of demand that already exists, plus whatever share of it the ad genuinely rescues from a competitor or a dead end. The second part is the only part the ad caused, and the platform report cannot separate it from the first. Google's own research gives a sense of how much the answer depends on where your free listing sits.

That is the seller of search ads measuring its own product, and even on its numbers the headline 89% falls to about half in exactly the situation brand campaigns live in: a query for your name, where you are almost always the first free result. Half is an average across advertisers in 2010 and 2011, measured in clicks. For a local business with a strong listing and no rival bidding, independent experiments put the figure far lower.

Where the demand behind a brand search comes from

A person searches your name because something put it in their head: a past job, a referral, a vehicle or sign, outdoor or broadcast advertising, or another digital ad they saw earlier. Brand search is the route most of those channels take to reach you, which is why it absorbs their credit.

Every row on that table is a channel whose own report looks weaker than it should, because its customers took a detour through Google. Our posts on billboards and referrals both run into the same wall from the other side: the board or the recommendation works, its dedicated phone number gets modest use, and the rest of its effect is recorded as search. The same mechanism inflates free brand clicks, which is why the post on organic search argues for judging SEO on non-branded traffic only.

The last row deserves its own sentence. Inside paid search, brand campaigns quietly take credit from non-brand ones whenever a customer clicks a generic ad, leaves, and returns days later by typing the name. On a last-click report the generic campaign that found the customer gets nothing and the brand campaign that answered the door gets everything. The general version of that dispute is in our post on channel conflict.

What the brand pause experiments found

Controlled experiments that switched brand ads off found that, for a famous brand with no rival bidders, almost all the paid clicks returned through the free result. Where rivals bid on the name and the brand stopped, a large share of clicks went to them. The value of a brand campaign depends mostly on who else is in the auction.

The best-known test is eBay's. In 2012 it stopped buying ads on queries containing "ebay" on Yahoo! and MSN while keeping them on Google as a control, and found that natural search replaced 99.5% of the lost traffic. That result, and the separate non-brand test that ran alongside it, are covered in our post on incrementality tests. Less often quoted is what happened when eBay repeated the brand test on Google itself.

Notice the qualifier: well-known brands in most consumers' consideration sets. eBay in 2012 had no serious rival bidding on its name and ranked first organically for it everywhere. A roofing company in a metro area with six competitors is a different case, and a later study measured that case directly.

Put the two studies side by side and the decision mostly makes itself. With nobody else in the auction, a brand ad buys a few percent more clicks at a steep effective price. With rivals in the auction, switching it off hands them up to two in five of the people who were looking for you by name. Those are different businesses, and the same campaign is a waste in one and a bargain in the other.

A brand campaign is insurance against competitors, not a source of customers. Where nobody bids on your name, it buys a few percent of extra clicks at a high price. Where rivals do, it keeps a large share of people who were already looking for you.

When a brand campaign is still worth running

Run a brand campaign when someone else would otherwise occupy the top of the page for your name: a competitor, a directory, a reseller or a lead marketplace. It is also worth running when your free result is weak, buried or out of date, and you need to control what a ready buyer sees first.

• Competitors bid on your name. The campaign's auction insights show which other advertisers appear on your brand terms and how often. If the list is empty, the defensive case is thin.
• Directories and marketplaces outrank you. For many local trades, the first free results for a company name are review sites and lead aggregators that will happily sell the inquiry to a competitor. An ad that puts your own number above them is defending a real customer.
• Your listing says the wrong thing. A brand ad can carry this month's offer, emergency hours or a direct link to booking, where the free result shows a stale description. That is message control, and it is legitimate, as long as it is valued as such and not as acquisition.
• A launch, a rename or a move. Searchers using an old name or a misspelling may not find you organically for a while, and the ad bridges it.
• Low cost by design. Brand clicks are usually cheap. A campaign whose job is defense can be capped at a small daily budget and left to do that job, rather than scaled because its ROAS looks heroic.

That is why the defensive case is real for so many local businesses. A competitor cannot usually write your name in their headline, but they can appear above your free listing every time somebody searches for you. Whether that is happening is a fact you can look up in an afternoon, and it should be looked up before anyone argues about the brand budget in either direction.

How to separate brand from non-brand in paid search

Put brand terms in their own campaign, keep them out of every other campaign, and report the two as separate lines from the ad platform all the way to closed sales. A blended paid search return lets a cheap brand campaign carry an expensive non-brand one that would not survive on its own numbers.

Here is how that plays out in one month for a home services business. The figures are illustrative assumptions; the arithmetic is the point.

The arithmetic. Brand: 22 × $3,000 = $66,000, and $66,000 ÷ $1,500 = 44.0 per dollar, $1,500 ÷ 22 ≈ $68 a sale. Non-brand: 15 × $3,000 = $45,000, and $45,000 ÷ $13,500 ≈ 3.3, $13,500 ÷ 15 = $900 a sale. Blended: $111,000 ÷ $15,000 = 7.4 and $15,000 ÷ 37 ≈ $405 a sale. Brand is 10% of spend and 59% of the revenue.

Presented as one line, paid search returns $7.40 a dollar and looks like the best channel in the business. Presented honestly, 90% of the budget returns $3.30 a dollar, which may still be fine at your margins and may not be, but is a different conversation. Now apply the experiments. If a pause test showed that half the brand campaign's sales would have arrived through the free listing anyway, its incremental revenue is about $33,000, or 22 per dollar. If nobody bids on your name and the true figure is nearer eBay's, it is close to nothing, and the blended number was mostly other channels' work.

