Holdback Experiment

What Is a Holdback Experiment?

Many advertisers overestimate the effect of their ads because some conversions would have happened without advertising anyway. Holdback experiments deliver the answer to the decisive question: how many additional conversions did your campaign actually generate? This insight protects you from budget misallocation and is the basis for evidence-based decisions in performance marketing.

A holdback experiment is a controlled test procedure in Google Ads in which a defined percentage of your target audience deliberately does not see any ads. This control group serves as a baseline to measure how many conversions were actually caused by advertising and how many would have occurred without ads.

The principle comes from scientific research and is also called a Randomized Controlled Trial (RCT). Google randomly divides users into a test group (sees ads) and a control group (sees no ads). By comparing the conversion rates of both groups, the incremental effect of your campaign can be calculated. This result is more meaningful than any attribution-based measurement because it measures causality rather than correlation.

In practice, holdback experiments are especially useful for evaluating brand campaigns and remarketing. Particularly with brand bidding, the question often arises whether users would not have come to your website even without the ad. A holdback experiment provides the answer. Plan a test period of at least two to four weeks and keep the control group at 10–20 percent to achieve meaningful results with acceptable revenue loss.