Presentation by Oliver J. Rutz PhD

Professor Daniel Wentzel, Foster School of Business, Wednesday, 19.12. 12:15-13:30, Seminar Room 1.202

Understanding Text Ads ? The Ad Genome Approach

Abstract

Text ads, traditionally mostly used in classified advertising, have become very popular with the rise of search engine marketing and content networks. These ads are unlike typical ads, as they only contain text and do not use graphical elements, e.g., color or pictures, to convey a message to the consumer. We propose a novel approach to gauge the performance of text ads and to allow identification of good ads. Our approach uses pairwise comparisons on an outcome dimension of interest (e.g., click-through) as well as dimensions of the ad itself which we call genes (e.g. the ad?s ?call to action?).  We use the pairwise comparisons to generate rankings on these dimensions. As a first step we investigate whether the ad genes can explain differences in the outcome of interest. In a second step, we use text mining to generate markers that inform the ad genes. We allow for direct effects as well as interactions in these text markers, resulting in a ?large p, small n? problem. We address this problem with a novel penalized regression model fully estimated in a Bayesian framework. Applying our model to three datasets of ads, we show that the ad genes help explain ad performance. Using text markers, we show that our approach ably forecasts the performance of an ad in terms of its rank compared to other, existing ads. 

Mehr Informationen über den Kollegen Rutz finden Sie unter www.foster.washington.edu/centers/facultyresearch/facultyprofiles/Lists/Faculty%20Contact%20Info/DispProfile.aspx.