The split has to exist before the sale does, because the closed-sales record will not tell you what the customer typed. In practice that means:

1. Write the definition down. Is "acme roofing reviews" brand? "acme roofing near me"? Misspellings? Decide once, in writing, because the brand share moves every time someone changes the rule.
2. Keep brand out of non-brand. Add your name as a negative keyword in generic Search campaigns, and use brand exclusions where the campaign type supports them. Broad match and automated campaigns will otherwise drift onto your name, because it converts.
3. Label the lead at first contact. A separate call tracking number on each campaign's call assets, or a campaign value captured from the landing page URL on form leads, so the lead file says "Google Ads brand" or "Google Ads non-brand" rather than just "Google".
4. Carry the label to the sale. Match closed sales to those leads on normalized phone numbers and email, and report brand and non-brand revenue on separate rows.

Performance Max is where this matters most, because a campaign without keywords will find your brand traffic on its own and report it as its own achievement. Our post on Performance Max covers what happens to its numbers when brand is excluded. If you also run Microsoft Advertising, split brand there too; the post on Microsoft Ads explains why imported campaigns are a common place for brand and non-brand to merge without anyone deciding they should.

How to test a brand campaign: the pause and the holdout

Switch the brand campaign off for a fixed period, or in part of your service area, and measure total clicks on your name, paid plus organic, and closed sales from people who searched for you. What the free listing absorbs, the ads were not adding. Check for competitor ads first, because they change the answer.

1. Check the auction. Look at auction insights on your brand terms for the past three months. If rivals appear often, expect a real loss and keep the test short.
2. Pick the design. A geographic holdout, pausing in half your area, is cleaner because the other half controls for season and weather; the method is in our incrementality tests post. A time-based pause of four weeks is simpler and is what most small accounts can run, as long as you compare against the same weeks last year or against non-brand trends.
3. Measure clicks from both doors. Paid brand clicks come from the ad platform; organic brand clicks come from Google Search Console's query report, filtered to queries containing your name.
4. Measure closed sales, not platform conversions. The platform will report conversions dropping to zero in the paused campaign by definition. The question is whether closed sales from name searches, paid and free combined, fell.
5. Set the stop rule in advance. If a competitor starts taking the top ad slot on your name, end the test: you have your answer.

A worked reading of a four-week pause. In the four weeks before, the brand campaign had 1,200 paid clicks at $1.50, so $1,800 of spend, and Search Console showed 2,800 organic clicks on brand queries: 4,000 in total. During the pause, organic brand clicks rose to 3,880 and paid clicks were zero.

The click reading is the stable one: nine in ten paid clicks came back through the free result, and the ones the ads really added cost $15 each, ten times the price the account reports. Compare that with what a non-brand click costs you and you have a defensible way to price the brand budget. The sales reading is a hint at most. A drop from 48 to 46 is two sales, well inside ordinary month-to-month variation for a business of this size, and the honest summary is "no detectable change in sales", not "brand ads are worth two sales a month".

In a brand pause, read the clicks first and the sales second. Total brand clicks tell you how much the free listing absorbs; closed sales tell you whether anything changed that matters, and at small volumes the honest answer is often that no change was detectable.

Reporting brand search honestly

Show brand and non-brand as separate rows, each with matched closed revenue and spend. Put the result of the last brand test beside the brand row, and keep sales nobody can trace in their own bucket. Never present a blended paid search return, and never describe brand search as having created the customers it answered.

• Brand paid search: spend, matched revenue, and the incremental share from the most recent test, with its date. If you have never tested, say that.
• Non-brand paid search: spend, matched revenue and cost per sale. This is the line that tells you whether paid search acquires customers.
• The channels behind the brand searches: vehicles, signs, referrals, outdoor and broadcast, reported with a note that part of their effect appears under brand search.
• Unattributed revenue: sales with no matching lead stay in their own line, for the reasons in our post on unattributed revenue. Moving them into brand because "they probably searched us" turns a guess into a number.

CloseRev does not see search terms, keywords, bids or ad platform data, and it cannot tell a brand click from a non-brand one. It matches the closed sales in one CSV export to the leads in another by normalized phone and email, and credits a matched sale to the source written in that lead's row; sales with no match stay in Direct / Unknown. Brand and non-brand appear as separate channels only if your lead export already labels them separately.

That last condition is the practical takeaway for whoever builds the lead file. If every call from Google arrives tagged "Google Ads", no amount of matching afterward can split the brand campaign from the generic ones, and the blended number is the only one anybody will ever see. The label has to be written when the lead is created.

What to do this month

Find out whether anyone bids on your name, split brand into its own campaign and its own lead label, take it out of every other campaign, and schedule a pause or holdout. Then judge paid search on the non-brand line alone.

1. Pull auction insights for your brand terms. Note who appears and how often.
2. Search your own name on a phone, signed out, and see what a customer sees above and below your listing.
3. Move brand terms into one campaign, add them as negatives elsewhere, and turn on brand exclusions in automated campaigns.
4. Give brand and non-brand separate labels in the lead file, through tracking numbers or captured campaign values.
5. Run a four-week pause or a geographic holdout, with the stop rule written down, and price the brand budget on cost per incremental click.

None of this argues for switching brand campaigns off. For a business whose competitors bid on its name, the experiments say the opposite. It argues for paying for them knowingly, as a cheap defensive line with a tested value, and for giving the credit for those customers back to the truck, the sign, the neighbor and the job well done that sent them.

Branded search answers demand; it rarely creates it. Keep it where rivals would otherwise take the top of the page, report it apart from non-brand, test it on a schedule, and judge whether paid search finds customers on the non-brand line alone.