Systems and methods of determining effectiveness of a test content item in a computer network environment are provided. A data processing system can receive a request for content from a computing device and predict a test content item as a winner of a content item auction. The data processing system can assign the computing device to a test group or a control group. The data processing system can execute the content item auction, while withholding the test content item from the auction if the computing device is assigned to the control group. The data processing system can transmit the winning content to the computing device. The data processing system can identify computer network activity associated with the test content item by the computing device. The data processing system can perform additional predictions if the computing device is the subject of additional experiments regarding additional test content items.