Google characterizes geo-based experiments as using non-overlapping geographic regions, which are randomly assigned to either a control or treatment group.

• Geo experiments use overlapping geographic regions that are assigned to a control or treatment group.
• Geo experiments use random geographic regions assigned to test group.
• Geo experiments use random geographic regions assigned to control group.
• Geo experiments use non-overlapping geographic regions that are randomly assigned to a control or treatment group.

The correct answer is: Geo experiments use non-overlapping geographic regions that are randomly assigned to a control or treatment group.

Explanation: Geo experiments, often referred to as GeoX, utilize non-overlapping geographic regions, like city or postal codes, designated randomly as control or treatment groups. This distinction allows for the exclusive attribution of any rise in success metrics between the groups to the ad campaign. With GeoX, businesses can monitor the efficacy of ad alterations by contrasting them with a control. Upon gathering data, offline conversion metrics such as store sales or visits can be analyzed, assigning the observed uplift directly to ad spend.