Invited Speakers  

Nan Wang

Data Science Lead, Databricks

Bio: Nan is a Data Science Lead at Databricks, where she spearheads a variety of high-priority projects across multiple domains, including Growth and Go-To-Market (GTM) strategies. Prior to her role at Databricks, Nan was a tech lead at LinkedIn, where she played a key role in driving user growth and engagement as founding member of the growth team. She later established the Growth Marketing Data Science team, developing state-of-the-art data products, predictive models, and causal measurement frameworks which supported million dollar marketing investments.

Title: Leveraging Machine Intelligence and Domain Expertise to Understand SaaS Consumption Patterns

Talk Abstract: With the shift from subscription-based to consumption-based business models in the Software as a Service (SaaS) industry, companies face the increasing challenge of understanding customer consumption beyond mere product usage. This is particularly complex among enterprises with multiple product offerings. This presentation introduces a novel approach integrating machine intelligence with deep domain expertise to map consumption data directly to customers' objectives. By analyzing how customers utilize Databricks in alignment with their business goals, this methodology not only aids Databricks in enhancing problem-solving support and growth strategies for customers but also empowers customers with critical insights into cost management. The presentation will include a discussion of the techniques employed, a case study on the application of this method, an explanation of our validation approach, and the measurable outcomes that highlight the value of this initiative.

Peter Coles

Head Economist of Policy, Airbnb

Bio: Peter Coles is the Head Economist for Policy at Airbnb. Previously he was Head Economist and Director of Global Strategy at eBay, and before that a professor at Harvard Business School. Peter's expertise is in the fields of Market Design, Strategy, and Public Policy. His research focuses on reducing frictions in marketplaces, and he helped design a "Signaling Mechanism" used by the American Economic Association to match students with jobs.  At Airbnb, his work pertains to understanding Airbnb's relationship to cities and the broader world. He is passionate about raising oysters, eating blue crabs, and putting Market Design principles into practice!

Mike Egesdal

Staff Data Scientist, Airbnb

Bio: Mike has worked as a data scientist at Airbnb for the past eight years, where his work has focused on the measurement and modeling of quality on the platform. His recent initiatives have involved estimating the long run impact of quality outcomes, developing model-based merchandising to better highlight great Airbnb listings, and improving the signals collected from reviews. Prior to Airbnb, he completed a PhD in Economics at Harvard specializing in industrial organization and econometrics.

Talk Abstract: It's often hard to differentiate the quality of customer experiences using simple review ratings, in part due to the tightness of their distribution. In this talk, we present an alternative notion of quality based on customer revealed preference: did a customer return to use the platform again after their experience? We describe how a metric – Guest Return Propensity (GRP)— leverages this concept and can differentiate quality, capture platform externalities, and predict future returns.

In practice, this measure may not be suited to many common business use cases due to its lagging nature and an inability to easily explain why it has changed. We describe a quality measurement system that builds on the conceptual foundation of GRP by modeling it as an outcome of upstream realized quality signals. These signals—from sources like reviews and customer support—are weighted by their impact on return propensity and mapped to a quality taxonomy to aid in explainability. The resulting score is capable of finely differentiating the quality of customer experiences, aiding tradeoff decisions, and providing timely insights.

Edwin Ng

Senior Applied Scientist, Amazon

Bio: Edwin Ng is the Applied Science Lead at Amazon for the Grocery Economics & Optimization Science (GEOS) Forecasting team. His team specializes in scalable forecasting and causal analysis to enhance inventory optimization and inform strategic decision-making. Prior to joining Amazon, Edwin was an Applied Science Manager at Uber, where his team developed financial forecasting, marketing measurement, and optimization solutions. He is the primary author of the open-source project "Orbit," a Bayesian time-series modeling package, and the Bayesian Time-Varying Coefficients (BTVC) model, which was published at AdKDD in 2021.

Title: Forecasting Grocery Demand with Bayesian State-Space Model and Causal Study

Talk Abstract: Accurate demand forecasting plays a pivotal role in optimizing Amazon's grocery inventory and supply chains. This talk will delve into the background and challenges currently faced in this area, setting the stage for our proposed solution. We will introduce Bayesian State Space Models (BSSM) as a powerful tool for addressing these challenges. BSSM stands out for its ability to incorporate prior knowledge, manage non-stationarity, and handle multivariate datasets, all while providing probabilistic forecasts. The talk will cover the technical aspects of our approach, including the model structure, sampling techniques, and implementation details. We will also outline the next steps for advancing this work and improving demand forecasting accuracy.