ORIE Colloquium: Divya Singhvi '15 (MIT) - From Little to Big Data: The Impact of Analytics in Operations Management

Description

Operations management is going through a transformational change where data is driving operational decisions. Nevertheless, firms inevitably find themselves in one of two settings: (i) where there is little to no data to drive decisions, (ii) where there is existing data to leverage and optimize decision making. In this talk, we discuss two problems related to each of these two settings. In the first part of the talk, we consider the problem of sequential product recommendations when customer preferences are unknown; i.e. very little prior data is available to estimate the customer’s preference. First, we present empirical evidence of customer disengagement using a sequence of ad campaigns from a major airline carrier. In particular, customers decide to stay on the platform based on the relevance of the recommendations. We then formulate this problem as a linear bandit problem, with the notable difference that the customer's horizon length is a function of past recommendations. We prove that any algorithm in this setting achieves linear regret. As a result, no algorithm can keep all customers engaged; regardless, we can hope to keep a subset of customers engaged. Unfortunately, we find that classical bandit learning as well as greedy algorithms provably over-explore, thereby incurring linear regret for every customer. As a result, we modify bandit learning strategies by constraining the action space upfront using integer optimization. We prove that this modification allows our algorithm to achieve sublinear regret for a significant fraction of customers. Furthermore, numerical experiments on real movie recommendations data demonstrate that our algorithm can improve customer engagement with the platform by up to 80%. In the second part of the talk, we consider the setting where the firm has access to data that can be used to inform and optimize decisions. In particular, we discuss the problem of reducing product returns in online retail. Working with one of the largest fashion retailers of India, and leveraging their rich data sets, we perform an extensive observational data study, and an RCT to investigate the causal relationship of delivery speed and delivery promise on product returns. We estimate cost savings of as much as $1.5 million per year on account of reduced returns, due to faster deliveries. We then propose a data-driven delivery expediting framework that optimally balances delivery expediting costs with reverse logistics costs to minimize the retailer’s overall costs. Bio: Divya Singhvi is a fifth-year Ph.D. student at the MIT Operations Research Center where he is being advised by Prof. Georgia Perakis. His research interests lie at the intersection of machine learning and operations management. Particularly, he has worked on problems related to optimal demand learning, pricing, recommendations and logistics in online and offline retail operations. His research has been tested with companies such as J&J, Myntra, Wayfair, IBM, and CitiBike, among others. Divya has interned at IBM and Ascend Analytics. Prior to attending MIT, Divya received a B.S. (2015) in operations research and engineering from Cornell University.