Deep learning for discovering new product opportunities

Abstract

Modeling customer preferences over multivariate attributes has long been an endeavor of marketing research. We provide an updated approach which can learn customer preferences for complex products with multiple multivariate attributes using modern Deep Neural Networks. In turn we outline approaches for framing managerial questions in the form of inference problems. With our empirical application to product identification in credit cards, we conclude Deep Learning results in significantly better performance than the state of the art. Our approach is scalable to Big Data and can derive superior predictive power from the inexpensive unstructured data exhaust of internet commerce.