As all agents, brokers, and home buyers know, searching for a home is a deeply personal process, and one of the most difficult challenges for buyers is narrowing down what they want. When a prospective buyer walks through a home or searches for one online, they are making hundreds of value judgments, often without ever consciously realizing them or expressing them to the real estate professional they are working with. Thankfully, artificial intelligence (AI) can now help bridge that gap and deliver a customized and personalized experience for consumers, without additional work by the agent or broker. For years, it has been easy to search for homes based on basic criteria like square footage, but what if a client wants something a little more specific, such as hardwood floors in all of the bedrooms, or homes with granite counters and white kitchen cabinets? That's where AI comes in.

We propose Style Conditioned Recommendations (SCR) and introduce style injection as a method to diversify recommendations. We use Conditional Variational Autoencoder (CVAE) architecture, where both the encoder and decoder are conditioned on a user profile learned from item content data. This allows us to apply style transfer methodologies to the task of recommendations, which we refer to as injection. To enable style injection, user profiles are learned to be interpretable such that they express users' propensities for specific predefined styles. These are learned via label-propagation from a dataset of item content, with limited labeled points. To perform injection, the condition on the encoder is learned while the condition on the decoder is selected per explicit feedback. Explicit feedback can be taken either from a user's response to a style or interest quiz, or from item ratings. In the absence of explicit feedback, the condition at the encoder is applied to the decoder. We show a 12% improvement on NDCG@20 over the traditional VAE based approach and an average 22% improvement on AUC across all classes for predicting user style profiles against our best performing baseline. After injecting styles we compare the user style profile to the style of the recommendations and show that injected styles have an average +133% increase in presence. Our results show that style injection is a powerful method to diversify recommendations while maintaining personal relevance. Our main contribution is an application of a semi-supervised approach that extends item labels to interpretable user profiles.

Like many online retailers, Overstock closely tracks the behavior of visitors on its site. But transforming all those individual page views and clicks into actual revenue is easier said than done. The company recently discussed with Datanami how it overcame challenges in building its own one-to-one marketing analytics system, and what results it's delivered this year. Overstock emerged from the wreckage of the first dot-com boom with a winning business plan: sell the excess merchandise of failed retail outfits at below-wholesale levels, build a loyal following, rinse, and repeat. Now nearly two decades in, Overstock has expanded to sell new products, and it all adds up to nearly $2 billion in revenue annually.