Key takeaways
- Branded search is partly incremental, and usually far less than the ad platform reports. In my geo holdout test at Shopify it came back about 35% incremental.
- That number belongs to Shopify. The share varies by company, so test your own branded search campaign before you borrow mine.
- With enough volume, run a geo holdout test: pause brand in matched markets for about eight weeks and keep non-brand in a separate cell.
- With a small account, pause the brand campaign for about three weeks and count total leads from every source. At Pipe17 the total barely moved.
01Is branded search incremental?
Branded search is partly incremental, and usually much less than the ad platform says. In the geo holdout test I ran at Shopify, the branded search campaign came back about 35% incremental, which means roughly 65% of the conversions it reported would have arrived through organic anyway.
The reason is who's searching. Someone who types your company name has already decided to visit you. The ad sits above your own organic result, takes the click and takes the credit. Inside the platform that reads as the cheapest, best-converting line in the account, and at Shopify it did before we tested it.
What incrementality means, the lift formula and how to size a test are in my note on incrementality testing. This note covers brand only.
02Two ways to test a branded search campaign
| Geo holdout test | Pause and count | |
|---|---|---|
| What you do | Pause brand in some matched markets and keep it running in the rest | Pause the whole brand campaign and compare the same number of days before and after |
| Use it when | Each market has enough conversions to show a difference | The account is too small to split by region |
| How long I ran it | About eight weeks at Shopify | 22 days each side at Pipe17 |
| What you read | Total conversions in holdout markets against test markets | Total leads from every source, before against after |
| Weak spot | Needs matched markets and someone who can do the statistics | Anything else that changed in the window, season included |
The geo holdout test
- Pick matched markets.We chose comparable mid-size US DMAs and split them into a test group and a holdout group.
- Pause brand entirely in the holdout.Branded search went fully dark in those markets for the length of the test.
- Give non-brand its own cell.Non-brand was tested separately so the two reads didn't mix.
- Run it to significance.For us that took about eight weeks.
- Read the total.The question is whether the holdout markets lost signups overall once organic picked up those searches. The brand campaign's own conversion count can't tell you that.
The pause-and-count test
- Pick equal windows.I used the 22 days before the pause and the 22 days after.
- Pause the brand campaign.Leave the rest of the account alone so you're changing one thing.
- Count leads by source in the CRM.Leads with a Google click ID, organic contacts and direct contacts, for both windows.
- Compare the totals.If the total holds while paid leads dip and organic and direct rise, most of the brand spend was buying people who were already coming.
03The Shopify example: about 35% incremental
At Shopify I ran paid merchant acquisition, $20M+ a year across six channels. I ran the search holdout with the Data Science team.
- Branded search came back about 35% incremental.
- Non-brand search came back about 70% incremental.
- About $4M a year moved out of brand and into non-brand search and prospecting. Part of it paid for tests in connected TV and radio.
- Blended incremental CAC improved by an estimated 15%.
At 35% incremental, each merchant that brand really added cost far more than the dashboard showed. The dashboard was reporting what it was built to report. It couldn't say which merchants would have signed up without the ad.
04The Pipe17 example: 22 days with brand off
Pipe17 sells an OrderOps platform to commerce brands and 3PLs, with deals that take six months or more. There was nowhere near enough volume for a geo split, and about half the Google budget was going to our own brand name. So I paused the brand campaign and counted.
| Source | 22 days before | 22 days after |
|---|---|---|
| Leads with a Google click ID | 9 | 7 |
| Non-brand leads | About 1 | 4 |
| Organic contacts | 13 | 15 |
| Direct contacts | 16 | 19 |
Total lead flow barely moved. It's a small sample and I'd call it early evidence, but it was enough to act on. The account ended up running 100% non-brand.
A second check pointed the same way. When I read the click paths, 14 of 15 paid-search sales-qualified opportunities had first arrived by searching the company's own name. They'd heard the name somewhere else. Form answers showed some of them had heard it from an AI assistant, and the brand ad had been collecting the credit.
05What I got wrong before I knew how to test
At Purolator I spent heavily on branded search and never tested whether it was incremental. I didn't know how yet. If I ran that account again, I'd test brand before defending that line in a budget meeting.
06Mistakes to avoid
- Judging brand on its own ROAS. It will always look good. Read total conversions with brand on and off.
- Mixing brand and non-brand in one cell. They came back at about 35% and about 70% for us. One blended number would have hidden both.
- Calling a small test final. Twenty-two days in a B2B account is a signal. Run it again before you treat it as settled.
- Letting several buyers bid on your name. At Dental Team FL, five agencies each bid on the group's own brand name for their own office. That pushed up the price of clicks from people who were already looking for the ten practices.
Testing brand is one of the first things I do in a Google Ads audit, and moving the budget it frees up is the work I do as a performance marketing consultant.
FAQQuestions people ask
- Is branded search incremental?
- Partly, and usually less than the ad platform reports. In a geo holdout test Farhan Munshi ran with Data Science at Shopify, branded search came back about 35% incremental, so roughly 65% of the conversions it claimed would have arrived through organic anyway. Non-brand search came back about 70% incremental. The share differs by company, so test your own account.
- How do you run a geo holdout test on a branded search campaign?
- Pick matched markets, pause the branded search campaign entirely in the holdout group, keep non-brand in its own separate cell and run until the difference is statistically significant. At Shopify, Farhan Munshi used comparable mid-size US markets and ran the test for about eight weeks. Read total conversions in each group, organic included, and ignore what the campaign reports about itself.
- Should I pause my branded search campaign to test it?
- If the account is too small for a geo holdout, yes. At Pipe17, Farhan Munshi paused the brand campaign and compared the 22 days after with the 22 days before. Leads with a Google click ID went from 9 to 7, while organic and direct contacts both rose. Total lead flow barely moved. It's a small sample, so treat a result like that as early evidence.
- How long should a branded search incrementality test run?
- Long enough for a clear read at your volume. Farhan Munshi's geo holdout at Shopify ran for about eight weeks to reach statistical significance. His pause test at Pipe17 compared 22 days before with 22 days after. That's enough to see whether total leads hold, and too short to call final in a B2B account with a six-month sales cycle.
- Why does branded search look so efficient in Google Ads?
- Because the people searching your name were already on their way. The ad takes the last click and the platform reports the conversion. Before Farhan Munshi tested it at Shopify, branded search looked like the most efficient line in the account. The geo holdout showed about 65% of those conversions would have arrived through organic, and about $4M a year moved out of brand.
