In this project, I investigate temporal externalities of attention on TV Ads. I use a dataset from TVision, which is unique in that it contains both 'traditional' features (viewing histories, user and item characteristics), as well as the attention ratio of viewers on Ads. I find that viewers' attention is persistent over a short period of time. I then apply the knowledge to incorporate temporal features such as previous Ad and previous attention to the traditional latent factor techniques, resulting highly effective advertising recommendations.