Brand search ads are often there for defense: when someone searches for your app, you don’t want a competitor’s ad sitting above your own listing. In Google App Campaigns, brand traffic also comes in by default, since the campaign serves on brand searches alongside everything else unless you exclude them. That makes it hard to see what the brand traffic actually adds, and it can quietly lift the ROAS that Google Ads reports.
For a productivity-focused mobile app, we wanted to test it properly: are brand search ads in App Campaigns incremental, or are we paying for installs that would have happened organically? If brand ads were really driving installs, excluding them should hurt revenue and ROAS. To find out, we ran a controlled A/B test using stratified sampling and difference-in-difference analysis.
Why we ran this experiment
Google Ads targets brand terms aggressively, and many of the users searching for the brand would install with or without an ad, so the campaign can end up taking credit for organic installs. The goal was to find out whether the brand traffic actually increased total revenue or just shifted attribution.
How we designed the experiment
Step 1: A fair comparison with stratified sampling
Simply splitting US states at random would not work, because revenue varies a lot between regions. Instead we used unsupervised learning to cluster the states into five groups based on daily revenue, then randomly assigned states from each cluster to either the control or the treatment group. This technique, known as stratified sampling, keeps the two groups comparable and reduces the variability between them.
Step 2: Checking the parallel trends assumption
Before the test started, we checked whether both groups followed a parallel trend in revenue. They did, so the parallel trends assumption passed: any difference after the brand exclusion would come from the experiment, not from trends that were already there.
Step 3: Structuring the campaigns for testing
First, we removed brand terms from the main US campaign, excluding all brand-related traffic from the primary Google App Campaign.
Then we created a separate brand campaign. It only included text assets, so that it mainly captured brand searches, and had a very high tROAS target to keep the bidding from getting aggressive. In half of the US regions (the treatment group), brand ads were excluded.
The two groups still had slight differences in revenue before the test, so we used difference-in-difference (DiD) analysis to remove that bias. The method adjusts for pre-existing differences, which gives a more accurate causal read.
The key findings
1. Revenue was slightly higher without brand ads, but not significantly
The treatment group, with no brand ads, had 10.63% higher revenue with outliers and 7.89% higher revenue without outliers. The difference was not statistically significant, though, so we cannot conclude that excluding brand caused the increase.
There was also an interesting change in dynamics: the brand-excluded group spent more overall on the main non-brand UAC campaign, which inflated its revenue. When we ran a regression analysis, we did not find that the non-brand spend increased because of the brand exclusion.
2. Blended ROAS was practically the same
After the difference-in-difference analysis, the effect of excluding brand on ROAS was 1.4%, which is negligible. We could not reject the null hypothesis: brand ads had no incremental impact on ROAS.
3. Total revenue was not affected
Despite turning off brand ads in half of the US, overall revenue stayed stable, as a regression analysis confirmed. This suggests that users searching for the app’s brand terms most likely installed organically when no ad was shown.
What this means for advertisers using UAC
Watch blended ROAS, not just what Google Ads reports. After we excluded brand, the ROAS reported in Google Ads dropped significantly for the campaign, but blended ROAS barely moved. The brand traffic had been making the campaign look better than it was.
Brand search ads in App Campaigns may not be necessary, especially when organic demand and the brand are strong. That makes it worth testing a brand exclusion rather than assuming brand ads drive incremental installs.
App rating probably matters here. The lower your rating, and the more your main competitors show up on your brand searches, the more likely you need to keep bidding on your brand, since a very low rating can put users off. In that case, having both your ad and your organic result take up space and push competitors down can help.
Every app and situation is different, but in this case, excluding brand ads did not affect ROAS.
Why run the test yourself instead of in the platform
Rigorous methods such as stratified sampling, parallel trends checks and difference-in-difference analysis are essential to avoid misleading conclusions, and so is running the test independently. A/B tests are how you make decisions based on causal inference, but research by Braun and Schwartz shows that tests run through the ad platforms are biased by divergent delivery: the algorithm decides which ad is relevant to which type of user, so the control and treatment groups are no longer random or comparable. That is why we chose the method based on the context and ran the experiment ourselves.