We took a promising new method of political polling - snap judgements of political candidates' facial appearance - from the lab to the real world within ternet-enabled mobile phones. Using iPhones and online multimedia-rich surveys, we collected over 6000 snap judgements of political candidates' faces,providing proof of concept for a new method of candidate pre-testing and political polling. Consistent with prior research, we find that snap judgements by small samples (178) of politically naive respondents can accurately predictelection outcomes. Further, we advance this method of research by testing designelements and providing practical details about the use of mobile technology to aid data collection.