We investigate the moderating effect of product attributes and review ratings on {views, conversion|views (conversion conditional on views), final conversion} of a purchase-based collaborative filtering recommender system on an e-commerce site. We run a randomized field experiment on a top...

Instrumental variables (IVs) are a commonly used technique for causal inference from observational data. In practice, the variation induced by IVs can be limited, which yields imprecise or biased estimates of causal effects and renders the approach ineffective for policy decisions. We confront...

We investigate the impact of collaborative filtering recommender algorithms (e.g., Amazon’s “Customers who bought this item also bought”) commonly used in e-commerce on sales diversity. We use data from a randomized field experiment run on the website of a top retailer in North America...

We propose an interpretable deep survival model that can capture human-understandable nonlinear patterns from big data while handling censored observations and time-varying customer dynamics. To this end, we build WTTE-TCN (Weibull Time to Event Temporal Convolutional Networks) and apply...

"Brushing"---online merchants placing fake orders of their own products---has been a growing phenomenon on major e-commerce platforms. On the one hand, brushing enables merchants to boost their rankings in search results, because products with higher sales volume are often ranked higher. On the...

Traditional online platforms (e.g., Amazon Marketplace) use Unilateral Rating System (URS), where customers rate sellers. However, sharing economy platforms (e.g., Uber, Airbnb) have adopted Bilateral Rating System (BRS) that also allows service providers to rate customers, and even selects...

We describe the effect of social media advertising content on customer engagement using data from Facebook. We content-code 106,316 Facebook messages across 782 companies, using a combination of Amazon Mechanical Turk and Natural Language Processing algorithms. We use this dataset to study the...

Recommender systems typically work over sparse matrices. Although most methods assume so, these matrices' entries are often not missing at random (NMAR). How problematic is this? We present a puzzle. Some methods explicitly account for NMAR processes. This has been shown to improve predictions....