Key points are not available for this paper at this time.
User profile is a summary of a consumer’s interests and preferences revealed through the consumer’s online activity. It is a fundamental component of numerous applications in digital marketing. McKinsey & Company view online user profiling as one of the promising opportunities companies should take advantage of to unlock “big data’s” potential. This paper proposes a modeling approach that uncovers individual user profiles from online surfing data and allows online businesses to make profile predictions when limited information is available. The approach is easily parallelized and scales well for processing massive records of user online activity. We demonstrate application of our approach to customer-base analysis and display advertising. Our empirical analysis uncovers easy-to-interpret behavior profiles and describes the distribution of such profiles. Furthermore, it reveals that even for information-rich online firms profile inference that is based solely on their internal data may produce biased results. We find that although search engines cover smaller portions of consumer Web visits than major advertising networks, their data is of higher quality. Thus, even with the smaller information set, search engines can effectively recover consumer behavioral profiles. We also show that temporal limitations imposed on individual-level tracking abilities are likely to have a differential impact across major online businesses, and that our approach is particularly effective for temporally limited data. Using economic simulation we demonstrate potential gains the proposed model may offer a firm if used in individual-level targeting of display ads. Data, as supplemental material, are available at http://dx.doi.org/10.1287/mksc.2015.0956 .
Trusov et al. (Thu,) studied this question.