This paper introduces several new methods aimed at improving ad testing. The power of interactive television and scientific statistical techniques are leveraged to create a unified methodology to test ads. The methodology makes three specific improvements, including the utilization of a random probability sample, the development of a system that allows for self-administered ad experiments given in a respondents own home, and the employment of a fully randomized, panel experimental research design. The paper draws on empirical evidence from the Yale Advertising Study, focusing on measuring the effect of political ads on voter attitudes and behavior. The study included 12,350 interviews.