Attention Thresholds for Brand Impact

Brand lift studies are a staple in the media industry for estimating the impact of advertising on consumer attitudes and preferences. They measure differences in multiple metrics between consumers who have been exposed to an ad and those who have not. One of the contributing factors to the impact of an ad is the amount of visual attention that the consumer gives to the creative, which was historically difficult to distinguish from exposure alone. In this work we develop a method for processing brand study data that contains both attention measurement and survey-based brand outcomes. This method focuses on exploiting inherent structural mechanisms in the nature of what upper funnel metrics measure in order better estimate the threshold of attention at which an impact occurs. We benchmark the approach against multiple alternative methods for estimating these thresholds and demonstrate superior performance in terms of mean error and width of the error bounds.