Measuring Brand Ad Effectiveness (303176)

Keywords: Causal modeling, bootstrap, randomized experiments, bias correction, online video ads, online surveys, Google searches

Display and video ads proliferate on the web, but are they effective? Or are they irrelevant in light of all the other advertising that people see? We describe a number of methods used at Google to estimate how effective these ads are through randomized experiments, where the treatment group sees ads from the ad campaign and the control group sees alternative ads. We measure brand awareness, ad recall and consideration through online surveys on YouTube, and we measure brand interest using searches on google.com for keywords relevant to the advertiser. Along with estimating overall ad effectiveness, our methods can also quantify effectiveness for slices of the population such as age groups and geographies.