A marketer wants to know which of two different video lengths drives a stronger CPA for their campaign. To do this, what should they set up?

A Demand Gen Experiment is the correct choice because it allows marketers to test different campaign elements, such as video lengths, against each other to determine which performs better for a specific metric like CPA. This type of experiment is designed for A/B testing and directly measures the impact of changes on performance. A Search Lift study measures the incremental impact of video ads on search queries and conversions, not direct CPA comparison of video lengths. A new ad group would simply run both video lengths simultaneously without a controlled testing framework to attribute performance differences. A brand survey measures brand perception and recall, not CPA.