Invited Speakers

Associate Professor at Emory University

Title: Promising Delivery Speed in Online Retail

Bio: Ruomeng Cui is an Associate Professor of Operations Management at the Goizueta Business School, Emory University (on leave). She currently is a full-time Amazon Visiting Academic at Amazon, working in the supply chain domain. Her research focuses on causal inference, machine learning and data-driven modeling, with applications in retail, supply chains, and platforms. She currently serves as an associate editor for Manufacturing & Service Operations Management and Production and Operations Management. She received her Ph.D. in Operations Management from the Kellogg School of Management, Northwestern University and B.Sc in Industrial Engineering from Tsinghua University.

Talks Abstract: Online retailers have to provide customers with an estimate of how fast an order can be delivered before they decide to make the purchase. Retailers can strategically adjust this delivery speed promise, and doing so may fundamentally impact customer journey. It can influence consumers' purchasing decisions and post-purchase experiences, often in the opposite direction. On one hand, an aggressive (i.e., faster) delivery estimate could ensure that more customers meet their deadlines and thus may increase their purchases ex ante. On the other hand, an aggressive estimate tends to overpromise, potentially leading to a longer than expected wait time, which can lower customer satisfaction and increase product returns ex post. In this research, we estimate the causal effect of retailers' delivery speed promise on customer behaviors and business performance. Collaborating with Collage.com, an online retailer that sells customized photo products across the US, we exogenously varied the disclosed delivery speed estimates online while keeping the physical delivery speed unchanged. Using the difference-in-differences identification, we find that a faster promise increases sales and profits, but it also increases product returns and reduces customer retention. In addition, we propose a data-driven optimization model with the estimated parameters as inputs to optimize delivery promises to maximize customer lifetime value. Our findings provide managerial insights and a data-driven policy that retailers can leverage to optimize and customize their delivery promises.

Sean O'Donnell

Jason Cai

Linsha Chen

Sean O’Donnell (Main Speaker)

Senior Data Scientist at Airbnb

Co-Author / Co-Speaker:

Jason Cai, Data Science Manager at Airbnb;
Linsha Chen, Senior Data Science Manager at Airbnb;

Title: Devoted to Long-Term Adventure: Growing Airbnb Through Measuring Customer Lifetime Value

Bio:

Sean is a senior data scientist at Airbnb. He has worked across many marketing initiatives, including email and push notification marketing, coupons, bot detection, attribution and measurement systems, and brand marketing. He is now focused on measuring the long-term value of Airbnb’s customer base, through CLV and incremental booking value modeling.
Jason is the Data Science Manager at Airbnb Marketing Technology. He leads a Data Science and Machine Learning team to drive sustainable growth and engagement for the Airbnb users through multiple marketing channels (Email, Push, Incentives, Referrals, Virality, etc) by leveraging Advanced Analytics, Causal Inference and Machine Learning models. He has rich experience in Forecasting, Targeting, Personalized Recommendations, Attribution; and has led multiple critical ML efforts in Airbnb, Microsoft and Amazon such as Airbnb Guest-to-Host propensity model, Xbox category recommendation model, Tree ensemble explainability system, etc.
Linsha is the Head of Growth & Marketing Technology Data Science at Airbnb. Her team builds state-of-the-art data products, models, and measurement capabilities to support $XB of marketing investment for both Guest and Host growth. The work of her team empowers Airbnb to optimize marketing and products that amplify the brand, and grow and engage with Airbnb’s Guest and Host communities throughout the lifecycle. Outside of work, Linsha is a mom of two lovely boys, and an active champion of women in tech.

Talks Abstract: Airbnb is a two-sided travel platform for homestays and experiences, and understanding the long-term value of our customers, both Guests and Hosts, is at the core of Airbnb’s product. It establishes the strategies to acquire users who resonate with our brand, improve product offerings to maximize value to users over time, and optimize our marketing to sustainably grow our platform while retaining our Guest and Host community. In our talk, we discuss how the nature of Airbnb's business, a two-sided marketplace where users can inhabit overlapping roles as Guest and Host, presents a unique challenge to measuring the long-term value of our users. Then, we describe how a blend of ML and statistical inference, along with marketplace dynamics and guest-to-host state modeling, enables us to perform customer lifetime value (CLV) modeling specific to our users. Finally, we explore further, targeted applications built in conjunction with CLV modeling to drive ad bidding strategy, cross-channel targeting, segmentation, and user understanding that will enable us to tailor our marketing and product experience.

Sean Taylor

Chief Scientist at Motif Analytics

Title: Discovering causal opportunities through deep learning models of event sequence data

Bio: Sean J. Taylor is co-founder and chief scientist at Motif Analytics. Previously he was head of Lyft's Rideshare Labs and spent seven years as a research scientist on Facebook's Core Data Science team. Sean's work is at the intersection of machine learning and causal inference, with a focus on applied problems and generating business value using the latest methods. He earned his PhD in Information Systems from NYU’s Stern School of Business as well as a BS in Economics from Wharton School.

Talks Abstract: The most common application of causal inference focuses on a hypothesized cause and estimation of its effect on some outcome of interest. Methods for this class of problem do not address the more common business problem of *finding opportunities*: currently unknown cause-effect relationships which could be used to improve customer experience and business outcomes. We propose an effective method for generating and ranking hypotheses about which events are potential causes for subsequent behavior of customers. Adapting the transformer architecture from text to abundant user-level event sequence data, we learn a latent representation of customer journeys, then fine-tune these representations to predict subsequent business outcomes such as signup, retention, and revenue. We cast this prediction problem in a causal inference framework, derive efficient estimators for bounds on causal effects of events, and show how they can be applied to surfacing promising opportunities for process improvement. Our method is highly general and shows promising results on real log data from a consumer application, as validated by domain experts.

Shant Torosean

Analytics Engineering Manager at Airbnb

Title: Unveiling the Guest & Host Journey: Session-Based Instrumentation on Airbnb Platform

Bio: Shant is an Analytics Engineering Manager at Airbnb. He leads a team of engineers that focus on Guest side data, event logging system, Booking system and the peripheral services. With his background in causal inference, Shant brings a unique lens of effectively bridging the gap between data consumers and producers by informing data model and event logging design. Shant currently resides in the Bay Area with his wife and two sons. When not at work, he enjoys spending quality time with family gardening and going on long hikes in Yosemite.

Talks Abstract: Airbnb has become a go-to platform for travelers seeking unique and memorable stays and experiences. Understanding guest and host interaction with our platform is a foundational step in improving the overall user’s product experience. In this talk, we’ll explore the power of a novel session-based client instrumentation to capture guest and host behavior without jeopardizing performance and downstream data consistency. By capturing user and system behavior using sessionization, we can gain new insights into how guests & hosts interact with the platform, how they visit the platform, what motivates their booking decisions and where they encounter pain points and obstacles. We’ll delve into real-world examples of how this type of instrumentation has helped data teams answer complex questions around user paths, usage patterns and attribution in a scalable way.